ASSESSMENT OF CLIMATE EFFECTS ON SCOTS PINE (Pinus sylvestris L.) GROWTH IN ESTONIA

Transcription

1 ASSESSMENT OF CLIMATE EFFECTS ON SCOTS PINE (Pinus sylvestris L.) GROWTH IN ESTONIA KLIIMA MÕJU HINDAMINE HARILIKU MÄNNI (Pinus sylvestris L.) KASVULE EESTIS SANDRA METSLAID A Thesis for applying for the degree of Doctor of Philosophy in Forestry Väitekiri filosoofiadoktori kraadi taotlemiseks metsanduse erialal Tartu 2017

2 Eesti Maaülikooli doktoritööd Doctoral Theses of the Estonian University of Life Sciences

3

4 ASSESSMENT OF CLIMATE EFFECTS ON SCOTS PINE (Pinus sylvestris L.) GROWTH IN ESTONIA KLIIMA MÕJU HINDAMINE HARILIKU MÄNNI (Pinus sylvestris L.) KASVULE EESTIS SANDRA METSLAID A Thesis for applying for the degree of Doctor of Philosophy in Forestry Väitekiri filosoofiadoktori kraadi taotlemiseks metsanduse erialal Tartu 2017

5 Institute of Forestry and Rural Engineering Estonian University of Life Sciences According to verdict No 2, of March 8, 2017, the Defence Board of PhD Theses in Forestry of the Estonian University of Life Sciences has accepted the thesis for the defence of the degree of Doctor of Philosophy in Forestry. Opponent: Supervisors: Prof. Gediminas Brazaitis, PhD Faculty of Forest Sciences and Ecology Aleksandras Stulginskis University Prof. Andres Kiviste, PhD Institute of Forestry and Rural Engineering Estonian University of Life Sciences Assoc. Prof. Ahto Kangur, PhD Institute of Forestry and Rural Engineering Estonian University of Life Sciences Defence of the thesis: Estonian University of Life Sciences, room 1B27, Kreutzwaldi 5, Tartu on June 9, 2017, at 10:00. The English and the Estonian language of the thesis were revised by Mrs. Karit Jäärats. Sandra Metslaid, 2017 ISSN ISBN (trükis) ISBN (pdf)

6 CONTENTS LIST OF ORIGINAL PUBLICATIONS... 7 ABBREVIATIONS INTRODUCTION REVIEW OF LITERATURE Shift towards empirical single tree growth models Effect of weather fluctuations on tree annual increment Tree-ring analysis to study annual climatic variability in tree growth Climate effects on site index Climate effects on radial growth Scots pine habitats and climatic conditions in Estonia Research needs AIMS OF THE STUDY MATERIALS AND METHODS Study sites Meteorological data Investigation of changes in height growth patterns (I) Dendroclimatic analysis (II, III, IV) Increment core sampling, preparation and measurements Standardization of tree-ring series Development of chronologies Assessment of chronology quality Analysis of growth-climate relationships Investigation of spatial and temporal response patterns (IV) Identification of pointer years (II) Modelling basal area increment (III) RESULTS Changes in long-term forest productivity Scots pine radial growth characteristics and relationship to inter-annual weather variations Scots pine chronologies Radial growth-climate relationships Temporal patterns of Scots pine response to climate variation Evidence of influences of extreme climatic conditions on radial growth... 55

7 5.3. Radial growth patterns and the BAI model for Scots pine on reclaimed areas DISCUSSION Tree growth deviations from expected growth Influence of annual climate variability on Scots pine growth in Estonia Modelling the basal area increment of Scots pine CONCLUSIONS REFERENCES SUMMARY IN ESTONIAN ACKNOWLEDGEMENTS ORIGINAL PUBLICATIONS CURRICULUM VITAE ELULOOKIRJELDUS LIST OF PUBLICATIONS

8 LIST OF ORIGINAL PUBLICATIONS The thesis is based on the following publications, referred to in the text by the corresponding Roman numerals. The papers are reproduced by the kind permission of the publishers. I II III Metslaid, S., Sims A., Kangur, A., Hordo, M., Jõgiste, K., Kiviste, A., Hari, P Growth patterns from different forest generations of Scots pine in Estonia. Journal of Forest Research, 16 (3): Hordo, M., Metslaid, S., Kiviste, A Response of Scots pine (Pinus sylvestris L.) radial growth to climate factors in Estonia. Baltic Forestry, 15 (2): Metslaid, S., Stanturf, J.A., Hordo, M., Korjus, H., Laarmann, D., Kiviste, A Growth responses of Scots pine to climatic factors on reclaimed oil shale mined land. Environmental Science and Pollution Research, 23: IV Metslaid, S., Hordo, M., Korjus, H., Kiviste, A., Kangur, A. 20xx. Spatio-temporal variability in Scots pine response to annual climate fluctuations in hemiboreal forests of Estonia. Submitted to Agricultural and Forest Meteorology. 7

9 The contributions from the authors to the articles are as follows: I II III IV Original idea AKa, SM MH AK, SM SM Study design AK, AKa, SM MH AK, HK, SM MH, AK Data collection AKa, MH, SM All All All Data analysis AK, AS, SM MH AK, SM SM Preparation of manuscript All All All All AK Andres Kiviste; AS Allan Sims, HK Henn Korjus; AKa Ahto Kangur; MH Maris Hordo, SM Sandra Metslaid; All all authors of the article. 8

10 ABBREVIATIONS AIC Akaike Information Criterion AC1 First-order autocorrelation ENFRP Estonian Network of Forest Research Plots EPS Expressed population signal ESTEA Estonian Environmental Agency BAI Basal area increment BCC Bootstrapped correlation coefficients HCA Hierarchical cluster analysis IC Mean interseries correlation ISL Islands JJp June July precipitation sum JJt June July temperature mean GAM Generalized additive model Glk Gleichläufigkeit MEE Mean estimation error MS Mean sensitivity NAO North Atlantic Oscillation NE Northeast PCA Principal components analysis PC1 First principal component SD Standard deviation SE Southeast SI x Site index at reference age x (SI 40 ) SNR Signal to noise ratio SW Southwest RMSE Root mean square error TBP T-values with Baillie Pilcher standardization TRW Tree-ring width 9

11 1. INTRODUCTION Forest growth and yield models are important tools in forest management planning (Hasenauer, 2006) for predicting short- and long-term growth, assessing management options and updating the information on available resources (Burkhart, 1990). As forest management decisions have a great impact on forest stand development, this considerably influences economic return, ecological conditions, and other ecosystem services (Duncker et al., 2012) of a managed area. Therefore, model predictions should be accurate and reliable enough to avoid wrong decisions and economic loss. Modellers and researchers continuously aim to build better models for forestry use (Eastaugh et al., 2013), to target questions that are more specific (Porté and Bartelink, 2002) and to improve the scope or scale of predictions. In addition, global climate change observed since the 1950s (Jones and Mann, 2004; IPCC, 2014) created the need for models capable of predicting growth dynamics with greater resolution and to take into account long-term changes in environmental conditions. In Estonia the mean annual air temperature over the second half of the 20 th century increased by C throughout the country (Jaagus, 1998; Jaagus, 2006a; Tarand et al., 2013). The greatest warming is observed for the spring season. For example air temperature in March over increased by 3 5 C, depending on the sub-region and weather conditions from cold and snowy, changed to mild and snow-free for that month (Jaagus, 2006a). Significant warming also took place in winter, especially in January and February (Jaagus, 2006a). However, the magnitude of the warming tendency varies across different parts of the country. Due to higher temperatures at the end of the cold season, the duration of the time period with snow cover has significantly decreased (Jaagus, 2006a, 2006b; Tarand et al., 2013) and the start of the growing period has shifted to earlier dates, as observed by the advanced spring and summer phenology of some plants (Ahas and Aasa, 2006). Besides that, temperature seasonality, defined as the difference between summer and winter mean temperatures, is also decreasing in this part of Europe, suggesting a latitudinal shift in climatic conditions equal to 4 7 N towards the equator (Xu et al., 2013). In addition to the increasing mean temperatures and altered precipitation regimes, rising CO 2 concentrations (Knorr, 2009; IPCC, 2014, BACC Author Team, 2015) and an increased deposition of nitrogen (Erisman and Vries, 2000) accompany 10

12 the climate change in the region. The warming tendency with an annual precipitation increase of 10 20%, mainly during the cold season is expected to continue in Estonia through the 21 st century (Jaagus and Mändla, 2014). The ongoing warming and processes associated with it force trees to respond to the environmental changes through physiological mechanisms, such like photosynthesis, transpiration, nutrient and water uptake (Morison and Lawlor, 1999; Niu et al., 2008) and consequently may have an impact on forest growth, productivity, survival and tree species distribution patterns (Lindner et al., 2010; 2014). In forestry, the concerns mainly emanate from how climate change will influence the forest growth in a certain region. It still remains uncertain, whether the impact will be positive or negative on forest productivity and tree vitality, and what economic, ecologic and social consequences it may cause for the countries. Uncertainties could be reduced and climate change effects on forest growth evaluated, based on future growth predictions, using models that incorporate tree response to climatic variations. Most of the forest growth models used in Estonia are stand-level growth models (Krigul, 1969; Tappo, 1982; Kiviste, 1999a, 1999b; Kasesalu, 2003; Nilson, 2005; Kiviste and Kiviste, 2009) developed based on an empirical approach. For a long time, these models were an appropriate instrument for defining stand growth over the time. However, due to shifts in environmental conditions and because of changing needs of forest managers such models become outdated. They are not applicable for predicting forest stand growth under various management alternatives and cannot be directly applied in heterogeneous stands (Teuffel et al., 2006). Besides that, most of the models developed during the last century are based on the assumption that site conditions are steady (Assmann, 1970), which relate the models to the time when they were developed. Thus, such models cannot account for the effects of the ongoing environmental changes on forest growth, most likely resulting in inadequate long-term predictions. Considering this shortcoming, during the early attempts to assess the effects of environmental changes on forest growth in Estonia, Nilson and Kiviste (1984, 1986) and Nilson et al. (1999) used site index (SI) models (which predict long-term mean height development of the stand) and compared model predictions and actual forest growth to assess differences in forest growth, possibly caused by environmental 11

13 changes. Uncertainties remained in the results because forest response to climate variation was not directly assessed. The increased need for more capable managerial tools forces development of more flexible forest growth models (Hasenauer, 2006) which could consider changes in growth conditions. Climate influence on tree growth can be modelled and forest growth in changing climatic conditions assessed (Girardin et al., 2008) using either a process-based, empirical approach or hybrid modelling approach, which is a combination of both these methods. Process-based modelling is a procedure where the behaviour of a system is derived from a set of functional components and their interactions with each other and with the environment, through physical and mechanistic processes occurring over time (Bartelink, 2000; Mäkelä, 2003). Process-based models commonly simulate basic growth processes, like photosynthesis, water balance, nutrient cycling, etc. within a tree (Mäkelä, 2003), but require a large number of input parameters, which are not commonly available through regular forest inventories. In Estonia, individual tree growth models based on empirical relationships could be used to predict forest growth in changing environmental conditions. Empirical forest growth and yield models for prediction of growth and yield use statistical techniques and are calibrated on comprehensive datasets and are widely used for decision support in different forest management scenarios (Peng, 2000). Such models may operate at different organizational levels (e.g. tree-, or stand-level), however, tree-level models are more flexible in the sense of tree species composition and for uneven-aged stands, compared to stand-level models. Jõgiste (2000) has developed an individual tree basal area increment model for Norway spruce (Picea abies (L.) Karst.), but climatic shifts cannot be predicted using this model since tree response to inter-annual climatic fluctuations is not considered in the model. Dendroclimatic analyses, based on tree-ring series could aid in quantifying climate influence on tree growth, by linking the annual growth rate and inter-annual weather variability over a longer time scale (II, III, IV). The most influential climate variables or variable sets are usually identified and could be further used in empirical growth models to make models climate-explicit. The inclusion of climate-related information into growth models could improve growth predictions as demonstrated by Huang et al. (2013). 12

14 Scots pine is an economically and ecologically important coniferous species in Europe (Pretzsch et al., 2014), including Estonia (Kaimre, 2010). As a predominant tree species, Scots pine accounts for 32.5 % (738,200 ha) of the forested land and 37.3 % (180.1 million m 3 ) of the growing stock (Raudsaar et al., 2016) in Estonia. Therefore, this thesis focuses on Scots pine growth and examines the influence of long-term local weather variation on Scots pine growth on a temporal and spatial scale (II, III, IV) using dendrochronological methods. A possibility for simultaneously incorporating climate-related information and response to thinning into a growth model is tested (III) by compiling an individual tree radial growth model for Scots pine. Possible climate change impacts on the height growth and site index of Scots pine are investigated (I), as the site index commonly describes site productivity in forest growth and yield models. In meteorology and climatology the terms weather and climate are commonly used to describe atmospheric conditions and are clearly distinguished. The weather describes short-term atmospheric variation, while climate is seen as an average pattern of weather fluctuations over a longer period ( 30 years long) for a particular area (Monkhouse, 1978). In dendrochronology (Cook and Kairiūkštis, 1990) and in this thesis, these two terms are used as synonyms, because variables (temperature and precipitation) describing weather and climatic conditions are related to growth using short (monthly) and long-time scales (time window longer than 20 years) simultaneously. 13

15 2. REVIEW OF LITERATURE 2.1. Shift towards empirical single tree growth models A forest growth model is a simplification of forest growth dynamics (Weiskittel et al., 2011), which allows performing a simulation of certain characteristics (e.g. height, diameter, basal area, volume) of the tree or forest stand over time (Vanclay, 1994). Generally, forest growth models are mathematical and biometric expressions of relationships between growth processes and the environment (Pretzsch, 2009). Forest growth can be modelled either at a stand or at individual tree level. Growth dynamics at individual tree level are usually more flexible because more details and interactions within the system surrounding the reference tree can be considered. Summing up or averaging the values of single trees produces stand level values. The use of tree level models is continuously increasing in Europe (Pretzsch et al., 2008) because of their ability to predict growth in heterogeneous stands (Porté and Bartelink, 2002). Most of the developed individual tree growth models are diameter or basal area growth models (Monserud and Sterba, 1996; Jõgiste, 2000, Nyström and Kexi, 1997; Andreassen and Tomter, 2003), with several models designated to predict the height of individual trees (Ritchie and Hamann, 2008; Vaughn et al., 2010). To describe the growth rate at a given time interval, input variables, such as age, stand density or basal area, commonly available through forest inventories are used in empirical models, making them widely applicable in forest management practice (Peng, 2000). However, empirical models have a drawback since they rely on descriptive relationships, implying that their applicability is limited to historical growth conditions, which are expected to remain relatively unchanged over time (Porté and Bartelink, 2002). This makes empirical models unreliable to be used in changing climatic conditions (Shugart et al., 1992). Several previous studies (Bravo-Oviedo et al., 2011; Nunes et al., 2011) showed that the productive capacity of the site and in turn stand dynamics (Pretzsch et al., 2014) are linked to climate and soil characteristics. Therefore, shortcomings in static empirical models could be eliminated by defining the environmental variables that control the growth of a given area (Bravo-Oviedo et al., 2011) and by incorporating them into a model (Girard et al., 2014) to 14

16 make it dynamic (González-García et al,. 2015) and suitable for use in varying environmental conditions Effect of weather fluctuations on tree annual increment The growth of trees is determined by their genetic potential and is a result of interactions of physiological processes, controlled by environmental factors (Kramer and Kozlowski, 1979, Landsberg and Sands, 2011). Trees, like other terrestrial plants, need sunlight, water, and nutrients to conduct photosynthesis and expand in size. Equally important for tree growth is the thermal regime, usually defined by air temperature. On an intra-annual scale, temperature regulates all biological processes in the trees, including cambial activity (Mäkinen et al., 2003, Rossi et al., 2006), and photosynthetic carbon assimilation (Mäkelä et al., 2004). However, biological processes in the trees might be restricted or even terminated when water shortage appears (Kramer and Kozlowski, 1979; Landsberg and Sands, 2011). Water is the primary component of photosynthesis and transpiration and is crucial for nutrient uptake (Bergh et al., 1999), therefore water deficit may reduce tree growth and vigour while extended periods of water deficit may lead to tree starvation, nutrient deficiency, wilting and even tree death. Nevertheless, different environmental factors have a positive effect on tree growth only when certain species-specific requirements are satisfied and interaction among them is possible. At northern latitudes, trees do not grow continuously all year round but follow the annual growth cycle, divided into four seasonal developmental phases: growth reactivation, meristematic activity, growth senescence and winter dormancy (Sarvas, 1972), emerging due to gradual changes in the photoperiod length and thermal regime. Because of this interaction, most of the hemiboreal zone trees produce one apical (shoot) and radial increment (tree-ring) every year. In general, phenological development in trees, from growth onset through winter dormancy, is endogenously controlled (Larcher, 2003), whereas daily height (Kanninen et al., 1982) and radial growth (Seo et al., 2007) rates are governed by temperature (Antonova and Stasova, 1993; Vaganov et al., 2006). Meristematic tissues, located between the xylem and phloem layers in 15

17 tree stems, branches and roots are responsible for shoot elongation (primary growth) and outward thickening, referred to as wood production or secondary growth (Plomion et al., 2001). Apical growth has priority compared to radial growth. Besides that, tree growth in height and expansion in diameter on the intra-annual scale do not coincide. For instance, Seo et al. (2010) monitored phenophases of Scots pine in the northern boreal part of Finland and found that height growth started in the middle of May followed by radial growth a few weeks later, with some variation among the sites and years. In general, wood formation onset and progress to a large extent is controlled by air temperature (Vaganov et al., 1999; Deslauriers and Morin, 2005; Vaganov et al., 2006; Gričar et al., 2006, 2007; Seo et al., 2007; Rossi et al., 2008). This process might be interrupted during the growing season, when sufficient moisture is not available, which is a common case for dry sites, and on soils with low water holding capacity. Plants, including trees, can use the received energy for growth, only if water is readily available, otherwise the received energy will only heat and stress the plant or vice versa, water is unused when energy supply is insufficient (Stephenson, 1990). A prolonged water deficiency may lead to premature cessation of wood formation (Güney et al., 2015) and lower increments. Therefore, climate factors, e.g. temperature and available moisture in the growing season, together with soil characteristics are important aspects that regulate tree growth and site productivity on the local scale Tree-ring analysis to study annual climatic variability in tree growth A large number of tree-ring based studies have revealed that a considerable part of the variability in tree-ring widths, resulting from complex interactions between trees and the environment, can be explained by weather fluctuations prior to or during the growing season (Fritts, 1976; Schweingruber, 1990). Wider tree-rings are produced when environmental conditions favour growth and narrower over adverse times. Therefore, tree-rings have been suggested to be natural environmental archives (Vaganov et al., 2006) and can be used for establishing a relationship between tree growth and weather fluctuations. Consequently, 16

18 most of the knowledge in reference to climate-growth relationships are being acquired from tree-ring analyses. Due to more laborious and expensive sampling, the use of other annually resolved growth measures, like height or volume increments for growth-climate calibrations, has been limited but possible (Mäkinen, 1998; Salminen and Jalkanen, 2005; Bouriaud et al., 2005; Rais et al., 2014). In the recent past, climate-radial growth relationships have been extensively used to reconstruct past climate (Fritts, 1976) and recently more often employed to forecast future growth (Williams et al., 2010; Bauwe et al., 2016) and to assess possible impacts of climate change on forest growth (Chhin et al., 2008). Knowledge on tree sensitivity acquired through dendroclimatological studies have been further used in growth and yield models (Girard et al., 2014), to make models climate-sensitive and improve their predictive capacities. Tree-ring measurement series show long-term fluctuations in tree growth and are advantageous to physiological observations because annually resolved growth data can be studied in a long-term context, and on a greater spatial scale (providing a greater variety of environmental conditions). Tree growth, however, is very complex, and annual (radial/height) increment is usually a result of complex interactions between intrinsic features of the studied individual and external factors (Fritts, 1976). Despite that fact, annual increments, wood density, and other structures of annual growth have been successfully related to basic climatic variables like monthly or seasonal temperatures and precipitation as well as to not so commonly used variables, such as vapour pressure deficit, cloudiness, radiation or derived indexes (e.g. drought, the Palmer Index and Standardized Precipitation Evapotranspiration Index (SPEI)) which incorporate the effect of several above-mentioned variables. For practical reasons or due to restrictions related to data availability, climatic variables used in dendroclimatic calibrations are resolved monthly, seasonally or annually. The longest meteorological series are usually available at monthly resolution. Aggregated meteorological data is used to test if longer time periods (seasons, several consecutive months, a year) can be related to a studied growth attribute. The results of correlation or response function analysis provide a direct 17

19 or indirect association between tree growth and a studied climatic component, however, they are not able to explain the cause. The physiological basis is usually found, based on the results of ecophysiological research, which is commonly used to interpret significant climate-growth relationships Climate effects on site index Forest site productivity and site index depend on soil characteristics, grown tree species and long-term climatic conditions (Assmann, 1970), which have a direct influence on height growth and physiological processes. Therefore, if any of these components is changing, site productivity is also affected. The inherent productive capacity of the site or site quality (Skovsgaard and Vanclay, 2008) has a great practical importance in forestry and usually describes the productive potential of the site to grow timber (Clutter et al., 1983). Site index, or mean dominant height of the forest stand at indexed age, has been widely used as an indirect and low-cost measure of site productivity (Clutter et al., 1983; Skovsgaard and Vanclay, 2008; Bontemps and Bouriaud, 2013). Therefore, it is an important input variable in stand growth simulators and yield tables (Pretzsch, 2009). Historically, when most of the site index equations were developed, long-term changes in forest soils and mean climatic conditions were minor, therefore it has been assumed that site index remains constant over time in a given area (Assmann, 1970). Variables, related to direct drivers of productivity such as nutrient availability or weather fluctuations were not included unless SI models were intended for large areas characterised by diverse climatic conditions. As tree height growth is climate-dependent (Mäkinen, 1998; Gamache and Payette, 2004, Kilpeläinen et al., 2006), the ongoing environmental changes (more favourable climatic conditions due to milder winters, extended growing periods; Menzel et al., 2003), including atmospheric fertilization (Kilpeläinen et al., 2006), may have an influence on forest productivity (Hari et al., 1984; Spiecker et al., 1996; Boisvenue and Running, 2006) in the long run. In Estonia, changes in forest productiv- 18

20 ity have been studied by Kiviste (1999a, 1999b), Nilson et al., (1999), and Paal et al. (2010). Site index increase and growth acceleration for Scots pine, birch and spruce stands were reported by Kiviste (1999a, 1999b) and Nilson et al. (1999). Whereas Paal et al. (2010) studied how uniform growing and soil conditions are for a single forest type in Estonia under different environmental conditions and found that tree layer productivity, described by tree height, diameter and timber volume were considerably lower in Western Estonia compared to the Southern part Climate effects on radial growth Tree growth response to weather variability is species-specific (Friedrichs et al., 2009), but may be mediated by soil properties (Lévesque et al., 2015), the hydrological regime (Dauškane and Elferts, 2011) and even tree size (Zang et al., 2012), tree age (Carrer and Urbinati, 2004), or tree social position within the stand (Lebourgeois et al., 2014). Topographic features of the landscape may also influence tree response to climate (Oberhuber and Kofler, 2000; Adams et al., 2014). Besides these factors, variability in local climate can contribute to contrasting tree responses (Sweingruber, 1988; Mazza et al., 2014). Tree response for example to negative drought effects can be modified to some extent through changes in stand structure (Martín-Benito et al., 2010; Guillemot et al., 2015). Due to a long list of factors causing differentiated tree response to annual climatic variability, it is strongly suggested to verify the climatic influence on tree growth using comprehensive tree-ring chronology networks (Büntgen et al., 2007), considering local climate variability (Mazza et al., 2014). Several researchers have investigated climate components in relation to the annual growth of Scots pine in the Eastern part of Europe. Most of the dendroclimatic studies on Scots pine rings in this region report its growth sensitivity to low winter temperatures (Bitvinskas, 1974; Špalte, 1978; Läänelaid, 1982; Cedro, 2001; Vitas, 2004). Growth response to precipitation is weaker but observed for several locations, including Southern Lithuania (Karpavičius et al., 1996; Vitas, 2004), Poland (Cedro, 2001), Southern Finland (Henttonen, 1984; Helama et al., 2005) and Northern Estonia (Hordo et al., 2011). 19

21 In Estonia, early dendroclimatological research on Scots pine has been focusing on developing tree-ring chronologies for wood dating purposes and for determining the climatic signal stored in tree-rings (Läänelaid, 1982; Läänelaid and Eckstein, 2003). Pärn (2003) studied Scots pine radial growth response to climate on dust-polluted sites in Northeast Estonia and Lõhmus (1992) investigated the climatic factors limiting the radial growth in Scots pine with respect to growing conditions. Recently, Scots pine radial growth variation in relation to weather fluctuations (Hordo et al., 2011), climatic variations and available soil water (Henttonen et al., 2014) were studied along a latitudinal gradient extending across Finland and including the Northern part of Estonia. In the recent past, high tree sensitivity to climate was one of the main prerequisite for dendroclimatic research (Fritts, 1976; Cook and Kairiūkštis, 1990), therefore, earlier studies targeted on sample trees growing in stressed environments. Under such site conditions, the radial tree growth would be limited preferably by a single climatic factor (Cook and Kairiūkštis, 1990). Due to this prerequisite, some of the regions and growing conditions could be underrepresented Scots pine habitats and climatic conditions in Estonia Scots pine is a tree species adapted to grow under various climatic conditions (Eckenwalder, 2013). Hence, its distribution range extends across Eurasia (Navasaitis, 2004; Eckenwalder, 2013; Sibul, 2014), from Northern Turkey to Northern Scandinavia, and continuing eastwards to Siberia and westwards reaching central Spain (Fig. 2.1). This indicates that Scots pine is a tree species with a wide ecological amplitude (Fritts, 1976). As a pioneer tree species, it is able to grow on a variety of different sites, including nutrient-poor dry soils. The optimal substrate conditions for Scots pine in North-Eastern Europe are suggested to be on sandy soils with some admixture of clay (sandy clay or clayey sand; Lõhmus, 2004). It is also common on waterlogged bog-type areas. In Estonia, Scots pine is present almost in all forest types, with the greatest share (40.3%) growing in mesotrophic forests. Scots pine in Estonia predominates in bog forests (89%), on the reclaimed areas of oil shale mining (74%) and is common in alvar forests (56 %) characterised by a thin topsoil layer. 20

22 Fig The natural distribution range (blue area) of Scots pine (EUFORGEN, 2009). Concerning the climate, Scots pine growth is better in continental climate, but the species can survive both very low and very high temperatures (Steven and Carlisle, 1959). Because of its great tolerance to drought, Scots pine has recently gained much attention as a potential tree species suitable for cultivation in changing climate conditions. Estonia is located almost in the centre of this tree species natural distribution range (Fig. 2.1), suggesting that climatic conditions are optimal for this species to grow here. The climate in Estonia is characterised by cold winters and warm summers and is strongly influenced by the North Atlantic Oscillation (NAO; Uvo, 2003). Despite the small size of the country, two climatic zones can be distinguished within the country. Maritime climate with milder winters prevails on the islands and in coastal areas and semi-continental climate is characteristic of the inland of the country. Climatic continentality in Estonia significantly increases from the west towards the east, and especially towards the Southeast (Jaagus, 2006a). Depending on the sub-region (reference period ; Tarand et al., 2013), mean annual temperature ranges from 4.6 to 6.8 C. The 21

23 coldest month is February, with mean temperature ranging from -6.5 to -2.5 C and the warmest month is July, with mean monthly temperature ranging from 16.5 to 17.8 C. Seasonal delay is characteristic of the coastal areas. Spring and summer start first in Southeast Estonia and after a short delay proceed towards North and the coastal areas (Jaagus and Ahas, 2000). Due to the warming effect of the sea surface, air temperatures from the end of summer until the middle of winter are considerably higher on the islands and along the coast, than in the inland. Thermal differences in air temperature between the coastal and inland areas increase towards the cold season, reaching the maximum in January (Tarand et al., 2013). The Northeast of Estonia is slightly cooler than the rest of the sub-regions of the country all year round, and the growing period is slightly shorter compared to southern Estonia and the coastal areas (Tarand et al., 2013). Deviations in mean temperature and precipitation patterns, as related to global climate warming are also smaller in the Northeastern part of Estonia, compared to the rest of the country. Estonia is located in a zone of excessive moisture, where total precipitation is greater than evaporation (Tarand et al., 2013). Annual precipitation ranges from 570 to 740 mm and is highly varying throughout the year. However, dry periods are common in the summer time (Tammets et al., 2011; Tammets and Jaagus, 2013). Spatial distribution of precipitation across the country is mainly influenced by landscape topography (Jaagus, 1992, 1999). Based on long-term observations, the Southern and Southwestern part of Estonia receives more rainfall compared to the other sub-regions (Jaagus, 2003) Research needs The environmental conditions under which forests grew in the recent past have changed (Jaagus and Ahas, 2000; Jaagus 2006a, Tarand et al., 2013), while previously elaborated empirical forest growth models, developed for past growing conditions are still used in practice. However, there are indications (Nilson et al., 1999) that predictions may not follow recent growth dynamics. Therefore, there is a need to re-evaluate the predictive power of existing forest growth models and to identify the influence of weather fluctuations on native tree species in Estonia. There is also a need to develop forest growth models that would include variables 22

24 related to climate variability, which would enable the models to be used under the conditions of a continuous environmental shift. Several researchers (Läänelaid, 1982; Lõhmus, 1992; Pärn, 2003; 2008; Läänelaid and Eckstein, 2003; Hordo et al., 2011) have been studying climate influences on Scots pine growth in Estonia, but most of these studies are restricted to a certain location, or neglect variation in site conditions. Lõhmus (1992) investigated climatic variables that could limit Scots pine growth in respect of site conditions, but he did not consider differences between local climates. Thus, spatially and temporally explicit knowledge of climatic influences on Scots pine growth, with respect to the site and local climate conditions is still limited on the spatial scale in Estonia. To predict Scots pine growth under changed climatic conditions simple linear models based on climate-growth relationships could be used. However, such predictions would be robust, and due to methodological drawbacks the responses to changes in stand structure (e.g. thinning) are removed during standardization and averaging. Whereas thinning, as a growth manipulation measure, is proposed as one of the activities of adaptive forest management, and tree response to stand structural changes needs to be included into the modelling process. Therefore, it would be reasonable to incorporate climate-related information further into growth models that already include the response to thinning. Individual tree growth models, based on empirical relationships, could be an option for growth prediction because structural and compositional diversity of forest stands is considered more accurately. Besides, tree response to management activities, particularly to thinning (Hynynen, 1995) and climate influence (Fritts, 1976), are predicted more properly at tree level. 23

25 3. AIMS OF THE STUDY With regard to the ongoing climate change, the aim of this doctoral dissertation is to study long-term weather variability influences on Scots pine growth by establishing relationships between tree radial growth and climatic variables and integrating climatic information into tree growth modelling. In this way, the effects of climate shifts can be taken into account when predicting future growth. As the radial growth of trees is not only driven by local weather fluctuations but it also responds to changes in stand structure, there is an aim to incorporate tree response to thinning and climatic information into the model. The specific aims of the study were: 1) To analyse long-term changes of Scots pine growth in Estonia following the alterations in temperature and precipitation regimes (I). 2) To identify the main climatic variables driving annual variability in Scots pine radial growth across a range of site and climate conditions in Estonia, using dendrochronological methods (II, III, IV). 3) To analyse the annual radial increment of individual Scots pine trees, and to assess whether the inclusion of thinning and climate variables improve the model of basal area growth predictions (III). 4) To investigate the temporal stability of tree response to local weather fluctuations (IV). The following hypotheses were tested: 1. Scots pine stands established after 1950 show higher growth rates than stands established several decades ago, growing on similar sites (I). 2. Relationships between climate and radial growth of Scots pine vary across the sub-regions of Estonia (II, III, IV). 3. Basal area growth of Scots pine can be modelled as a function of tree size, site productivity and climatic variables, whereas changes in competition can be included as tree response to thinning (III). 4. Climate-growth relationships, established between monthly climatic factors are not stable in time due to shifts in environmental conditions (IV). 24

26 4. MATERIALS AND METHODS 4.1. Study sites The studies presented in the thesis were carried out at 135 sites distributed across Estonia, the area between 59 50ʹN, 28 00ʹE and 58 00ʹN, 22 00ʹE, located in North-East Europe (Fig. 4.1). Radial increment core sampling was carried out, using a layout of the Estonian Network of Forest Research Plots (ENFRP; Kiviste et al., 2015) restricting sampling to plots (n=131) established in Scots pine dominated stands. Twelve of these plots are established in Scots pine plantations on reclaimed areas of oil shale (Narva quarry) open-surface excavation (III). The remaining 119 plots (II, IV) are located in Scots pine forests growing on forest land (Table 4.1). Such a sampling strategy was subjectively chosen in anticipation of the results obtained during this thesis to be coupled with other information available in the ENFRP for further use in tree-growth modelling. Besides that, the ENFRP is distributed across Estonia, covering a large geographical area and since it was designed to provide forest growth information for modelling purposes (Kiviste et al., 2015), it includes stands of different age, densities and covers a range of different growing conditions (Table 4.1). Sampling and analysis of pine forests on forest land were restricted to the forests growing on mineral soils, classified into mesotrophic and heath forest types and forest site types related to these forest types. The heath forest type is characterised by sparsely stocked pine stands with slow growth on dry and nutrient-poor soils. Heath forests include Scots pine forests growing on Cladonia and Calluna forest site types. Mesotrophic forests are found on moderately humid or temporarily moist sandy soils with a slightly better nutritional status (Lõhmus, 2004). Growing conditions in mesotrophic forests are described in more detail in article IV with Myrtillus and Rhodococcum forest site types. 25

27 Table 4.1. Summarised characteristics of the sampled Scots pine stands. SD - standard deviation. FOREST TYPE Forest site type Range Range Mean SD Mean SD Mean SD Basal area (m 2 ha -1 ) Diameter (cm) Height (m) Stand density (trees ha -1 ) Stand age (years) Region No. of sites No. of trees* Islands HEATH Calluna Cladonia MESOTROPHIC Myrtillus Rhodococcum Northeast HEATH Calluna Cladonia MESOTROPHIC Myrtillus Rhodococcum RECLAIMED Southeast HEATH Calluna Cladonia MESOTROPHIC Myrtillus Rhodococcum Southwest MESOTROPHIC Myrtillus Rhodococcum Total *Refers to the number of trees used in the analysis.

28 NE ISL Pärnu SW SE Fig Map showing the geographical location of the sampling sites (dots) and meteorological stations (triangles) included in the study. The colour of the dots refers to different datasets as follows: blue dots indicate stem-analysis sites (I), red dots represent Scots pine plantations on reclaimed oil shale areas (III), and green dots refer to the rest of the ENFRP plots (II, IV). Dashed lines show site division into geographical sub-regions. Abbreviations refer to the following sub-regions: ISL Islands, SW Southwest, SE Southeast, NE Northeast. To investigate the local climate influence on Scots pine growth, sampling sites were assigned to one of four geographical sub-regions (Northeast (NE), Southeast (SE), Islands (ISL) and Southwest (SW)), defined by the geographical site location. In addition to the ENFRP, four Scots pine stands (Table 1 in I) belonging to the NOLTFOX network, growing in South-eastern Estonia in an area managed by Järvselja Training and Experimental Forest Centre (58 25ʹN, 27 46ʹE) were chosen to study deviation in height growth patterns (I). To retain similarities among the stands and growing conditions as much as possible, even-aged Scots pine stands, growing in a close vicinity of each other but with different germi- 27

29 nation years (1891, 1952 and 1969) were selected. All four stands grow on Oxalis-Rhodococcum site type and are managed for timber production (more details in I) Meteorological data The Estonian Weather Service (the Estonian Environmental Agency, ES- TEA) provided monthly temperature and precipitation data from five meteorological stations, including Kunda (Northeast; II, IV), Ristna (Islands, Southwest II), Pärnu (IV) and Tõravere (Southeast; I, II, IV). Table 4.2. General information about meteorological stations and mean climatic conditions over the period of (except Jõhvi meteorological station) in the study sub-regions. Means, based on Jõhvi meteorological records, are calculated for the period of Region Islands Southeast Southwest Northeast, Northeast, coast inland Station name Ristna Tõravere Pärnu Kunda Jõhvi Latitude 59 16ʹ27ʺN 58 16ʹ09ʺN 58 23ʹ09ʺN 59 29ʹ54ʺN 59 21ʹ33ʺN Longitude 23 43ʹ55ʺE 26 27ʹ30ʺE 24 29ʹ49ʺE 26 31ʹ33ʺE 27 25ʹ15ʺE Record start year Mean annual temperature ( C) Mean annual precipitation (mm) Growing period temperature ( C) Growing period precipitation (mm) Mean July temperature ( C) Mean February temperature ( C)

30 Temperature ( C) T= 5.3 C Southeast (Tõravere) P = 606 (mm) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Precipitation (mm) Temperature ( C ) Southwest (Pärnu) 100 T= 5.9 C P= 673 (mm) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Temperature ( C ) Northeast inland (Jõhvi) T= 4.6 C Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec P= 695 (mm) Precipitation (mm) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Northeast coast (Kunda) 100 T= 5.2 C P = 558 (mm) Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec 0 Precipitation (mm) Temperature ( C ) Precipitation (mm) Temperature ( C ) T= 6.4 C Islands (Ristna) TEMP P = 621 (mm) PREC Precipitation (mm) Fig Climate diagrams for separate study sub-regions. Temperature (T) and precipitation (P) means for individual stations/sub-regions are calculated for the reference period of The length of climatic data records differed among the stations (Table 4.2). Daily meteorological data for the period of was available from Jõhvi meteorological station (Northeast, reclaimed areas; III). Seasonal climatic variables were calculated based on the provided monthly meteorological records, as an arithmetic mean of three consecutive months: December, January and February for winter, March, May and April for spring, June, July, August for summer and September, October, November for autumn. Seasonal precipitation sums were calcu- 29

31 lated respectively. Mean annual temperatures and total rainfall were calculated for a calendar (January December) and for a hydrological year, starting from September of the previous growing season until August of the current year. The growing season was defined as a period from April to September unless specified otherwise Investigation of changes in height growth patterns (I) For studying deviations from long-term growth, the site index comparison approach (Untheim, 1996; Bontemps et al., 2010) was applied by comparing cumulative height growth (expressed as the site index (SI)) of three Scots pine generations growing in similar conditions but established with different germination dates. The growth of the oldest pines (mean age 115 years) was used as a reference for the expected growth and trees from two consequently younger generations (mean age 56 and 40 years, respectively) were used for analysing significant growth deviations. Since long-term observations on tree height growth were not available, we applied the stem analysis technique (Spiecker, 2002), which allows retrospective reconstruction of annual height increments for the whole tree lifespan, including early tree height development (Karlsson, 2000). Only dominant and co-dominant trees were selected to study changes in the height growth pattern because the height growth of trees growing above and within tree crown level are not strongly affected by thinning (Valinger et al., 2000). Selected pine trees were felled and cut into sections of 2.5 m in length and annual height increments were reconstructed for the sectioned trees, with an accuracy of 0.1 cm by measuring distances between two consecutive whorls along the split trunk of the tree (I). Radial increments were measured on cross-sections taken at 1.30 m height above the ground and at the base of each section. For time-dependent growth comparisons, height and diameter cumulative growth was calculated for each age group, based on measured height and radial increments (Fig. 3 in I). A three-parameter Richards (1959) growth function and a non-parametric generalized additive model (GAM; Wood, 2006) were fit on the measured cumulative height growth curves of individual trees to approximate height development and to calculate 30

32 the SI at a reference age of 40 years (SI 40 ). Differences in height growth rates were analysed with one-way ANOVA. In addition, the height development of three cohorts was compared to mean Scots pine height development on Rhodococcum site in Estonia, using the model for stand height prediction (Kiviste, 1997; Kiviste and Kiviste, 2009). In addition, mean climatic conditions, over three 40-year-long periods when three studied tree generations grew, were investigated using pair-wise one-way ANOVA with Tukey s post hoc test. The period of refers to old tree growth, the period of to middle-aged trees and to the youngest tree growth period. Climatic conditions were characterised by mean annual temperature and total precipitation, mean seasonal temperatures and seasonal rainfall, mean temperature and total precipitation of the growing period, defined as a period from May to August as well as mean monthly temperatures and monthly precipitation sums. Meteorological data (monthly mean temperature and total precipitation) from Tõravere weather station was used for this analysis Dendroclimatic analysis (II, III, IV) Increment core sampling, preparation and measurements Tree-ring increment cores were extracted from the living Scots pine trees at 1.3 m above the ground level, using a Pressler increment borer. Double radii containing cores were taken only from dominant and co-dominant trees in the stand by selecting the trees at four locations, described by cardinal directions (North, East, South, West) outside the permanent sample plot (II, III, IV). The number of sampled trees per site ranged from 8 to 15. The trees were inspected for visual stem and top damage before coring and excluded from sampling if any damage was detected. The cores were mounted into a holder and their surface was shaved with a razor blade to enlighten the borders of the rings (Stokes and Smiley, 1968). If a temporal holder was used (II), then the cores were soaked in water for minutes to regain full size before preparation. Chalk was applied on the cores to enhance the visibility of rings. Ring widths were measured with an accuracy of 0.01 mm using LINTAB TM 5 measuring device and TSAPWin TM software (RINNTECH, Heidelberg, 31

33 Germany). The measured series were visually and statistically cross-dated to inspect for measuring mistakes and missing or false rings (Schweingruber, 1996). The measurements of two radii were averaged for each tree to reduce within tree growth variability. The resulting average treering series were further used to control the cross-dating quality with the statistical software COFECHA (Lamont-Doherty Earth Observatory, Palisades, USA; Holmes, 1983; Grissino-Mayer, 2001), which estimates tree-ring series correlation with the master chronology (built from the rest increment series in the set) and identifies poorly matching segments (Grissino-Mayer, 2001). Tree-ring series with synchronous variation within and between the sites are expected to contain a common signal, which is suggested to be caused by common environmental influence, e.g. climate (Fritts, 1976). The series poorly correlating with the master chronology (p<0.05) were discarded from the analysis, since they could have contained a non-climatic signal, associated with stand dynamics Standardization of tree-ring series Raw tree-ring width is a result of complex interactions between a tree and physiological drivers, among which non-climatic factors, like age trend, are responsible for ring width decrease with increasing tree age (Fritts, 1976; Cook and Kairiūkštis, 1990). In a growing tree, the surface, where cells are being laid down increases, while wood volume remains considerably constant. Before conducting a climatic analysis, non-climatic age trend should be removed, otherwise, tree-ring values will reflect the age effect rather than the climate signal (Fritts, 1976). In the thesis, to emphasise the climatic signal, tree-ring series were detrended (standardized) individually by fitting either a negative exponential curve (II) or spline (III, IV). Tree-ring indices were computed for each individual tree-ring series by dividing the observed ring width value by the value of the curve. Spline rigidity was determined to be 67% of the length of each individual ring width series (Cook and Peters, 1981). Standardization for studies III and IV was performed using the functions provided in the package dplr (Bunn, 2008) within the R software (R Core Team, 2016) while detrending of the tree-ring series in study II was carried out with the program ARSTAN (Lamont-Doherty Earth Observatory, Palisades, USA; Cook and Krusic, 2006). 32

34 Due to the particularity of physiological processes within the tree, radial growth increments tend to autocorrelate, which means that the current year s growth depends to some level on the previous year s growth. Such dependency is not desired in modelling and may cause biased estimations (Salminen et al., 2009). Therefore, autoregressive modelling (Box and Jenkins, 1970) was applied to remove temporal autocorrelation and to enhance high-frequency variation, which is assumed to be caused by weather fluctuations. Akaike (1974) information criteria were used to determine the order of the autoregressive process Development of chronologies Detrended series were further used for construction of the chronologies. It is noted that trees growing in similar site conditions tend to synchronise the physiological processes, like cambium or other living cell reaction to a particular combination of factors, resulting in ring-width sequences with great similarity (Schweingruber, 1988; 1996). Based on that knowledge, Scots pine response to weather fluctuations was analysed by grouping annual growth patterns and developing chronologies according to ecological site conditions. Growing conditions in pine forests on forest land, during current dendroclimatic studies, were characterised by forest type (II) and forest site type (IV) in pine forests on forest land. The grouping according to forest site type was done because it was thought that tree response by forest type can be too robust, since with averaging, some of the signal characteristic of a certain forest site type may become lost. Besides, forest site type is the main classification unit in forest management (Lõhmus, 2004; Forest Act, 2006) and is often used as an indirect indicator of potential site productivity. Scots pine response to climatic variability on reclaimed oil shale areas where growing conditions vary greatly with respect to soil characteristics (including continuous substrate formation) as well as stand structures, was studied at site level by combining index series by plot and by grouping growth patterns into three sites, characterized by similar age and stand structure. In all cases, composite chronologies were developed by averaging all the series in the set into mean value chronology. The averaging process aims to reduce the growth impulses inhered in a given stand, arising due to local disturbances (e.g. thinning). The averaging of 33

35 tree-ring indices was achieved and standard chronologies for each target group were produced by using Tukey s biweight robust mean that minimises the influence of possible outliers (Mosteller and Tukey, 1977). The index values surpassing the estimated mean value by 6 or more standard deviations were excluded along the averaging procedure. By using the residuals from the autoregressive modelling, residual chronologies were computed. Pre-whitened (residual) chronologies tend to contain a stronger climatic signal (Cook and Pederson, 2011). Chronologies for study II were developed in the program ARSTAN (Cook and Holmes, 1996), and in studies III and IV for chronology calculation, the package dplr (Bunn, 2008; Bunn et al., 2013) within the R software (R Core Team, 2016) was used Assessment of chronology quality Before using chronologies for dendroclimatic analysis, chronology quality and its reliability for growth-climate analysis needs to be evaluated and the strength of a common signal assessed. For this purpose, a set of statistics commonly used in dendrochronology were calculated for the ring width index series. Coherence and common signal among the growth patterns were assessed by calculating the series inter-correlation (IC) and Gleichläufigkeit (Glk). The IC is an average correlation of each index series to a master chronology, which is compiled from all other series. A higher IC indicates a stronger common signal in the studied population. The Glk is another measure of accordance among the series (Eckstein and Bauch, 1969), showing mean percentage of years in which ring widths in two series being compared synchronously increase or decrease (Fritts, 1976). Values above 60% indicate acceptable correspondence of growth patterns, whereas Glk values above 70% indicate a great level of common variance (Esper et al., 2008) and are characteristic of climate-sensitive trees (Frank and Esper, 2005). Average mean sensitivity (MS) shows mean relative change between two adjacent ring widths. It has been suggested that the MS in trees containing a stronger common signal tends to be greater compared to trees with complacent growth. Similarly to the MS, average standard deviation (SD) indicates the magnitude of annual growth fluctuations. Greater standard deviations are expected in highly responsive trees (Fritts, 1976). 34

36 To assess the strength of the common signal, signal to noise ratio (SNR) as well as the expressed population signal (EPS) were calculated for each set of indices. The EPS is used to assess common variability among the ring width index series and to evaluate chronology reliability for dendroclimatic analysis (Briffa and Jones, 1990). The EPS indicates how close the developed chronology is to the hypothetical one. An EPS equal or higher than 0.85 was proposed by Wigley et al. (1984) as a threshold for chronology confidence and is widely used among dendrochronologists. The SNR and EPS for each chronology were defined using the following equations: (2) where rbt is the average correlation between tree-ring series in the set, calculated for the common period of all index records (Briffa and Jones, 1990) and n is the number of mean tree-ring width series in the set. First-order autocorrelation (AC1) was calculated to determine the current year s growth dependency on the previous year s influence, and to evaluate the need for autoregressive modelling. For the rest of the above-listed statistical indicators, higher values indicate greater tree sensitivity and common signal, which is assumed to be caused by large-scale external factors, like the climate (Fritts, 1976). (1) Analysis of growth-climate relationships The influence of annual climate variability on Scots pine growth was assessed based on the relationships established between tree-ring widths and climatic variables (temperature and precipitation). Residual chronologies were used in all climate-growth analyses by relating standardized ring-width data to monthly meteorological data (temperature and precipitation) from the closest meteorological station (Fig. 4.2) and calculating Pearson s correlation coefficients and response function analysis also known as principal components regression. Response func- 35

37 tion is a kind of multiple regression where ring-width chronologies are used as dependent variables and principle components calculated on the weather data are used as growth predictors (Fritts, 1976; Briffa and Cook, 1990). Response function analysis enables identifying climatic variables that are not inter-correlated between each other, so multicollinearity problems in multiple regressions can be avoided. The significance of the correlation and response function coefficients was tested with the bootstrapping procedure (Guiot et al., 1982; Guiot, 1990; Zang and Biondi, 2013) based on 1,000 iterations, using the 95 th percentile rank. The length of the time window used for climate-growth calculations within each study (II, III, IV) varied depending on the available growth and meteorological data. The climate-growth calibration period was the shortest for Scots pine plantations, restricted to the period of (21 years; III). Relationships between radial growth and climatic variables were investigated in interval between (52 years), which was a common period for all forest site types in the four studied sub-regions (IV). The analysis was restricted to a common time interval of (62 years) for relating radial growth to weather fluctuations at forest type level (II). The climate-growth analysis was performed within the R software (R Core Team, 2016), using the functions bootres (III; Zang and Biondi, 2013) and treeclim (IV; Zang and Biondi, 2015) Investigation of spatial and temporal response patterns (IV) To identify sites with similar growth patterns and better understand the tree responses to weather fluctuations across the studied area, chronologies for each forest site type, based on the common interval, were compared using Pearson s correlation and hierarchical cluster analysis (HCA). In HCA, chronologies were assigned into clusters according to dis/similarities, expressed as the Euclidean distance that was established with Ward s minimum variance method, using an algorithm containing the initial Ward s clustering criterion where dissimilarities are squared before cluster updating (Murtagh and Legendre, 2014). Moreover, to identify groups of forest site types with similar response and sensitivity to climate at regional level, the principal components analysis (PCA) was performed on the BCC matrix (Weber et al., 2007; Méri- 36

38 an et al., 2011). The PCA was conducted based on variance-covariance, considering that the magnitude of descriptors was similar (Legendre and Legendre, 1998; Mérian et al., 2011). To evaluate whether climate-growth relationships between tree-ring width indices and monthly climate variables remained stable over time moving correlations were calculated over 30-year-long intervals shifted by one year. Bootstrapping was used to estimate the 95% confidence intervals for correlation intervals. Temporal stability of growth-climate relationships over period (over for pine plantations on reclaimed areas) were tested using Scots pine chronologies developed for separate forest site types at different sub-regions of Estonia Identification of pointer years (II) Growth response of Scots pine to extreme climatic events was studied at forest type level (II) using pointer year analysis (Schweingruber et al., 1990; Neuwirth et al., 2007). The pointer year analysis together with the correlation or response function analysis facilitate more comprehensive understanding of growth limiting factors on the corresponding site (Neuwirth et al., 2007). Pointer year analysis aims to identify and explain the appearance of markedly narrow or wide rings, which are formed due to extreme climatic conditions, commonly referred to as climatic anomalies. Two terms are used in pointer year analysis: event years, referring to years of extraordinary growth at the individual tree level, and pointer years when certain event years replicate in a larger number of individuals (e.g. in most of the trees at the site; Schweingruber et al., 1990; van der Maaten-Theunissen et al., 2015). In this thesis, pointer years were indicated by applying the method proposed by Cropper (1979), according to which normalisation was applied in a moving 13-year window. Pointer years were calculated using the following equation: (3) 37

39 where C i is a Cropper value, TRW i is tree-ring width formed in the year i; TRW mean[window] is the arithmetic mean tree-ring width in the moving window, and TRW stdv[window] is the standard deviation of ring widths in the moving window. Pointer years were identified on non-standardised tree-ring series considering the maximum data period, restricted to replication by at least 5 series. Years containing C i values above 1 or below -1 were identified as positive or negative pointer years, respectively (Neuwirth et al., 2007; Pourtahmasi et al., 2007). Moreover, following Neuwirth et al. (2007) pointer years were divided into three intensity groups, defined by C i values as follows: C i >1, weak pointer year; C i >1.28, strong pointer year, C i >1.645, extreme pointer year Modelling basal area increment (III) Annually resolved growth data from 128 Scots pine trees growing on 12 sites of reclaimed areas was used to fit individual tree radial growth model. The nonlinear modelling approach was applied to model basal area growth of individual Scots pine trees growing on reclaimed areas of open surface-mining of oil shale (III). Tree-ring widths were converted into annual basal area increment (BAI) series to account the differences related to tree size (Biondi, 1999; Biondi and Qeadan, 2008), using the following equation: (4) where R is tree radius and t is the year when the tree-ring was formed. Since the aim was to elaborate the model as parsimonious as possible, the constructed model had a multiplicative structure, containing three components: the basic growth function, tree response to thinning and climate variability, and was expressed as follows: (5) where the dependent variable iba is the annual basal increment in the year i (cm 2 year -1 ), is a growth function describing individual tree basal area growth assuming open-space conditions, is a function describing tree response to thinning, and is a function describing annual growth fluctuations as related to weather variation, and is 38

40 the error term. Each function was described by a sub-model. Considering that basal area increment can be described by an asymptotic pattern (Pokharel and Dech, 2012), Weber growth function (cf. Kiviste et al., 2002) in relation to tree size and site quality was used to model the component : where p tree is the parameter related to tree growth potential (current tree size/diameter) and defining the magnitude of the asymptote; p site is the parameter related to site productivity potential; A is the tree breast height age (years). One of the components in the radial growth model is usually attributed to competition. Instead of that, a function depicting tree response to a reduced competition level was introduced in developed model. Individual tree response to thinning was modelled as an asymptotic growth increase, lasting for 13 years, as proposed by Hynynen (1995). Since information on thinning intensities are rarely available in practice and for this study, time elapsed since thinning was used to describe tree response to reduction in competition. Therefore, the equation proposed by Hynynen (1995) was adjusted to the available variables as follows: (6) (7) where th is the time elapsed since thinning (years); b and c are the estimates of parameters, respectively b=13.8 and c=1.58; and p th is the parameter related to thinning to be estimated. Annual growth fluctuations, as related to variations in weather conditions, in the model were described by function and were modelled by including indices (III), since tree-ring width indices are usually calculated to reflect climatic forcing on tree radial growth. It was additionally tested whether the use of a set of the most influential climatic variable would improve model importance. Therefore, based on the dendroclimatic study results (Fig.4. in III) a least-squares stepwise multiple regression model was compiled taking into account climatic variables explaining the most of annual variation and indices replaced by the following model: 39

41 (8) where I t is the tree-ring index in the year t; P JJ is total June July precipitation (mm), T spring is the mean spring (March-April-May) temperature ( C). The stepwise procedure with a forward selection approach was used to include model climatic components and to estimate the parameters. General chronology developed as mean composite chronology for reclaimed areas was used as a dependent variable when compiling the regression analysis. Model evaluation was carried out based on the graphical analysis of residual distribution and by evaluating model fit statistics, including the root mean square error (RMSE), which indicated the precision of estimates, mean estimation error (MEE), which shows mean deviation of predicted values from the observed, and pseudo R 2 estimating, the total explained variance. The fit statistics were calculated using the following equations: (9) (10) (11) Where is predicted basal area increment (cm 2 /year), is observed basal increment (cm 2 /year) and n is number of observations, SD is mean standard deviation of all BAI observations, used to build the model. The Akaike Information Criterion (AIC) was used in model evaluation (Burnham and Anderson, 2002). 40

42 5. RESULTS 5.1. Changes in long-term forest productivity Both mean annual height and radial increment rates (Fig. 5.1a; Fig. 3 in I) were significantly (p<0.0001) greater in two younger age groups, established between , compared to the oldest group, established before 1900, resulting in the different development of cumulative growth curves (Fig. 5.1a; Fig. 3 in I). At the reference age of 40 years, mean height (SI 40 ) for the young, middle-aged and old cohorts were 24.0, 21.8 and 19.5 m, respectively (Fig. 4 in I), or 12% greater in the middle-aged cohort and 23% greater in the youngest group compared to the growth of the oldest trees. The difference in SI 40 between the youngest and middle-aged trees was 11%, indicating greater growth of the youngest generation, established around When differences between average Scots pine height in Rhodococcum site type and mean height growth of each group were calculated, diverse trends for age groups were observed (Fig. 2b in I). Fig Mean cumulative height curves of the studied trees (a) and mean relationship trends between annual height increment and cumulative tree growth (b) for three studied cohorts. Grey continuous lines represent single tree growth, dark lines refer to cohort mean and dashed dark red lines are 95% confidence intervals. 41

43 Fig Significant differences in climatic conditions in which three studied cohorts grew for the first 40 years. The period of refers to old (Old), to middle-aged (Mid) and to young (Young). Periods with significantly different (p<0.05) climatic conditions, described by mean annual temperature (T annual), mean spring temperature (T spring), mean temperature in April (T April), mean temperature in August (T August), total winter precipitation (P winter) and total precipitation in February (P February) are identified with dashed lines and p-values. The comparison of climatic conditions in which three studied cohorts grew for their first 40 years, provided evidence that there were statistically significant differences (Fig. 5.2). It was revealed that thermal growth conditions throughout the year were significantly warmer for the youngest trees, established around 1970 (T annual = 5.4±1.1 C) than for middle-aged, established around 1950 (4.8±1.1 C) and the oldest pine trees, established a century ago (4.7±0.9 C). Mean temperature during the growing season, defined as the period from May to August, was also significantly higher over the period, when the youngest pine trees grew 42

44 (T veg = 14.9±0.9 C) compared to other two periods (middle-aged 14.4±0.9 C vs. old 14.3±1.1 C). Springtime (March-April-May) was warmer by 1.0 C and mean temperature in the summer season was greater by 0.5 C for the youngest pine generation compared to the earlier growth periods. On a monthly basis, significant warming appeared in April and August. No significant differences in thermal conditions were detected between two earliest periods ( (old) vs (middle-aged)). However, Tukey s post hoc test displayed significant differences in the precipitation regime for these two periods, suggesting that winter precipitation (total P winter = 115.3±39.7 mm vs. 97.9±26.6 mm) and particularly precipitation in February was greater when the oldest trees grew (total P Feb = 34.7±17.7 mm), compared to the growth period of middle-aged pine stands (total P Feb = 25.9±13.3 mm). No significant differences in winter precipitation were detected when two recent periods ( vs ) were compared Scots pine radial growth characteristics and relationship to inter-annual weather variations Scots pine chronologies Dendroclimatic Scots pine studies (II, III, IV) conducted during compilation of this thesis supported the hypothesis (2) of climate-growth relationship variability across the sub-regions of Estonia and also with respect to site conditions. For identifying the most influential climatic variables on Scots pine growth, tree-ring data from more than 900 living Scots pine trees (Table 5.1.) were used and a number of ring-width chronologies were developed (Fig. 5.2.; II, III, IV), considering similarities in ecological and climatic growth conditions, representing radial growth at the investigated site. Separate chronologies were built for Scots pine trees growing in mesotrophic and heath forests at four sub-regions of Estonia (Figures 4, 5 and 6, in II) and fourteen residual ring-width chronologies were constructed for five forest site types at four regions (Fig. 5.3). Twelve plot-level (data not presented) and three site-level residual chronologies (Fig. 3 in III) 43

45 were developed to investigate in more detail the climate influence on the growth of Scots pine plantations established on reclaimed oil shale areas. Fig Mean residual chronologies by forest site types within four sub-regions of Estonia. The chronologies abbreviations are explained in Table 5.1. Chronologies developed for heath and mesotrophic Scots pine forests were the longest ( years; Table 1 in II) and well replicated, while chronologies for pine plantations on reclaimed areas were only years long, but replicated by a sufficient number of samples (Table 5.1; Table 3 in III). Pine ring-width chronologies by forest site type spanned years with mean replication of 62 trees, (range from 14 to 173; Table 5.1). 44

46 Statistical parameters (Table 5.1) calculated for each Scots pine chronology by forest site type indicated that the chronologies are reliable for use in growth-climate calibrations. Mean series inter-correlation characterising similarities among the individual tree growth patterns ranged from to (overall mean 0.530). Myrtillus forest site type chronologies tended to exhibit a lower correspondence among the index series, except on the Islands. Mean Glk, characterising the synchrony among the growth patterns was 61% with little variation between 57 to 66%. The mean sensitivity ranged between 0.17 to 0.23 (overall mean 0.19), with no clear pattern along the ecological gradient. On the other hand, year-to-year variation was slightly higher for all sites in the Northeastern sub-region. The first-order autocorrelation, estimated on standardized series and indicating influence of the previous year s growth on the current growth was moderate ( ; mean 0.42 for all sites, except Scots pine plantations on the reclaimed areas (0.16)). Common signal strength, as measured by the EPS ranged from 0.85 to For Calluna-NE chronology, the replication threshold of 5 trees was not sufficient to achieve the EPS value of 0.85, therefore chronologies were truncated to the year when this threshold was fulfilled. Common variance, accounted by the PC1, ranged between 24.5 to 53.9% (overall average 36.0%) and tended to be greater for Northeast sub-region sites, except for Myrtillus site. More than 50% of common variance summarised in PC1 was detected for Cladonia forest site type and reclaimed areas from the Northeast sub-region. 45

47 Table 5.1. Summary of the basic statistics of Scots pine forest site type chronologies in four sub-regions. Abbreviations: IC-inter correlation between ring-width measurement series, Glk-Gleichläufigkeit, MS-mean sensitivity, SD-standard deviation, AC1-first order autocorrelation, SNR-signal to noise ratio, EPS-expressed population signal, PC1-variance explained by the first principal component. MS SD AC1 SNR EPS PC1 (%) IC Glk (%) Span (n>4 series) Range (n>4 series) Series length (years) min/mean/max No. of trees* Sub-region Chronology name Forest site type Calluna Islands CaluNE 19 (38) 17/73/ Cladonia Islands CladISL 29 (58) 41/85/ Myrtillus Islands MyrtISL 90 (180) 21/65/ Rhodococcum Islands RhodISL 77 (154) 20/74/ Calluna Northeast CaluNE 21 (42) 38/89/ Cladonia Northeast CladNE 173 (346) 35/68/ Myrtillus Northeast MyrtNE 40 (80) 26/58/ Rhodococcum Northeast RhodNE 24 (48) 52/65/ Reclaimed Northeast ReclNE 128 (258) 15/28/ Cladonia Southeast CladSE 14 (28) 27/74/ Myrtillus Southeast MyrtSE 24 (48) 39/67/ Rhodococcum Southeast RhodSE 97 (194) 20/67/ Myrtillus Southwest MyrtSW 31 (62) 20/40/ Rhodococcum Southwest RhodSW 48 (96) 34/45/ *Number of trees used to compute the chronology.

48 The hierarchical cluster analysis applied on standardised Scots pine growth patterns grouped chronologies into three clusters (Fig. 5.4; Fig. 5 in IV), implying closest similarities among the growth patterns within the same sub-region, and suggesting that local climate is the main driver of a common growth. Fig Dendrogram of hierarchical cluster analysis applied on the 14 residual forest site type chronologies, sharing a common period of Grey lines show clusters. Chronology abbreviations are explained in Table 5.1. A separate cluster was produced for the Islands and Northeast sub-regions, while forest sites from the Southeast and Southwest sub-regions were aggregated into one cluster. Chronology for Scots pine plantations on reclaimed areas was aggregated into a cluster with other chronologies from the Northeast, however there was a considerable separation between Scots pine chronology of reclaimed areas and other chronologies, indicating that growth in the disturbed areas differs from the growth patterns of natural forest sites. Scots pine chronologies by forest type were compared to the general chronology for Scots pine (reference chronology) previously developed by Lõhmus (1992) (Figure 4, Table 1 in II). The accordance between separate new chronologies and the reference chronology ranged between 56.1 to 80.7% based on the Glk and was greater for heath forests. T-values with Baillie-Pilcher standardization (TBP) were also relatively high (ranging between ) except for two chronologies of mesotrophic forests (Northeast TBP=3.7 and Southwest TBP=3.4), indicating a lower similarity between these chronologies and the reference chronology. 47

49 Radial growth-climate relationships The results of correlation and response function analysis indicated that climate influence on Scots pine growth was not homogenous across the studied area and varied both on the spatial scale and with respect to site conditions (II, III, IV). Tree-ring width growth on both studied forest types (heath and mesotrophic) was positively correlated with mean winter temperature prior to the growing season and mean temperature of the growing period. The radial growth in the Northeast and on the Islands was negatively associated with mean August temperature of the prior growing season and positively related to the total rainfall in that month. Radial increment in the mesotrophic forest of Southwest Estonia was positively related to precipitation in February. The annual radial increment of Scots pine in the Southwest and on the Islands for both mesic sites was significantly positively correlated to mean annual temperature. Since the relationships established between Scots pine growth at forest type level is a robust summary of pine response at forest site type level, in addition Scots pine response to climate analysis was conducted to define climate-growth relationships at forest site type level. Therefore, tree-ringwidth dataset from mesotrophic forests was downscaled and grouped into Myrtillus and Rhodococcum forest site types and heath forests into Calluna and Cladonia forest site types according to sub-region. Therefore, the response patterns at forest site type level represent ecological growth conditions more precisely, and consequently obtained response patterns describe Scots pine response to climate variability better. The greatest number of significant positive correlations at forest site type level (Fig. 5.5) were established between radial growth and mean temperature in February (n=10), March (n=9) and April (n=7) and with the total precipitation of the previous August (n=8). The number of significant correlations per chronology ranged from one, very strong (r=0.49) for Scots pine plantations on reclaimed areas to nine moderately strong monthly climatic variables for Rhodococcum forest site on the Islands. A positive relationship to mean air temperatures from December/January to April was established for both mesic habitats (Myrtillus and Rhodococcum) from the Islands, the Southwest and Southeast (MyrtIsl, MyrtSW, 48

50 RhodIsl, RhodSW, RhodSE). For Myrtillus in the Southeast, the relationship was stronger only for mean temperature in March (r=0.32). The sites in the Southwest (MyrtSW, RhodSW) showed a strong correlation (r= ) with mean temperatures in March and April, whereas this relationship was weaker for the sites on the Islands. Sensitivity to mean monthly temperatures of the cold period and early spring was characteristic of a mesic sites and weaker for Cladonia and Calluna sites and all sites in the Northeast region. The mean temperature of the previous August was significantly negatively related to the growth in all sites on the Islands and mesic sites of the Northeast region (MyrtNE, RhodNE) and Rhodococcum site in the Southeast, whereas a positive response to the total rainfall for this month was displayed by eight chronologies from the Northeast and Islands. Sensitivity to the total rainfall in the previous August was stronger for the Scots pine trees growing on the Islands (r= ) than in the Northeast region (r= ), suggesting that the climatic water deficit in these sites might be influential for radial growth during the next growing season. Several Scots pine chronologies for mesic sites, mainly in the Southwest and Islands (RhodSW, RhodISL, MyrtISL, RhodSE) showed a significantly negative relationship (r= ) to the total rainfall of the previous September. A considerably weaker negative correlation between the total precipitation of the previous October was displayed for radial growth in Cladonia site in the Southeast. The response function verified the significant relationships established by the correlation analysis, but the number of climatic variables and relationship strength was considerably lower. No significant relationships were observed between radial growth on Calluna site on the Islands and climatic variables. 49

51 Fig. 5.5 Visualisation of Pearson s correlation (bars) and response function (lines) coefficients between pre-whitened Scots pine chronologies by forest site type in four sub-regions and mean monthly climatic variables. Dark bars and filled dots indicate statistically significant (p<0.05) relationships from the previous June (j) to the current August (A) of the current year of ring formation. 50

52 The PCA analysis applied on BCC suggested that Scots pine climatic sensitivity depends on geographical sub-region. The first two principal components explained 70.9 % of total variance (Fig. 6 in IV). The first principal component captured 48.3% of the total inertia and described Scots pine sensitivity to current growing season climatic conditions. Most of the chronologies from Southeast, Islands and Southwest were positively related to this component while Northeast chronologies were clearly separated. The second principal component captured 22.6% of the total variance and described Scots pine sensitivity to water availability at the end of previous year summer. It was the most limiting climatic factor for chronologies from Islands, but had opposite influence to pine populations in the Southwest. Scots pine response to climate on reclaimed areas of oil shale was investigated at a finer scale, by grouping trees into three sites (more details in III). A strong positive relationship (r range of ) between the radial growth of Scots pine plantations on reclaimed areas and rainfall in the period of June July for all age groups was established through Pearson s moment correlation analysis. Warmer springs were positively related to the tree-ring widths of the two oldest sites (r=0.50 and r=0.69, site 1 and site 2, respectively). A moderately strong relationship between mean temperature in January and ring width was established for trees in the densest stands (site 2; r=0.39) and a much stronger relationship (r=0.56) for the youngest trees growing on more fertile soils (site 3). The radial growth of trees on this site was also negatively associated (r=-0.45) with mean temperatures in the summer. Pearson s correlation analysis between the BCC and variables describing the soil and stand (Table 1 and Table 2 in III) indicated that the sensitivity of Scots pine trees to temperature in reclaimed areas was defined by stand characteristics, while the sensitivity to precipitation was dependent on soil characteristics and got stronger with increasing tree dimensions (Table 4 in III). Denser stands, characterised by a higher stand volume, relative density and basal area were sensitive to spring temperatures (r= ). Radial growth sensitivity to mean temperatures in January were negatively related to age and increased with site productivity, as defined by greater SI 50 and content of clay in topsoil. Trees growing in stands where thicker layers of the organic matter were apparent tended 51

53 to be less sensitive to the negative influence of temperatures in August. A significant negative relationship was established between sensitivity to rainfall in June and soil ph Temporal patterns of Scots pine response to climate variation Moving correlation functions calculated for 30-year windows and shifted by one year revealed the growth response to inter-annual weather variation dynamics over the period of (Fig. 5.6, Fig. A.2 in IV). Relationships between radial growth of Scots pine and winter and early spring temperatures were significant and quite stable over the last fifty years for mesic sites (Rhodococcum and Myrtillus) in the Southwest of Estonia and Islands, but associations with dry and nutrient poor sites of Cladonia and Calluna were weaker, and not always significant. Relationships to winter and spring temperatures for these sites were strongest at the beginning of analysis, but towards the twentieth century end associations with winter temperature became weaker (e.g. MyrtISL, MyrtSW, MyrtSE). Relationships between Scots pine growth in all forest site types (except Myrtillus) in the Northeast were weakly linked to cold season temperatures already in the middle of 20 th century. Positive linkage between growth and August precipitation of the previous year over studied period got stronger for pine on Cladonia site in the Northeast and Islands. Climate influence on radial growth was the least stable for all sites in the Southeast sub-region. 52

54 Fig Moving correlation functions between residual Scots pine chronologies and mean monthly air temperatures and precipitation sums for the period of , except ReclNE. The moving correlations were calculated using a 30-year moving window shifted by one year. Climatic variables are arranged from the previous June to the current August. Capital letters refer to current year (T - temperature, P - precipitation) and small letters (p, t) indicate climatic variables of the previous year. Numbers define calendar months (1 - January, 2 - February, etc.) Asterisk indicates significant correlations (p<0.05). 53

55 Fig Continued 54

56 Fig Continued Evidence of influences of extreme climatic conditions on radial growth A different number of years with exceptionally small or large radial growth was detected depending on the studied site and region (Fig. 5.7; Table 2 in II). The strongest growth reduction when Cropper values were below (extreme pointer years) was observed in most forest types and regions in 1940 and 1985, indicating large-scale extreme climatic conditions. 55

57 Fig Pointer years in Scots pine radial growth by forest type and sub-region. Grey bars show strong pointer years, C i >1.28 and black bars refer to extreme pointer years, C i > According to documented meteorological records (Tarand et al., 2013) summer of 1939 was warm and dry and followed by exceptionally cold winter of The winter of was also cold. The number of event and pointer years varied among the sites and regions due to different chronology lengths, therefore a greater number was detected for heath forests than for mesotrophic forests, because dry-forest chronologies were longer (II). 56

58 5.3. Radial growth patterns and the BAI model for Scots pine on reclaimed areas Radial growth patterns, expressed as the BAI, for all three sites on reclaimed oil shale mining areas (Fig. 5 in III) showed an exponential increase. Sudden increases in BAI patterns following thinning were observed on the oldest and the youngest sites (site 1 and site 3, respectively), whereas growth curve flattening appeared at middle-aged site (site 2) after the year A sharp growth reduction in was observed among the youngest trees, which was a consequence of the dry summer in BAI reduction was significantly lower for that year in older trees. However, pre-commercial thinning on the youngest site contributed to a fast radial growth recovery, as indicated by the sudden growth increase in As a response to thinning, the BAI of the remaining dominant trees on the oldest site increased by 58%, on average. In total, sixteen alternative models were fitted (Table 5.3 and Table 5 in III) on more than 3,600 annually resolved Scots pine increment data, expressed as the BAI, expanding over the period of (37 years). In addition, two models (M 15 M 16, including hypothetical thinning and without it) were compiled to test the effects of growth index substitution with climatic variables. The parameter estimates in all models were highly significant (p<0.0001). The explained variance by different combinations of independent variables ranged from 19% to 63%. The most important variables in our developed model were tree size, described by tree diameter, followed by the response to thinning and site factor. Tree diameter defined the asymptote better than SI 50 and explained 19% of the variance, therefore it was kept in the model. Fine soil depth was used to describe site quality for radial growth because it was a slightly better predictor than SI 50. In general, the difference between these two variables was marginal. The inclusion of the response to thinning improved model prediction power significantly and increased the explained variance by 14%. 57

59 Table 5.3. BAI model fit statistics and parameter estimates. p tree - tree size parameter, D max -tree DBH (cm); p th -response to thinning parameter, described as years since thinning ; p site - site productivity parameter described by either SI 50 site index at reference age of 50 years or fine soil - fine soil thickness (cm); f 1 (A)-basic grow function, f 2 (th)-response to thinning function, f 3 (av)-function describing annual variation related to climate; f 2 (hth)-response to thinning function with included hypothetical thinning. RMSE - root mean square error; general chron general Scots pine tree-ring width chronology for reclaimed areas; site chron tree-ring width chronologies for three study sites (site1, site 2, site 3); plot chron tree-ring width chronologies for twelve study plots, P JJ total precipitation in June-July (mm), T spring mean spring season temperature( C), AIC - Akaike Information Criterion; MEE - the mean error of estimate. The best models with and without hypothetical thinning is marked in bold. Model Components RMSE AIC pseudo-r 2 MEE M 1 f 1 (A),p tree M 2 f 1 (A),p tree, p site M 3 f 1 (A),p tree - D max, p site M 4 f 1 (A),p tree - SI 50, p site M 5 f 1 (A),p tree - D max, p site - SI M 6 f 1 (A),p tree - D max, p site - fine soil M 7 f 1 (A),p tree - D max, p site - fine soil * f 2 (th) M 8 f 1 (A),p tree - D max, p site - fine soil* f 2 (th)*f 3 (av) - general chron M 9 f 1 (A),p tree - D max, p site - fine soil* f 2 (th)*f 3 (av) - site chron M 10 f 1 (A),p tree - D max, p site - fine soil* f 2 (th)*f 3 (av) - plot chron M 11 f 1 (A),p tree - D max, p site - fine soil* f 2 (hth) M 12 f 1 (A),p tree - D max, p site - fine soil* f 2 (hth)*f 3 (av) - general chron M 13 f 1 (A),p tree - D max, p site - fine soil* f 2 (hth)*f 3 (av) - site chron M 14 f 1 (A),p tree - D max, p site - fine soil* f 2 (hth)*f 3 (av) - plot chron M 15 f 1 (A),p tree - D max, p site - fine soil* f 2 (th)*f 3 (av) - P JJ +T spring M 16 f 1 (A),p tree -D max, p site - fine soil* f 2 (hth)*f 3 (av) P JJ +T spring

60 The variable transformation was not applied during model compilation. The performance of alternative models was compared based on the Akaike Information Criterion (AIC), the root mean square error (RMSE), pseudo-r 2, the mean error of estimate (MEE) and analysis of residuals in relation to the observed data (data not presented). The model M 14 was able to explain the greatest variance (63%) and contained the smallest RMSE and AIC, therefore it was selected as the final one. By using this model, the annual BAI of Scots pine trees can be predicted based on tree diameter, mean soil depth in the stand and time since thinning. Furthermore, climatic variability could be accounted in the modelling process by replacing the growth index through the regression equation with the total precipitation in June July and mean spring temperature. The total explained variance was 57 59% for the models (M 15 M 16 ) including climatic variables. A visual comparison of the observed and predicted BAI curves shows a good reproduction of growth dynamics by the model (Fig. 5a and 5b, in III). In the beginning, the lack of fit was observed for the youngest sites, when the model was not able to predict a smaller peak, but the inclusion of hypothetical thinning improved the model performance significantly (Fig. 5b, Table 5, in III). The replacement of growth indices with climatic variables weakened the accuracy of model predictions just marginally, with the total explained variance reaching 59% (compared to 63% based on growth indices). 59

61 6. DISCUSSION 6.1. Tree growth deviations from expected growth The results obtained during compilation of this thesis provided evidence that height growth of Scots pine on Oxalis-Rhodococcum site in Southeastern Estonia is not following the trajectory of the reference growth (Figs. 2a, 3a, 3b and 4 in I) and supported the hypothesis (1) that pine stands with a later establishment date grow faster than those established several decades earlier. Because of the small sample size (only one forest site type was considered, in a single location), the findings from this study are of indicative nature, and can be valid only for trees growing in similar conditions as our studied stands. Differences in height growth patterns may not be the same for sites with a poorer nutritional status (Pretzsch et al., 2014) and may depend on the local climatic conditions (Sharma et al., 2012). Despite the limitations of result applicability, our study results are in agreement with previously reported findings where a representative data set was used and the SI increase for Scots pine and other tree species in Estonia was reported (Kiviste, 1999b; Nilson et al., 1999). Observed tree growth deviation from the expected growth in Estonia is not a phenomenon on its own. An increased growth of forest tree species and also a decline (Spiecker, 1996) have been reported in Europe (Kahle et al., 2008) and other parts of the world (Bontemps et al., 2009). For instance, Elfving and Tegnhammar (1996) studied Scots pine and Norway spruce growth in Sweden and found that the annual height and basal area growth rate over the period of increased about %. In other studies carried out in Sweden (Elfving and Nyström, 1996) and in Norway (Sharma et al., 2012), where growth trends in pine stands were explored by comparing the site index of plantations established in different years, researchers discovered a site index increase of 2.4 cm per annum over , with the most significant differences detected between stands established before and after The evaluation of height growth in our study was based on data from repeated measurements in the experimental plots, therefore height growth over time was approximated with two different site index functions. Some researchers (Elfving and Nyström, 1996) indicated that the results were dependent on the function selection. The findings from our study are based on actual annual height and radial growth increments, reconstructed from individual tree covering the full lifespan of the trees, therefore the results 60

62 are precise and do not include estimation bias related to the function choice (Elfving and Nyström, 1996; Subedi and Sharma, 2011). A similar increasing trend in the site index with a breakpoint in the 1930s in pine and spruce stands with different germination dates was observed by Sharma et al. (2012), with a significantly greater deviation in climatically warmer parts of Norway compared to the colder regions. Similarly to our study, Lebourgeois et al. (2000) compared the growth of young and old black pine stands in France, growing under the same climatic and soil conditions, and also concluded that younger stands grow faster. A study conducted by Pretzsch et al. (2014) used long-term experiment data and in parallel to evidence on increased forest stand growth, provided proof on accelerated forest growth dynamics in spruce and beech forests of Central Europe. To define the causes of the accelerated growth in the younger stands is a challenging task. The greater growth in the younger stands could be a result of changes in a site or long-term climatic conditions. By selecting pine stands growing on the same forest site type we attempted to exclude any differences in site conditions. However, considering that the assignment to a forest site type was based on the ground vegetation composition, some nutritional differences could still have persisted (Paal et al., 2010). While most of the studies referred above, attribute the observed deviations in forest tree growth to environmental changes, caused by the ongoing climate warming, Elfving and Tegnhammar (1996) and Elfving and Nyström (1996) do not exclude the possible influences on growth due to changes in management and suggest that stand tending at an early age (Kuliešis et al., 2010) and initial spacing (Elfving, 1975) or spacing shape (Gil, 2014) might have had a positive effect on height growth. The stand establishment and management differed among our studied stands, therefore more favourable growing conditions could have been created for pine stands established later (I). On the other hand, it is known for a long time that the top height growth and yields of the forest stands growing on the sites of comparable quality but in climatically different regions, are not the same, and differences are climatically conditioned. The comparison of mean climatic conditions that prevailed over the first 40 years for each cohort indicates warmer climatic conditions for middle-aged trees and significantly warmer for the youngest trees. Hy- 61

63 drological conditions remained almost the same, except for lower precipitation in winter over the period of when middle-aged trees grew (Fig. 5.2). Peltola et al. (2002) investigated the impacts of elevated temperature and CO 2 concentration under controlled conditions on diameter growth in 20-year-old Scots pine trees and discovered that under warmer conditions, the cumulative diameter growth over a three-year-period increased by 26% while higher CO 2 concentrations and greater temperatures enhanced pine radial growth up to 67% compared to the control group. In natural conditions, the assessment of climate change effects on tree growth is complicated, due to a large number of factors acting simultaneously. For elucidating the influence of climate, all the other factors should be kept constant, which is not straightforward bearing in mind the longevity of trees. Stem analysis allows to reconstruct individual tree growth, however, it fails to provide information on the tree s social status over its lifetime, therefore some uncertainty related to the tree s dominance remains. Site index deviations (I) from the expected trajectory indicate changes in productivity. It also shows that site index models are not able to predict the accelerating height growth in younger pine generations. Top-height growth models in the recent past have been developed assuming that the climate remains unchanged, therefore such models are unable to account for time dependent variability. Given that, tree stem volume is commonly not directly measured in the forest, but is rather calculated based on tree dimensions described by tree diameter and height. Accurate estimations of height development over time (SI) would enable more precise stem volume calculations (Zhang et al., 2004) Influence of annual climate variability on Scots pine growth in Estonia Scots pine radial growth response to annual climate fluctuations varied at the site level (III), along the ecological gradient (forest site types and forest types) (II, IV), among the regions (II, IV), suggesting that Scots pine sensitivity in Estonian forests is site-specific and depends on the local climatic conditions. Following the results of the studies in this thesis, late winter-early spring 62

64 temperatures and meteorological water availability during late summer prior to the ring formation season, are the main climatic factors driving Scots pine growth in Estonia (II IV). In general, the influence of temperature on pine growth is stronger than the importance of precipitation and this is in agreement with previous findings for the Baltic Sea region (Lõhmus, 1992; Läänelaid and Eckstein, 2003; Pärn, 2008; Hordo et al., 2011; Henttonen et al., 2014). Läänelaid et al. (2012) compared several Scots pine chronologies from the Baltic Sea region and confirmed that the control of winter NAO from December to March on Scots pine growth is characteristic of the studied region. However, negative influences of dry conditions, arising as a result of high temperatures and low precipitation at the end of the previous growing season, have not been extensively reported (Hordo et al., 2011). Warm late summers with low precipitation in August have been reported to be negatively related with ring-width for this species in the dust polluted areas (Pärn, 2003), but also for pines growing in the South and East Finland (Lindholm et al,. 2000) of Northern Europe. A limited number of sampling locations in Estonia could be an explanation for the low number of studies reporting this relationship. When investigating growth-climate relationships at a finer spatial scale (II, III, IV), it was also revealed that Scots pine response patterns to climate can be very diverse, especially at site level (Fig. 4 in III), even when the area studied is not very large. High diversity in responses was also obtained by Oberhuber et al. (1998) who studied Scots pine growth on sites exposed to dryness and Bauwe et al. (2013) who studied the climatic sensitivity of Scots pines growing on soils with varying water availability. In our case, Scots pine sensitivity to weather variability on reclaimed areas varied due to soil and stand structure conditions (III), pointing out that physical substrate characteristics and stand structure manipulations through management (thinning activities) are very important factors, defining the dependency on climatic conditions. Scots pine plantations, established on reclaimed areas of oil shale excavation in Estonia were analysed for the first time by applying dendrochronological techniques (III). Radial growth patterns of pine plantations on reclaimed areas (Fig. 5.3 and Fig.5.4) as well as the response to climatic variables, were distinctive from the pine populations growing on the natural forest sites (Fig. 5.5). Correlation analysis of the three Scots pine chronologies from reclaimed areas indicated that year-to-year variation in these areas can be quite dif- 63

65 ferent at the site level. In the HCA analysis (Fig. 5.4), a general residual chronology of reclaimed areas was segregated from other forest site type chronologies and could be assigned to a separate cluster, still bearing some resemblance to the chronologies from the same region. In boreal forests, the climate is one of the major factors controlling forest tree growth (Briffa et al., 1998), however, the climatic signal is weaker where optimal growth conditions for trees are met (Fritts, 1976). It has been reported by a number of studies that several climatic drivers simultaneously influence radial growth in temperate forests and climatic sensitivity tends to be lower compared to the relationships established for trees growing at the borders of a distribution area or on a site where extreme growth conditions persist. The final size and other features of tree rings is a result of environmental conditions that prevailed in the tree environment over the growing season and sometimes during the time before the start of the radial growth. According to Vaganov et al. (2006), the final width of an annual tree ring depends mainly on the xylem cell production rate. The time interval when cell division takes place during the growing season is considerably short, i.e. less than a month (Vaganov et al., 1999). For that reason, dendroclimatic calibrations make use of monthly climatic variables, aiming to establish stronger relationships. Climatic variables calculated on the basis of calendar months are one of the drawbacks of dendrochronological methods. In reality, phenological periods do not follow the calendar months and such use of resolution seems not to have biological meaning. Some researchers (Rathgeber et al., 2005) suggest calibrating radial growth to more bioclimatic variables, like temperature sums, indexes representing moisture availability. However, strong relationships may not always be guaranteed even if bioclimatic variables are used. Besides, some of the relationships established between monthly climate variables and radial growth during this thesis were not stable in time (IV), which makes them unsuitable for future growth predictions. Extreme yearly growth reductions caused by extreme climatic events slightly varied among the sites, however, low radial growth was observed in 1940 and 1985 across the sites and sub-regions of Estonia. Karpavičius (2001) noted that trees of different species tend to synchronise their growth after extreme climatic events. For instance due to the cold winter 64

66 in and in Lithuania, the radial growth of the common oak and Scots pine growing on soils with normal humidity decreased. He suggests that the growth dynamics and dependency on climatic conditions is preconditioned by soil characteristics and the depth of ground water. According to the results of the HCA analysis (Fig.5.4; Fig.5 in IV), applied on index series and PCA analysis conducted on BCC (Fig. 6 in IV), Scots pine populations investigated in our study could be divided into two groups, following the geographical sub-regions of Estonia. Similarities in growth patterns and growth responses were observed between the Islands and the Northeast sub-region and as well as between Southeast and Southwest sub-regions. Such groupings in dendroclimatological studies are common. For instance, in Lithuania Kairaitis and Karpavičius (1996) have proposed to distinguish between four types of growth-climate responses for oak. Despite the extensive Scots pine tree-ring sampling network in Estonia, some parts of the country still remain underrepresented and need further investigations to verify Scots pine responses to climate. The studies presented in this thesis have a low representation in sampling on the island of Saaremaa and in central Estonia. Also, not all the sites have a comparable number of replications for chronology building in the investigated sub-regions, therefore additional dendroclimatological studies should be carried out to make the tree-ring network more equally representative. Since growth-climate relationships change in time, temporal stability in growth-climate relationships and its causes should be studied in more detail. Therefore, the need for new dendroclimatic studies on Scots pine remains Modelling the basal area increment of Scots pine The study also aimed to incorporate the effect of thinning and climatic variation simultaneously into a basal area increment model of individual trees (III). A model with a multiplicative structure (Eq. (3)) consisting of three sub-components was selected to achieve the goal. Models of a similar structure have been used earlier (Quicke et al., 1994), and because each individual subcomponent of the model is fitted simultaneously to the data, overall model errors are minimised (Gadow and Hui, 1999). 65

67 The linear mixed effect model is the other commonly used approach for annual basal area increment modelling in individual trees (e.g upon the need to handle multilevel data). The modelled time step in the BAI model developed during the study is one year, which is appropriate for evaluating climatic forcing on growth, considering that increment forms over the calendar year. Besides that, tree response to different silvicultural measures can be easily simulated since the length of response may vary depending on site conditions, species composition and silvicultural treatments. In the current study, it was assumed that medium intensity thinning is applied to pine stands, therefore the thinning intensity was not differentiated but could be tested as proposed by Hynynen (1995) in subsequent studies. In models with the multiplicative structure, the potential growth function (basic function) aims to describe the maximum tree growth, whereas modifiers represent the restrictions commonly related to competition. In our case we targeted at describing the response to reduced competition. To avoid over-parameterisation, we maintained a low number of parameters in the model. For the same reason, the two-parameter Weber function was chosen to describe unrestricted tree growth. The function itself was able to explain 18% of total variation in the annual basal increment (M 1, M 2 in Table 5.3). Tree diameter was used to describe the asymptote because tree radial growth is size dependent (Wykoff, 1990) and it is able to explain a considerable amount of variation. According to Weiskittel et al. (2011) tree size already integrates competition to some extent. In addition to tree size, fine soil depth describing the potential site fertility was included in the model. The use of site-related variables enables using the model for mixed stands where site index estimation is complicated. However, if data about fine soil depth is not available, the site index could be reliably used in the model due to a marginal difference between the effect of using either one of these two variables. The inclusion of the thinning effect significantly improved predictions, and reduced prediction errors. However, climate related variables (index chronologies and climate variables) did not have the expected effect. This contradicted to earlier findings where climate-related variable inclusion improved the performance of the model. In our model, climate-related variables explained around ~3% of BAI variability but enabled the model to mimic annual fluctuations. The results of the present study 66

68 demonstrate the potential of using climate and management related information simultaneously for assessing the management alternatives in changing climatic conditions (Pukkala et al., 2009). However, the model still needs to be improved and verified on an independent data set to make the predictions more accurate and applicable in practice. 67

69 7. CONCLUSIONS 1. The comparison of growth patterns of three generations, growing in different time periods showed long-term changes in Scots pine growth (I). Therefore, the first hypothesis was proved. Height and radial growth rates of both studied Scots pine stands established after 1950 were greater compared to the tree growth that were established before 1900, resulting in different growth development over long-term. Consequently, SI 40 on average was 2.2 m (or 12%) greater for younger stands. While, causes of the accelerated growth are not very clear, meteorological data show that climatic conditions after 1970 was significantly warmer compared to the earlier periods, which could favour Scots pine growth. This means, that site index equations for Scots pine used nowadays might not be able to account for an accelerated tree growth and need to be re-developed. 2. The second hypothesis was proved. Scots pine response to climate in Estonia strongly depends on the local climatic conditions but also varies across the ecological site gradient (II, III, IV). The growth of Scots pine on mesic forest sites (Myrtillus, Rhodococcum), especially on the Islands and in the Southwest and Southeast are determined by winter and early spring (February April) temperatures. Whereas radial growth of Scots pine stands on all sites in the Northeast and on dry sites on the Islands are negatively related to high temperatures of the previous August and positively to the total rainfall of the same month, suggesting that the climatic water deficit during this month negatively influences next year s radial growth. 3. The third hypothesis was partly proved. Scots pine radial growth (expressed as the BAI) on reclaimed areas can be predicted with the annual resolution, using the following input variables: tree DBH, fine soil depth, time since thinning, and climate depicting variables precipitation sum of June July and mean spring temperature (III). The inclusion of the thinning effect to model behaviour was very important (R 2 15%) for radial growth, while the inclusion of climate-related variables did not have the expected effect and improve model predictions just marginally (R 2 3%). Our study shows the potential of using climate and management related information simultaneously, however model needs further improvements and verification. 68

70 4. Scots pine response to inter-annual weather variation was not stable in time in the case of all studied forest site types (IV), therefore the forth hypothesis was only partly proved. Relationships between radial growth and winter season temperatures got progressively weaker over period in all sites in the Southwest and Southeast and dry sites on the Islands. While associations with summer precipitation are becoming stronger especially for pine populations growing in the Northeast of Estonia and on the Islands. 69

71 REFERENCES Adams, H.R., Barnard, H.R., Loomis, A.K., Topography alters tree growth-climate relationships in a semi-arid forested catchment. Ecosphere, 5(11): 148. Ahas, R., Aasa, A., The effects of climate change on the phenology of selected Estonian plant, bird and fish populations. International Journal of Biometeorology, 51: Akaike, H., A new look at the statistical model identification. IEEE Transections on Automatic Control, 19(6): Andreassen, K., Tomter, S.M., Basal area growth models for individual trees of Norway spruce, Scots pine, birch and other broadleaves in Norway. Forest Ecology and Management, 180(1 3): Antonova, G.F., Stasova, V.V., Effects of environmental factors on wood formation in Scots pine stems. Trees, 7: Assmann, E., The Principles of Forest Yield Study. Pergamon Press, New York, 500 p. BACC Author Team, Assessment of Climate Change for the Baltic Sea Basin. Springer. 473 p. BACC Author Team, Second Assessment of Climate Change for the Baltic Sea Basin. Springer-Verlag, Berlin, 515 p. Bartelink, H.H., Effects of stand composition and thinning in mixed-species forests: a modeling approach applied to Douglas-fir and beech. Tree Physiology, 20: Bauwe, A., Koch, M., Kallweit, R., Konopatzky, A., Strohbach, B., Lennartz, B., Tree-ring growth response of Scots pine (Pinus sylvestris L.) to climate and soil water availability in the lowlands of northeastern Germany. Baltic Forestry, 19(2): Bauwe, A., Jurasinski, G., Scharnweber, T., Schröder, C., Lennartz, B., Impact of climate change on tree-ring growth of Scots pine, Common beech and Pedunculate Oak in northeastern Germany. IForest, 9: Beck, P.S.A., Juday, G.P., Alix, C., Barber, V.A., Winslow, S.E., Sousa, E.E., Heiser, P., Herriges, J.D., Goetz, S.J., Changes in forest productivity across Alaska consistent with biome shift. Ecology Letters, 14(4): Bergh, J., Linder, S., Lundmark, T., The effect of water and nutrient availability on the productivity of Norway spruce in northern and southern Sweden. Forest Ecology and Management, 119:

72 Biondi, F., Comparing tree-ring chronologies and repeated timber inventories as forest monitoring tools. Ecological Applications, 9(1): Biondi, F., Qeadan, F., A theory-driven approach to tree-ring standardization: Defining the biological trend from expected basal area increment. Tree-Ring Research, 64(2): Bitvinskas, T., Dendroclimatological research. Hidrometeoizdat, Leningrad, 172 p, [in Russian]. Boisvenue, C., Running, S.W., Impacts of climate change on natural forest productivity-evidence since the middle of the 20th century. Global Change Biology, 12: Bontemps, J.D., Hervé, J.C., Dhôte, J.F., Long-term changes in forest productivity: A consistent assessment in even-aged stands. Forest Science, 55(6): Bontemps, J.-D., Hervé, J.-C., Dhôte, J.-F., Dominant radial and height growth reveal comparable historical variations for common beech in north-eastern France. Forest Ecology and Management, 259: Bontemps, J.-D., Bouriaud, O., Predictive approaches to forest site productivity: recent trends, challenges and future perspectives. Forestry, 87(1): Bouriaud, O., Bréda, N., Dupouey, J.-L., Granier, A., Is ring width a reliable proxy for stem-biomass increment? A case study in European beech. Canadian Journal of Forest Research, 35(12): Box, G.E.P., Jenkins, G.M., Time Series Analysis: Forecasting and Control. Holden-Day, San Francisco, 553 p. Bravo-Oviedo, A., Roig, S., Bravo, F., Montero, G., del-rió, M., Environmental variability and its relationship to site index in Mediterranean maritime pine. Forest Systems, 20: Briffa K.R, Cook, E., Methods of response function analysis, in: Cook E.R., Kairiūkštis L.A. (Eds), Methods of dendrochronology. Applications in the environmental sciences, Kluwer Academic, Dordrecht, pp Briffa K.R., Jones P.D., Basic chronology statistics and assessment, in: Cook E.R., Kairiūkštis L.A. (Eds), Methods of dendrochronology. Applications in the environmental sciences, Kluwer Academic, Dordrecht, pp Briffa, K.R., Trees tell of past climates: but are they speaking less clearly today? Philosophical Transactions of the Royal Society B: Biological Sciences, 353(1365): Bunn, A.G., A dendrochronology program library in R (dplr). Dendrochronologia, 26(2):

73 Bunn, A., Korpela, M., Biondi, F., Campelo, F., Mérian, P., Qeadan, F., Zang, C., dplr: Dendrochronology Program Library in RR Package Version Burkhart, H.E., Status and future of growth and yield models. In LaBau V.J., Cunia T. (Eds), State-of-the-art methodology of forest inventory: a symposium proceedings. USDA Forest Service, Pacific Northwest Forest Experiment Station, Portland, General Technical Report, PNW-GTR-263, pp Burnham, K.P., Anderson, D.R., Model selection and multi-model inference. A practical information-theoretic approach. Springer, New York, 488 p. Büntgen, U., Frank, D., Kaczka, R., Verstege, A., Zwijacz-Kozica, T., Esper, J., Growth responses to climate in a multi-species tree-ring network in the Western Carpathian Tatra Mountains, Poland and Slovakia. Tree Physiology, 27(5): Carrer, M., Urbinati, C., Age-dependent tree-ring growth responses to climate in Larix decidua and Pinus cembra. Ecology, 85(3): Cedro, A., Dependence of radial growth of Pinus sylvestris L. from Western Pomerania on the rainfall and temperature conditions. Geochronometria, 20: Chhin, S., Hogg, E.H.(Ted), Lieffers, V.J., Huang, S., Potential effects of climate change on the growth of lodgepole pine across diameter size classes and ecological regions. Forest Ecology and Management, 256(10): Clutter, J.L., Fortson, J.C., Pienaar, L.V., Brister, G.H., Bailey, R.L., Timber Management: A Quantitative Approach. Wiley, New York. 335 p. Cook, E.R., Holmes, R.L., Users manual for program Arstan. In: Holmes, R.L., Adams, R.K., Fritts, H.C. (Eds), Tree-ring chronologies of western North America: California, eastern Oregon and northern Great Basin. Laboratory of Tree- Ring Research, University of Arizona. Tucson, Chronology Series Vol. 6, pp Cook, E.R., Kairiūkštis, L.A., Methods of Dendrochronology. Kluwer Academic Publishers, Dordrecht, The Netherlands, 394 p. Cook, E.R., Krusic, P.J., Program ARSTAN: A tree-ring standardization program based on detrending and autoregressive time series modeling, with interactive graphics. Lamont-Doherty Earth Observatory, Columbia University, Palisades, N.Y. Cook, E.R., Pederson, N., Uncertainty, emergence, and statistics in dendrochronology. In: Hughes, M.K., Swetnam, T.W., Diaz, H.F. (Eds) Dendroclimatology: progress and prospects, developments in paleoecological research. Springer, Berlin, pp

74 Cook, E.R., Peters, K., The smoothing spline: a new approach to standardizing forest interior tree-ring width series for dendroclimatic studies. Tree-Ring Bulletin, 41: Cropper, J.P., Tree-ring skeleton plotting by computer. Tree-ring Bulletin, 39: Dauškane, I., Elferts, D., Influence of climate on Scots pine growth on dry and wet soils near Lake Engure in Latvia. Estonian Journal of Ecology, 60(3): Deslauriers, A., Morin, H., Intra-annual tracheid production in balsam fir stems and the effect of meteorological variables. Trees, 19: Duncker, P.S., Barreiro, S.M., Hengeveld, G.M., Lind, T., Mason, W.L., Ambrozy, S., Spiecker, H., Classification of forest management approaches: a new conceptual framework and its applicability to European forestry. Ecology and Society, 17(4): Eastaugh, C.S., Kangur, A., Korjus, H., Kiviste, A., Zlatanov, T., Velichkov, I., Srdjevic, B., Srdjevic, Z., Hasenauer, H., Scaling Issues and Constraints in Modelling of Forest Ecosystems: a Review with Special Focus on User Needs. Baltic Forestry, 19(2): Eckenwalder, J.E., Conifers of the World: The Complete Reference, Portland, London, Timber Press, 720 p. Eckstein, D., Bauch, J., Beitrag zur Rationalisierung eines dendrochronologischen Verfahrens und zur Analyse seiner Aussagesicherheit. [Contribution to the rationalization of a dendrochronological procedure to the analysis of confidence], Forstwissenschaftliches Centralblatt, 88(1): , [In German]. Elfving, B., Volume and structure in unthinned stands of Scots pine. Royal College of Forestry. Research Notes, 35: Elfving, B., Nyström, K., Yield capacity of planted Picea abies in northern Sweden. Scandinavian Journal of Forest Research, 11: Elfving, B., Tegnhammar, L., Tveite, B., Studies on growth trends of forests in Sweden and Norway. In: Spiecker H, Mielikäinen, K., Kohl, M., Skovsgaard, J. (Eds) Growth trends in European forests: studies from 12 countries. Springer, Berlin, pp Erisman, J.W., de Vries, W., Nitrogen deposition effects on European forests. Environmental Reviews, 8: Esper, J., Niederer, R., Bebi, P., Frank, D., Climate signal age effects-evidence from young and old trees in the Swiss Engadin. Forest Ecology and Management, 255:

75 EUFORGEN, (accessed on January 26, 2016). Forest Act, (Amendment of 1 July 2015), English translation of the Estonian legal act, (accessed on February 15, 2016): Frank, D., Esper, J., Characterization and climate response patterns of a high-elevation, multi-species tree-ring network for the European Alps. Dendrochronologia, 22: Friedrichs, D., Trouet, V., Büntgen, U., Frank, D.C., Esper, J., Neuwirth, B., Löffler, J., Species-specific climate sensitivity of tree growth in Central-West Germany. Trees: Structure and Function, 23: Fritts, H.C., Tree Rings and Climate. Academic Press, New York, 567 p. Gadow, K.v., Hui., G., Modelling forest development. Kluwar Academic Publishers, Dordrecht, The Netherlands. 213 p. Gamache, I., Payette, S., Height growth response of tree line black spruce to recent climate warming across the forest-tundra of eastern Canada. Journal of Ecology, 92: Gil, W., The influence of initial spacing on growth and survival of Scots pine in 40 years period of cultivation in varied habitat conditions. Forest Research Papers, 75(2): Girard, F., Beaudet, M., Mailly, D., Messier, C., Integrating climatic response in competition dependent tree-level growth models for northern hardwoods. Forest Ecology and Management, 323: Girardin, M.P., Raulier, F., Bernier, P.Y., Tardif, J.C., Response of tree growth to a changing climate in boreal central Canada: A comparison of empirical, process-based and hybrid modelling approaches. Ecological Modelling, 213(2): González-García, M., Hevia, A., Majada, J., Calvo de Anta, R., Barrio-Anta, M Dynamic growth and yield model including environmental factors for Eucalyptus nitens (Deane & Maiden) Maiden short rotation woody crops in Northwest Spain. New Forests, 46(3): Gričar, J., Zupančič, M., Čufar, K., Koch, G., Schmitt, U., Oven, P., Effect of local heating and cooling on cambial activity and cell differentiation in the stem of Norway spruce (Picea abies). Annals of Botany, 97(6): Gričar, J., Zupančič, M., Čufar, K., Oven, P., Wood formation in Norway spruce (Picea abies) studied by pinning and intact tissue sampling method. Wood Research, 52(2):

76 Grissino-Mayer, H.D., Evaluating crossdating accuracy: a manual and tutorial for the computer program COFECHA. Tree-Ring Research, 57: Guillemot, J., Klein, E.K., Davi, H., Courbet, F., The effects of thinning intensity and tree size on the growth response to annual climate in Cedrus atlantica: a linear mixed modeling approach. Annals of Forest Science, 72: Guiot, J., 1991.The bootstrapped response function. Tree-Ring Bulletin, 51: Güney, A., Kerr, D., Sökücü, A., Zimmermann, R., Küppers, M., Cambial activity and xylogenesis in stems of Cedrus libani A. Rich at different altitudes. Botanical Studies, 56(1): 1 9. Hari, P., Arovaara, H., Raunemaa, T., Hautojärvi, A., Forest growth and the effects of energy production: a method for detecting trends in the growth potential of trees. Canadian Journal of Forest Research, 14(3): Hasenauer, H., Sustainable Forest Management: Growth Models for Europe. Springer, Berlin, 398 p. Helama, S., Lindholm, M., Meriläinen, J., Timonen, M., Eronen, M., Multicentennial ring-width chronologies of Scots pine along north-south gradient across Finland. Tree-Ring Research, 61: Henttonen, H., The dependence of annual ring indices on some climatic factors. Acta Forestalia Fennica, 186 p. Henttonen, H.M., Mäkinen, H., Peltoniemi, M., Laurén, A., Hordo, M., Response of radial increment variation of Scots pine to temperature, precipitation and soil water content along a latitudinal gradient across Finland and Estonia. Agricultural and Forest Meteorology, : Hordo, M., Henttonen, H., Helama, S., Mäkinen, H., Kiviste, A., Annual growth variation of Scots pine in Estonia and Finland. Baltic Forestry, 17 (1): Holmes, R.L., Computer-assisted quality control in tree-ring dating and measurement. Tree-Ring Bulletin, 43: Hynynen, J., Predicting the growth response to thinning for Scots pine stands using individual tree growth models. Silva Fennica, 29(2): IPCC, Climate change 2014: impacts, adaptation, and vulnerability. In: Barros, V.R., Field, C.B., Dokken, D.J. (Eds), Part B: Regional Aspects. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press Cambridge, United Kingdom and New York, 688 p. Huang, J.-G., Bergeron, Y., Berninger, F., Zhai, L., Tardif, J.C., Denneler, B.,

77 Impact of Future Climate on Radial Growth of Four Major Boreal Tree Species in the Eastern Canadian Boreal Forest. Plos One, 8(2): e Jaagus, J., Periodicity of precipitation in Estonia. Man and Nature, Talinn,pp Jaagus, J., Climatic fluctuations and trends in Estonia in the 20 th century and possible climate change scenarios. In: Kallaste, T., Kuldna, P., (Eds), Climate change studies in Estonia. Stockholm Environment Institute, Tallinn, pp Jaagus, J., Uusi andmeid Eesti kliimast. [New data about Estonian climate], Publicationes Instituti Geographici Universitatis Tartuensis, Tartu, 85: 28 38, [In Estonian]. Jaagus, J., Kliimamuutuse tendentsid Eestis 20. Sajandi teisel poolel seostatuna muutustega atmosfääri tsirkulatsioonis. [Climate change tendencies in Estonia on the second half of 20 th century in relation to shifts in atmospheric circulations] Publicationes Instituti Geographici Universitatis Tartuensis, Tartu, 93:62 78, [In Estonian]. Jaagus, J., 2006a. Climatic changes in Estonia during the second half of the 20th century in relationship with changes in large-scale atmospheric circulation. Theoretical and Applied Climatology, 83: Jaagus, J., 2006b. Trends in sea ice conditions in the Baltic Sea near the Estonian coast during the period 1949/ /2004 and their relationships to large-scale atmospheric circulation. Boreal Environmental Research, 11: Jaagus, J., Ahas, R., Space-time variations of climatic seasons and their correlation with the phenological development of nature in Estonia. Climate Research, 15: Jaagus, J., Mändla, K., Climate change scenarios for Estonia based on climate models from the IPCC Fourth Assessment Report. Estonian Journal of Earth Sciences, 63: Jones, P.D., Mann, M.E., Climate over past millennia, Reviews of Geophysics, 42: RG2002. Jõgiste, K., A basal area increment model for Norway spruce in mixed stands in Estonia. Scandinavian Journal of Forest Research, 15(1): Kaimre, P., Männipuistute väärtusest ja majandamise tasuvusest [The profitability of pine stand management] In: Kurm, M., (Ed), Mänd Eestis, Vali Press OÜ, Tartu, pp , [In Estonian with English summaries]. Kahle, H.P., Karjalainen, T., Schuck, A., Ågren, G.I., (Eds), Causes and consequences of forest growth trends in Europe-Results of the RECOGNITION Project. EFI Research Report 21, Brill Leiden, Boston, Köln. 76

78 Kanninen, M., Hari, P., Kelloma ki, S., A dynamic model for above-ground growth and dry matter production in a forest community. Journal of Applied Ecology, 19: Karlsson, K., Height growth patterns of Scots pine and Norway spruce in the coastal areas of western Finland. Forest Ecology and Management, 135: Karpavičius, J., Yadav, R., Kairaitis, J., Radial growth responses of pine (Pinus sylvestris L.) and spruce (Picea abies L. Karst.) to climate and geohydrological factors. Paleobotanist, 45: Karpavičius, J., Specifics of tree growth in Lithuania and its dependence on various factors. Paleobotanist, 50: Kasesalu, H., Andres Mathiesen-elu ja töö, [Andres Mathiesen-life and work], Estonian Society of Foresters,158 p, [In Estonian]. Kiviste, A., Eesti riigimetsa puistute kõrguse, diameetri ja tagavara vanuseridade diferentsmudel a. metsakorralduse takseerkirjelduste andmeil, [Algebraic difference equations for stand height, diameter, and volume in Estonian state forest based on measurement data]. EPMÜ teadustööde kogumik, Eesti Põllumajandusülikool, Tartu, pp , [In Estonian]. Kiviste, A., 1999a. Eesti puistute kasvumudelitest. Pidev metsa korraldus. [Regarding forest stands models in Estonia. Continuous forest management planning ]. EPMÜ Metsandusteaduskonna toimetised. (32): 28 36, [In Estonian]. Kiviste, A., 1999b. Site index change in the 1950s-1999s according to Estonian forest inventory data. In: Karjalainen, T., Spiecker, H., Laroussinie, O. (Eds), Causes and Consequences of Accelerating Tree Growth in Europe. EFI Proceedings, 27: Kiviste, A., Álvarez Gonzáles, J.G., Rojo Alboreca, A., Ruiz Gonzáles, A.D., Funciones de crecimiento de aplocacion en el ámbito forestal, [Growth functions applied in forestry], Instituto Nacional de Investigacion y Tecnología Agraria y Alimentaria Ministerio de Ciencia y Tecnología, INIA, Madrid, 190 p., [In Spanish]. Kiviste, A., Kiviste, K., Algebraic Difference Equations for Stand Height, Diameter, and Volume Depending on Stand Age and Site Factors for Estonian State Forests. International Journal of Mathematical and Computational Forestry and Natural-Resource Sciences, 1(2): Kiviste, A., Hordo, M., Kangur, A., Kardakov, A., Laarmann, D., Lilleleht, A., Metslaid, S., Sims, A., Korjus, H., Monitoring and modeling of forest ecosystems: the Estonian Network of Forest Research Plots. Forestry Studies, 62(1):

79 Kilpeläinen, A., Peltola, H., Rouvinen, I., Kellomäki, S., Dynamics of daily height growth in Scots pine trees at elevated temperature and CO 2. Trees: Structure and Function, 20(1): Knorr, W., Is the airborne fraction of anthropogenic CO 2 emissions increasing? Geophysical Research Letters, 36 (21): L Kramer, P.J., Kozlowski T.T., Physiology of woody plants. Academic Press, Orlando, Florida, 881 p. Krigul, T., Metsataksaatori teatmik. [Forest inventor s handbook], EPA rotaprint, Tartu, 139 p, [In Estonian]. Kuliešis, A., Saladis, J., Kuliešis, A.A., Development and productivity of young scots pine stands by regulating density. Baltic Forestry, 16(2): Larcher, W., Physiological Plant Ecology: Ecophysiology and Stress Physiology of Functional Groups. 4 th edition. Springer-Verlag Heidelberg, New York, Germany, 517 p. Lebourgeois, F., Becker, M., Chevalier, R., Dupouey, J.-L., Gilbert, J.-M., Height and radial growth trends of Corsican pine in western France. Canadian Journal of Forest Research, 30(5): Lebourgeois, F., Eberlé, P., Mérian, P., Seynave, I., Social status-mediated treering responses to climate of Abies alba and Fagus sylvatica shift in importance with increasing stand basal area. Forest Ecology and Management, 328: Legendre, P., Legendre, L., Numerical Ecology. Elsevier, New York, USA, 1006 p. Lévesque, M., Walthert, L., Weber, P., Soil nutrients influence growth response of temperate tree species to drought. Journal of Ecology, 104(2): Lindholm, M., Lehtonen, H., Kolström, T., Meriläinen, J., Eronen, M., Timonen, M., Climatic signals extracted from ring-with chronologies of Scots pines from the northern, middle and southern parts of the boreal forest belt in Finland. Silva Fennica, 34(4): Lindner, M., Maroschek, M., Netherer, S., Kremer, A., Barbati, A., Garcia-Gonzalo, J., Seidl, R., Delzon, S., Corona, P., Kolström, M., Lexer, M.J., Marchetti, M., Climate change impacts, adaptive capacity and vulnerability of European forest ecosystems. Forest Ecology and Management, 259(4): Lindner, M., Fitzgerald, J.B., Zimmermann, N.E., Reyer, C., Delzon, S., van der Maaten, E., Schelhaas, M.-J., Lasch, P., Eggers, J., van der Maaten-Theunissen, M., Suckow, F., Psomas, A., Poulter, B., Hanewinkel, M., Climate change and European forests: What do we know, what are the uncertainties, and what are 78

80 the implications for forest management? Journal of Environment Management, 146: Läänelaid, A., Radial increment of bog pines and climatic changes. International Biological Programme, 9: Läänelaid, A., Eckstein, D., Development of a tree-ring chronology of Scots pine (Pinus sylvestris L.) for Estonia as a dating tool and climatic proxy. Baltic Forestry, 9: Läänelaid, A., Helama, S., Kull, A., Timonen, M., Jaagus, J Common growth signal and spatial synchrony of the chronologies of tree-rings from pines in the Baltic Sea region over the last nine centuries. Dendrochronologia, 30(2): Lõhmus, E., Hariliku männi radiaalkasvu seosest meteoroloogiliste teguritega [The dependence of Scots pine annual ring indices on some climatic factors]. Forestry Studies, 25: 50 59, [In Estonian]. Lõhmus E., Eesti metsa kasvukohatüübid, [Estonian forest site types]. Eesti Loodusfoto, Tartu, 80 p., [in Estonian]. Martín-Benito, D., Del Río, M., Heinrich, I., Helle, G., Cañellas, I., Response of climate-growth relationships and water use efficiency to thinning in a Pinus nigra afforestation. Forest Ecology and Management, 259(5): Mazza, G., Cutini, A., Manetti, M.C., Site-specific growth responses to climate drivers of Pinus pinea L. tree rings in Italian coastal stands. Annals of Forest Science, 71(8): Mérian, P., Bontemps, J.D., Bergès, L., Lebourgeois, F., Spatial variation and temporal instability in climate-growth relationships of sessile oak (Quercus petraea [Matt.] Liebl.) under temperate conditions. Plant Ecology, 212: Menzel, A., Jakobi, G., Ahas, R., Scheifinger, H., Estrella, N., Variations of the climatological growing season ( ) in Germany compared with other countries. International Journal of Climatology, 23: Monkhouse., F.J., A Dictionary of Geography. Edward Arnold Publishers Ltd., London, 320 p. Monserud, R.A., Sterba, H., A basal area increment model for individual trees growing in even- and uneven-aged forest stands in Austria. Forest Ecology and Management, 80(1-3): Morison, J.I.L., Lawlor, D.W., Interactions between increasing CO 2 concentration and temperature on plant growth. Plant Cell and Environment, 22:

81 Mosteller, F., Tukey, J.W., Data analysis and regression. Addison-Wesley, Reading, Mass, 588 p. Mäkelä, A., Process-based modelling of tree and stand growth: towards a hierarchical treatment of multiscale processes. Canadian Journal of Forest Research, 33(3): Mäkelä, A., Hari, P., Berninger, F., Hänninen, H., Nikinmaa, E., Acclimation of photosynthetic capacity in Scots pine to the annual cycle of temperature. Tree Physiology, 24(4): Mäkinen, H., The suitability of height and radial increment variation in Pinus sylvestris (L.) for expressing environmental signals. Forest Ecology and Management, 112(1-2): Mäkinen, H., Nöjd, P., Saranpää, P., Seasonal changes in stem radius and production of new tracheids in Norway spruce. Tree Physiology, 23(14): Murtagh, F., Legendre, P., Ward s hierarchical agglomerative clustering method: which algorithms implement Ward s criterion? Journal of Classification, 31: Navasaitis, M., Dendrologija, [Dendrology]. Margi Raštai, Vilnius, 856 p. [In Lithuanian]. Neuwirth, B., Schweingruber, F.H., Winiger, M., Spatial patterns of central European pointer years from 1901 to Dendrochronologia, 24(2-3): Nilson, A., Kiviste, A., Männikute kasvukäigu mudel tüpiseerimata kasvukohatingimuste järgi. [Pine stand growth model on undefined site type.] In: Laas, E., Kurvits, P., Pihelgas, E., Nilson, A. (Eds), Eesti Põllumajanduse Akadeemia teaduslike tööde kogumik 151. Eesti Põllumajanduse Akadeemia, Tartu, pp , [in Estonian with Russian and English summaries]. Nilson, A., Kiviste, A., Reflection of environmental changes in models of forest growth composed using different methods. In: Antanaitis, V.V., Baltrunaitis, R.J., Mastauskis, M.M., Juknis, R.A. (Eds) Monitoring of forest ecosystems. Abstracts of scientific conference, June 5 6, Akademija, Kaunas, pp , [in Russian]. Nilson, A., Kiviste, A., Korjus, H., Mihkelson, S., Etverk, I., Oja, T., Impact of recent and future climate change on Estonian forestry and adaptation tools. Climate Research, 12(2-3): Nilson, A., Puistute kõrguse kasvu üldistus boniteerimiseks sobivate diferentsmudelitena. Forestry Studies, 43: , [In Estonian with English summary ]. Niu, S., Li, Z., Xia, J., Han, Y., Wu, M., Wan, S., Climatic warming changes 80

82 plant photosynthesis and its temperature dependence in a temperate steppe of northern China. Environmental and Experimental Botany, 63: Nunes, L., Patricio, M., Tomé J., Tomé, M., Modeling dominant height growth of maritime pine in Portugal using GADA methodology with parameters depending on soil and climate variables. Annals of Forest Science, 68: Nyström, K., Kexi, M., Individual tree basal area growth models for young stands of Norway spruce in Sweden. Forest Ecology and Management, 97(2): Oberhuber, W., Stumböck, M., Kofler, W Climate-tree-growth relationships of Scots pine stands (Pinus sylvestris L.) exposed to soil dryness. Trees, 13: Oberhuber, W., Kofler, W., Topographic influences on radial growth of Scots pine (Pinus sylvestris L.) at small spatial scales. Plant Ecology, 146: Paal, J., Rajandu, E., Köster, T., Vegetation-Environment Relationship in Estonian Hepatica Site Type Forests in the Light of A.K. Cajander s Forest Site Type Approach. Baltic Forestry, 16 (2): Peltola, H., Kilpeläinen, A., Kellomäki, S., Diameter growth of Scots pine (Pinus sylvestris) trees grown at elevated temperature and carbon dioxide concentration under boreal conditions. Tree Physiology, 22: Peng, C., Growth and yield models for uneven-aged stands: Past, present and future. Forest Ecology and Management, 132(2-3): Plomion, C., Leprovost, G., Stokes, A., Wood formation in trees. Plant Physiology, 127: Pokharel, B., Dech, J.P., Mixed-effects basal area increment models for tree species in the boreal forest of Ontario, Canada using an ecological land classification approach to incorporate site effects. Forestry, 85(2): Pourtahmasi, K., Parsapajouh, D., Bräuning, A., Esper, J., Schweingruber, F.H., Climatic analysis of pointer years in tree-ring chronologies from Northern Iran and neighboring high mountain areas. Geoöko, 28: Porté, A., Bartelink, H.H., Modelling mixed forest growth: a review of models for forest management. Ecological Modelling, 150(1-2): Pretzsch, H., Grote, R., Reineking, B., Rötzer, T., Seifert, S., Models for forest ecosystem management: A European perspective. Annals of Botany, 101(8): Pretzsch, H., Forest dynamics, growth and yield. From measurement to model, Springer, Berlin, 664 p. Pretzsch, H., Biber, P., Schütze, G., Uhl, E., Rötzer, T., Forest stand growth 81

83 dynamics in Central Europe have accelerated since Nature Communications, 5: Pärn, H., Radial growth response of Scots pine to climate under dust pollution in Northeast Estonia. Water Air and Soil Pollution, 144: Pärn, H., Männipuistute radiaalkasvust muutuvates keskkonnatingimustes [The radial growth of pine stands in changing environmental conditions]. Forestry Studies, 48: 41 52, [In Estonian]. Pukkala, T., Lähde, E., Laiho, O., Growth and yield models for uneven-sized forest stands in Finland. Forest Ecology and Management, 258: Quicke, H.E., Meldahl, R.S., Kush, J.S., Basal area growth of individual trees: a model derived from a regional longleaf pine growth study. Forest Science, 40(3): Rais, A., van de Kuilen, J.-W.G., Pretzsch, H., Growth reaction patterns of tree height, diameter, and volume of Douglas-fir (Pseudotsuga menziesii [Mirb.] Franco) under acute drought stress in Southern Germany. European Journal of Forest Research, 133(6): Rathgeber, C.B.K., Cuny, H.E., Fonti, P., Biological Basis of Tree-Ring Formation: A Crash Course. Frontiers in Plant Science, 7(May): 1 7. Raudsaar, M., Pärt, E. Adermann, V., Forest resources. In: Yearbook Forest 2014, Keskkonnaagentuur, Tallinn, pp R Core Team, R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL org/. Richards, F.J., A flexible growth function for empirical use. Journal of Experimental Botany, 10: Ritchie, M.W., Hamann, J.D., Individual-tree height-, diameter- and crownwidth increment equations for young Douglas-fir plantations. New Forests, 35(2): Rossi, S., Deslauriers, A., Anfodillo, T., Morin, H., Saracino, A., Motta, R., Borghetti, M., Conifers in cold environments synchronize maximum growth rate of tree-ring formation with day length. New Phytologist, 170(2): Rossi, S., Deslauriers, A., Gričar, J., Seo, J.-W., Rathgeber, C.B., Anfodillo, T., Morin, H., Levanič, P.O., Jalkanen, R., Critical temperatures for xylogenesis in conifers of cold climates. Global Ecology and Biogeography, 17(6): Salminen, H., Jalkanen, R., Modelling the effect of temperature on height increment of Scots pine at high latitudes. Silva Fennica, 39(4):

84 Salminen, H., Jalkanen, R., Lindholm, M., Summer temperature affects the ratio of radial and height growth of Scots pine in northern Finland. Annals of Forest Science, 66: Sarvas, R., Investigations on the annual cycle of development of forest trees. Active period. Communicationes Instituti Forestalis Fenniae, 76: Schweingruber, F.H., Tree rings and environment. Basics and Applications of Dendrochronology. D.Reidel Publishing Company, Dordrecht, 276 p. Schweingruber, F.H., Eckstein, D., Serre-Bachet, F., Bräker, O.U., Identification, presentation and interpretation of event years and pointer years in dendrochronology. Dendrochronologia, 8: Schweingruber, F.H., Tree Rings and Environment. Dendroecology. Birmensdorf, Swiss Federal Institute for Forest, Snow and Landscape Research. Berne, Stuttgart, Vienna, Haupt. 609 p. Sibul, I., Männi üldiseloomustus [General characterization of the pine]. In: Kurm, M., (Ed), Mänd Eestis, Vali Press OÜ, Tartu, pp , [In Estonian with English summaries]. Seo, J.-W., Eckstein, D., Schmitt, U., The pinning method : From pinning to data preparation, Dendrochronologia, 25: Seo, J.W., Salminen, H., Jalkanen, R., Eckstein, D., Chronological coherence between intra-annual height and radial growth of Scots pine (Pinus sylvestris L.) in the Northern Boreal Zone of Finland. Baltic Forestry, 16(1): Sharma, R.P., Brunner, A., Eid, T., Site index prediction from site and climate variables for Norway spruce and Scots pine in Norway. Scandinavian Journal of Forest Research, 27(7): Shugart, H.H., Smith, T.M., Post, W.M., The application of individual-based simulation models for assessing the effects of global change. Annual Review of Ecology and Systematics, 23: Skovsgaard, J.P., Vanclay, J.K., Forest site productivity: A review of the evolution of dendrometric concepts for even-aged stands. Forestry, 81(1): Spiecker, H., Mielikäinen, K., Köhl, M., Skovsgaard, J.P., (Eds), Growth Trends of European Forests: Studies from 12 Countries. Springer-Verlag, Berlin, 371 p. Spiecker, H., Tree rings and forest management in Europe. Dendrochronologia, 20(1-2): Stephenson, N.L., Climatic control of vegetation distribution : the role of the water balance. The American Naturalist, 135(5):

85 Steven, H.M., Carlisle, A., The native pinewoods of Scotland. Oliver and Boyd Publications, Edinburgh. 368 p. Stokes, M.A., Smiley, T.L., An introduction to tree-ring dating. University of Chicago Press, Chicago, USA. 73 p. Subedi, N., Sharma, M., Individual-tree diameter growth models for black spruce and jack pine plantations in northern Ontario. Forest Ecology and Management, 261(11): Špalte, E.P., Influence of meteorological factors on the radial growth of pine in Latvian SSR. Lesovedenie, 3: 11 18, [in Russian]. Tammets, T., Jaagus, J., Sepp, M., Ekstreemsed saju- ja kuivaperiodid Eestis, [Periods with extreme rainfalls and droughts in Estonia]. Publicationes Instituti Geographici Universitatis Tartuensis, Tartu, 109: 23 50, [In Estonian]. Tammets, T., Jaagus, J., Climatology of precipitation extremes in Estonia using the method of moving precipitation totals. Theoretical and Applied Climatology, 111: Tarand, A., Jaagus, J., Kallis, A., Eesti kliima minevikus ja tänapäeval [Climate in Estonia in the past and nowadays]. Tartu Ülikooli kirjastus, Tartu, 632 p., [In Estonian]. Tappo, E., Eesti NVS puistute keskmised takseertunnused puistu enamuspuuliigi, boniteedi ja vanuse järgi [Mean forest stand characteristics of Estonian NVS forests by species, productivity class and age], Eesti Metsakorralduskeskus, 72 p., [In Estonian]. Teuffel, K., Hein, S., Kotar, M., Preuhsler, E.P., Puumanainen, J., Weinfurter, P., End user needs and requirements. In: Hasenauer, H. (Ed), Sustainable Management, Growth Models for Europe. Springer, Berlin Heidelberg, pp Untheim, H., Has site productivity changed? A case study in the Eastern Swabian Alps, Germany. In: Spiecker, H., Mielikäinen, K., Köhl, M., Skovsgaard, J.P., (Eds), Growth trends in European forests. Springer, Berlin, pp Uvo, C.B., Analysis and regionalization of northern European winter precipitation based on its relationship with the North Atlantic Oscillation. International Journal of Climatology, 23(10): Valinger, E., Elfving, B., Mörling, T., Twelve-year growth response of Scots pine to thinning and nitrogen fertilisation. Forest Ecology and Management, 134(1-3): Vaganov, E.A., Hughes, M.K., Kirdyanov, A.V., Schweingruber, F.H., Silkin, P.P., Influence of snowfall and melt timing on tree growth in subarctic Eurasia. Nature, 400(6740):

86 Vaganov, E.A., Hughes, M.K., Shaskin, A.V., Growth Dynamics of Conifer Tree Rings: Images of Past and Future Environments, Springer-Verlag, Berlin, Heidelberg, New York. 358 p. Vanclay, J.K., Modelling forest growth and yield : applications to mixed tropical forests, CAB International, Wallingford, UK. 312 p. Vaughn, N.R., Turnblom, E.C., Ritchie, M.W., Bootstrap evaluation of a young Douglas-fir height growth model for the Pacific Northwest. Forest Science, 56(6): Vitas, A., Dendroclimatological research of Scots pine (Pinus sylvestris L.) in the Baltic coastal zone of Lithuania. Baltic Forestry, 10(1): van der Maaten-Theunissen, M., van der Maaten, E., Bouriaud, O., pointres: An R package to analyze pointer years and components of resilience. Dendrochronologia, 35: Weber, P., Bugmann, H., Rigling, A., Radial growth responses to drought of Pinus sylvestris and Quercus pubescens in an inner- Alpine dry valley. Journal of Vegetation Science, 18(6): Weiskittel, A.R., Hann, D.W., Kershaw, J.A., Vanclay, J.K., Components of treelist models. Forest Growth and Yield Modeling. John Wiley & Sons Ltd., 415 p. Wigley, T.M.L., Briffa, K.R., Jones, P.D., On the average value of correlated time series, with applications in dendroclimatology and hydrometeorology. Journal of Applied Meteorology, 23: Williams, A.P., Michaelsen, J., Leavitt, S.W., Still, C.J., Using tree rings to predict the response of tree growth to climate change in the continental United States during the twenty-first century. Earth Interactions, 14(19): Wood, S.N., Generalized Additive Models: An Introduction with R. Chapman and Hall/CRC Press, Boca Raton, Florida. 384 p. Wykoff, W.R., A basal area increment model for individual conifers in the northern Rocky Mountains. Forest Science. 36(4): Xu, L., Myneni, R.B., Chapin III, F.S., Callaghan, T.V., Pinzon, J.E., Tucker, C.J., Zhu, Z., Bi, J., Ciais, P., Tømmervik, H., Euskirchen, E.S., Forbes, B.C., Piao, S.L., Anderson, B.T., Ganguly, S., Nemani, R.R., Goetz, S.J., Beck, P.S.A., Bunn, A.G., Cao, C., Stroeve, J.C., Temperature and vegetation seasonality diminishment over northern lands. Nature Climate Change, 3: 1 6. Zang, C., Pretzsch, H., Rothe, A., Size-dependent responses to summer drought in Scots pine, Norway spruce and Common oak. Trees: Structure and Function, 26(2):

87 Zang, C., Biondi, F., Dendroclimatic calibration in R: The bootres package for response and correlation function analysis. Dendrochronologia 31(1) : Zang, C., Biondi, F Treeclim: An R package for the numerical calibration of proxy-climate relationships. Ecography, 38(4): Zhang, L., Bi, H., Cheng, P., Davis, C.J., Modeling spatial variation in tree diameter-height relationships. Forest Ecology and Management, 189(1 3):

88 SUMMARY IN ESTONIAN KLIIMA MÕJU HINDAMINE HARILIKU MÄNNI (Pinus sylvestris L.) KASVULE EESTIS Sissejuhatus Metsanduslike tegevuste kavandamine tugineb tänapäeval olulisel määral puistute kasvumudelite kasutamisele (Hasenauer, 2006). Seetõttu on vajalik, et koostatud mudelid oleksid võimalikult täpsed ja usaldusväärsed ning aitaksid võtta vastu metsamajanduslikke otsuseid ka muutuvate kliimatingimuste korral. Sarnaselt paljudes teistes riikides kasutusel olevate empiiriliste kasvumudelitega (Assmann, 1970) on ka Eestis nii teaduslikus kui ka praktilises metsanduses seni olnud kasutusel puistu kasvufunktsioonid ja -mudelid (Krigul, 1969; Tappo, 1982; Kiviste 1999a, 1999b; Kasesalu, 2003; Nilson, 2005; Kiviste ja Kiviste, 2009), mis on välja töötatud valdavalt eelmise sajandi metsade kasvuandmetel põhinevalt. Praeguseks on aga jõutud olukorda, kus puude kasvukeskkond, sealhulgas nii metsamullad kui ka kliima, on varasemaga võrreldes oluliselt muutunud, seetõttu ei ole puistute kasvu prognoosimine olemasolevate mudelitega enam piisavalt usaldusväärne. Globaalset kliima muutumist on mõõtmisandmetele põhinevalt kirjeldatud alates 1950ndatest aastatest üle kogu maailma (Jones ja Mann, 2004) ja sarnaste muutuste täheldamine on nii Läänemeremaades (BACC II Autor Team, 2008) kui ka Eestis (Jaagus, 1998, 1999; Jaagus, 2006a, 2006b; Tarand jt, 2013) toonud avalikkuse tähelepanu alla toimuvad pikaajalised muutused keskkonnatingimustes. Alates 20. sajandi teisest poolest on Eestis kliima olnud tunduvalt soojem kui varasematel perioodidel. Eesti erinevates paikades on aasta keskmine temperatuur tõusnud 1,0 1,7 C (Jaagus, 2006a; Tarand jt, 2013). Järk-järgult on talvised keskmised temperatuurid tõusnud ja lumekattega periood jäänud lühemaks, mille tulemusena on ka vegetatsiooniperioodi algus nihkunud varasemaks (Ahas ja Aasa, 2006). Sademerežiimi muutumisele on Eestis iseloomulik järjest kasvav ruumiline ja ajaline varieeruvus, seejuures sademete hulga suurenemise tendents on täheldatav eelkõige külmal aastaajal (oktoober märts) ja suvel juunis. Piirkondliku kliimamudeli (Jaagus ja Mändla, 2014) prognoosid ennustavad kliima soojenemise jätkumist vähemalt käimasoleva sajandi lõpuni. Samuti on mõnedes Eesti piirkon- 87

89 dades sademete kasvu 10 20% võrra talvisel perioodil ning varasemaga võrreldes vähesema sademete hulgaga perioodi suve keskpaigas. Temperatuur ja sademed avaldavad vahetut mõju peamistele puude füsioloogilistele protsessidele nagu fotosüntees, hingamine, toitainete ja vee transport (Morison ja Lawlor, 1999; Niu jt, 2008) ning mõjutavad ka kasvuks olulist kambiumi aktiivsust ning ksüleemi rakkude jagunemiskiirust ja dünaamikat. Viimati nimetatud protsess määrab omakorda puude aastasisese kasvurütmi ja kasvuperioodil toodetud puidu koguse. Pikaajalised soodsad või ebasoodsad (puudele stressi tekitavad) ilmastikutingimused mõjutavad vahetult üksikpuude aastasisest kasvu (positiivselt või negatiivselt), suunates sellega metsade kasvudünaamikat ja tootlikkust laiemalt (Beck jt, 2011). Kliimamuutuse mõju metsa kasvule prognoositakse mudelitega, mis võimaldavad arvestada seoseid puudes toimuvate protsesside ja muutuvate kasvutingimuste vahel ning võimaldavad nende seoste rakendamist koos teiste kasvudünaamikat kirjeldavate tunnustega. Sellest tulenevalt on vajalik hinnata erinevate kliimatingimuste mõju puude kasvule ja tuvastada need ilmastikunäitajad, mida on võimalik kasutada puistu kasvumudelites. Mahukat puude aastarõngaste mõõtmisandmestikku ning varasemaid meteoroloogiliste mõõtmiste andmeid kasutav dendroklimaatiline analüüs võimaldab puistute kasvu kirjeldamiseks välja sõeluda olulise mõjuga ilmastikunäitajad ning kirjeldada nende tunnuste omavahelist mõju ja sesoonset ajastatust. Kliimatingimuste pikaajalisest muutumisest sõltuvalt on oluline kontrollida ka teadaolevate empiiriliste seoste ajalist püsivust ja ruumilist muutlikkust. Käesoleva doktoritöö fookuses on hariliku männi (Pinus sylvestris L.) kasvuanalüüs, kuna mänd on Eesti metsades enim levinud ning ühtlasi ka majanduslikult ja ökoloogiliselt olulisem puuliik. Doktoritöö peamine eesmärk oli uurida ilmastiku pikaajalise varieerumise mõjusid hariliku männi kasvule, tehes seda läbi puude radiaaljuurdekasvu ja kliimatunnuste vaheliste seoste kirjeldamise. Doktoritöös uuritakse ka kliimanäitajate kasutamise võimalusi puude kasvu täpsemaks modelleerimiseks ja puude kõrguskasvu pikaajalise dünaamika muutumist. 88

90 Doktoritöö detailsemad eesmärgid olid: 1. analüüsida pikaajalisi muutusi hariliku männi kasvus Eestis seotult muutustega kasvukeskkonna keskmises temperatuuris ja sademete režiimis (I); 2. dendrokronoloogilisi meetodeid kasutades selgitada välja peamised kliimanäitajad, mis mõjutavad hariliku männi radiaaljuurdekasvu aastast varieeruvust erinevate kliima- ja kasvukoha tingimuste korral Eestis (II, III, IV); 3. analüüsida hariliku männi üksikpuude aastast radiaaljuurdekasvu ja hinnata, kas harvendusraie ja ilmastikunäitajate lisamine parandab puistu rinnaspindala kasvu prognoosimudelit (III); 4. uurida lokaalsete kliimatingimuste muutumisest tulenevat puude kasvureageeringu stabiilsust ajas (IV). Doktoritöös püstitati järgnevad hüpoteesid: 1. Sarnastes kasvukohtades kasvavad kuid kronoloogiliselt hilisema rajamisaastaga (20 saj. teisel poolel) hariliku männi puistud on kiirema kasvuga kui puistud, mis on rajatud aastakümneid varem (20 saj. esimesel poolel ja 19 saj. teisel poolel)(i). 2. Kliimanäitajate ja hariliku männi radiaaljuurdekasvude vahelised seosed varieeruvad Eesti erinevate piirkondade vahel (II, III, IV). 3. Hariliku männi rinnaspindala kasvu saab modelleerida puude suuruse, kasvukoha viljakuse ja kliimanäitajate funktsioonina ning puudevaheliste konkurentsitingimuste muutusi saab mudelitesse lisada sarnaselt kui puude reageeringut harvendusraiele (III). 4. Kliimanäitajate ja radiaaljuurdekasvu vahelised seosed ei ole ajas stabiilsed keskkonnatingimuste pideva muutumise tõttu (IV). Materjal ja metoodika Katsealad Kagu-Eestis viidi läbi uurimus (I) männikute erinevate rajamisaegadega puistute kõrguskasvu erinevuste väljaselgitamiseks. Järvselja Õppe- ja Katsemetskonnas (58 25ʹN, 27 46ʹE) valiti tüveanalüüside jaoks puud 89

91 (n=28) neljast hariliku männi enamusega (Pinus sylvestris L.) üksteisele lähedal asuvast puistust. Hariliku männi radiaaljuurdekasvude andmed (puursüdamikud) koguti kokku 135 erinevalt proovitükilt üle Eesti (joonis 4.1, tabel 4.1), mis paiknesid Eesti metsa kasvukäigu püsiproovitükkide võrgustiku proovialadel (Kiviste jt, 2015). Proovide (puursüdamikud ja tüveanalüüsid) võtmiseks valiti hariliku männi enamusega proovitükid nii mineraalmuldadel kasvavatesse puistutes (n=119) (I, II, IV) kui ka Kirde-Eestis põlevkivi kaevandamise järgselt taastatud aladele rajatud metsakultuurides (n=12) (III). Hariliku männi kasvutingimuste varieeruvuse uurimiseks jaotati proovialad nende geograafilise asukoha järgi nelja piirkonna vahel: Kirde-Eesti (NE), Kagu-Eesti (SE), Edela-Eesti (SW) ja saared (ISL). Muutused kõrguskasvu pikaajalises dünaamikas Uurimaks, kas puistute kõrguskasvus esineb kolme kasvu alguse poolest ajaliselt oluliselt erineva hariliku männi põlvkonna (keskmiste vanustega 115, 56 ja 40 aastat) vahel erinevusi, võeti 28 männipuude tüveanalüüsid, mille tulemusena rekonstrueeriti puude tegelikud kõrguse ja läbimõõdu juurdekasvud. Iga puu kohta arvutati kumulatiivsed kõrguskõverad ja võrreldi kõrgusindekseid kontrollvanusel 40 aastat. Võrreldi kolmeparameetrilise Richardi kasvufunktsiooni (1959) ja mitteparameetrilise üldistatud aditiivse mudeli (Wood, 2006) arvutatud lähendeid. Kõrguskasvu erinevusi analüüsiti ühefaktorilise dispersioonanalüüsiga. Kolme erineva puistupõlvkonna kõrguskasvude dünaamikat võrreldi ka hariliku männi üldise keskmise kõrguskasvuga pohla kasvukohatüübis. Täiendavalt analüüsiti kasvuandmeid (kolme 40-aastase perioodi kohta , vana; , keskmise vanusega; , noor) vastavate perioodide Tõravere ilmajaama meteoroloogilistele andmetega. Erinevate perioodide kliimatunnuseid analüüsiti kasutades ühefaktorilist dispersioonanalüüsi ja Tukey post-hoc testi. Dendrokronoloogilised meetodid Kasvuaastate vahelist radiaalkasvu varieerumise uurimiseks kasutati rohkem kui 900 hariliku männi analüüsipuu mõõteandmeid ning saadud aastarõngaste kronoloogiad omakorda rühmitati vastavalt 90

92 metsa kasvukoha tüübirühmade (II) või siis metsakasvukohatüüpide (IV) (Lõhmus, 2004) lõikes. Põlevkivi kaevandamise järgselt taasmetsastatud aladel analüüsiti hariliku männi kasvu oluliselt detailsemal tasemel, seda eelkõige tasandatud puiste suure varieeruvuse ja sellest tuleneva puistu takseertunnuste suure varieeruvuse tõttu. Kokku koostati kolm kasvukohatüübipõhist kronoloogiat ja kaksteist proovitükipõhist kronoloogiat kliimamõjude hindamiseks kultuurmännikutes. Doktoritöös koostatud kronoloogiaid võrreldi alale lähimast meteoroloogiajaamast pärinevate igakuiste kliimaandmetega (eelkõige temperatuur ja sademed). Seoste uurimiseks aastase juurdekasvu ja kliimanäitajate vahel kasutati korrelatsioon- (II, III, IV) ja vastavusfunktsiooni analüüsi, mis teostati puude kasvuandmete ja meteoroloogiliste andmeseeriate kattuva perioodi kohta. Kliimanäitajate ja puude kasvu vaheliste seoste ajalist stabiilsust uuriti libiseva korrelatsioonanalüüsiga. Varasemate ekstreemsete kliimanähtuste mõju radiaalkasvule uurimiseks kasutati näitaasta analüüsi (Schweingruber jt, 1990; Neuwirth jt, 2007) võttes aluseks Cropperi (1979) analüüsimetoodika. Vastavalt Neuwirth jt (2007) kriteeriumitele jaotati näitaastad signaali intensiivsuse alusel kolme erinevasse rühma. Hariliku männi populatsioonide sarnaselt reageerivate gruppide tuvastamiseks kasutati peakomponentanalüüsi, kus korrelatsioonikoefitsiendid arvutati radiaalkasvu indeksite ja igakuiste kliimanäitajate vahel bootstrap-meetodiga. Täiendavalt hinnati puude radiaalkasvu dünaamika sarnasusi kasutades hierarhilist klasteranalüüsi. Rinnaspindala juurdekasvu mudel Kliimamuutuste ja majandamistegevuste pikaajaliste mõjude prognoosimiseks koostati üksikpuude kohta rinnaspindala juurdekasvu mudel, kasutades selleks 128 puu puursüdamike andmeid endistesse põlevkivikarjääridesse rajatud hariliku männi kultuuridest. Saadud multiplikatiivne mudel koosneb kolmest komponendist, mis (1) kirjeldab puude radiaalset juurdekasvu piiramata tingimuste korral, (2) esitab juurdekasvu muutumist vastavalt harvendusraie intensiivsusele ja (3) kirjeldab aastasisest ilmastikutingimuste varieeruvust. Puude rinnaspindala juurdekasvu assümptootilise dünaamika kirjeldamiseks kasutati Weberi kasvufunktsiooni ja selle seost puude suuruse ja kasvu- 91

93 koha headusega. Aastase juurdekasvu järjepidevat häiringuteta kasvu, mis mändidel kestis 13 aastat Hynyneni (1995) järgi, modelleeriti kirjeldamaks puudevahelise konkurentsi mõju. Iga-aastaseid juurdekasvu muutumisi, mis tulenesid ilmastikutingimuste varieerumisest, modelleeriti kas kliimaindeksite või kliimanäitajate lisamisega mudelisse (nt. sademete hulk juunist juulini ja keskmine kevadine temperatuur). Mudeli hindamine põhines mudeli headust kirjeldavate statistikute ja mudeli jääkide jaotuste analüüsil. Tulemused ja arutelu Muutused metsade pikaajalises produktsioonis Täheldatud erinevused hariliku männi kolme põlvkonna kõrguskasvu dünaamikas toetasid pikaajalist kasvu muutumise hüpoteesi (hüpotees 1), mille kohaselt kasvavad nooremad puistud (rajatud pärast 1950 a.) kiiremini võrreldes nendega, mis on rajatud mitukümmend aastat varem. Kuna uurimuses kasutatud mõlema noorema põlvkonna keskmine aastane radiaal- ja kõrgusjuurdekasv olid statistiliselt oluliselt (p<0.0001) suuremad, saadi tulemuseks erineva kujuga kumulatiivsed kasvukõverad (joonis 5.1a; joonis 3 artiklis I), kus noore ning keskmise vanusega puistus oli kõrgusindeks (SI 40 ) vastavalt 12% ja 23% võrra suurem võrreldes kõige vanema puistuga (joonis 4 artiklis I). Kuigi seda antud uurimuses ei käsitletud, võib saadud trendi võimalike põhjustena välja tuua ka selgelt jälgitavad pikaajalised muutused kliimatingimustes (aasta keskmise temperatuuri kasv ja kasvuaasta pikenemine). Meteoroloogiliste andmete täiendav analüüs näitas, et hilisema kasvualgusega puistute kasvuajal valitsesid oluliselt soojemad temperatuurid (joonis 5.2) võrreldes kahe varasema kasvuperioodiga. Kevaded, suved ja vegetatsiooniperioodid (maist augustini) olid keskmiselt samuti soojemad hilisema (pärast 1960 a.) kasvualgusega puistus võrreldes varem (enne 1950 a.) kasvama hakanud puistute vanuseliselt võrreldava kasvuperioodiga. Hariliku männi ja hariliku kuuse aastase kõrguskasvu suurenemist on täheldatud samuti Skandinaaviamaades (Elfving ja Tegnhammar, 1996; Elfving ja Nyström, 1996; Subedi ja Sharma, 2010; Sharma jt, 2012) ja Kesk-Euroopas (Lebourgeois jt, 2000; Pretzsch jt, 2014). Seega ei ole 92

94 käesolevas doktoritöös leitud trend pikaajalises mändide kõrguskasvus iseenesest midagi erakordset. Eestis varasemalt saadud tulemused (Kiviste, 1999b; Nilson jt, 1999) koos käesoleva uurimuse tulemustega osutavad aga sellele, et konstantsete kasvutingimuste eeldusel põhinevad baaskõrguse mudelid ei anna täna enam usaldusväärseid kõrguskasvu prognoose kiirenenud kasvuga nooremate männipõlvkondade puhul ning nende kasutamisel hindame kasvu olulisel määral alla. Aastaste kliimamuutuste mõju hariliku männi aastarõngaste laiusele Aastarõngaste laiuse dendroklimaatiline analüüs näitas, et hariliku männi kasv Eestis on positiivses seoses hilistalviste/varakevadiste temperatuuridega. Negatiivne seos eelmise aasta augusti keskmise temperatuuriga ja positiivne seos sama kuu sademete hulgaga näitas, et niiskusdefitsiit vegetatsiooniperioodi lõpus omab järgmise aasta aastarõnga laiuse kujunemisele negatiivset mõju. See seos oli tugevam kuivades ja toitainevähestes kasvukohtades (sambliku ja kanarbiku kasvukohatüübid), eriti Kirde-Eestis ja saartel kasvavate harilike mändide puhul. Viljakamatel ja parasniisketel muldadel (mustika ja pohla kasvukohatüüp) kasvavatel mändidel olid tugevamad seosed talvise puhkeperioodi lõpu temperatuuridega. Kuivade ja toitainevaeste kasvukohtade mändidel oli nõrgem seos külma perioodi temperatuuridega. Hariliku männi kasv Edela-Eesti parasniisketes kasvukohtades ja saartel näitas tugevat korrelatsiooni külma perioodi temperatuuridega. Endistesse põlevkivikarjääridesse rajatud hariliku männi kultuuride kasvu ja kliimanäitajate vahelisi seoseid uuriti kasvuperioodil Puistut ja kasvukohta iseloomustavad näitajad mängisid olulist rolli hariliku männi reageerimisel ilmastikunäitajate muutumisele. Aastarõngaste laius oli positiivses seoses suurenenud sademete hulgaga sama aasta juulis ja kõrgemate temperatuuridega kevadel. Siiski oli täheldav negatiivne seos suurema kivisusega substraadil ja tihedamates puistutes kasvavate männipuude radiaaljuurdekasvu ning suve, iseäranis just kasvuaasta augusti, keskmiste temperatuuride vahel. Hariliku männi metsakasvukohatüüpide indeksite kronoloogiad jagunesid hierarhilise klasteranalüüsi põhjal kolmeks klastriks. Kagu-Eesti ja Edela-Eesti kasvukohatüüpide kronoloogiad koondusid ühte klastrisse ning kuigi endiste põlevkivikarjäärialade kronoloogia oli sarnasem 93

95 Kirde-Eesti puistutega, saab nende alusel moodustada eraldi klastri. Enim sarnasusi täheldati sama piirkonna kronoloogiate vahel, sama metsakasvukohatüübi kronoloogiad erinevates piirkondades olid üksteisest suhteliselt erinevad. Hariliku männi kasvu Kirde-Eestis ja saartel piiras enim sademete vähesus kasvuperioodile eelnenud vegetatsiooniperioodi lõpus. Edela- ja Kagu-Eestis kasvavatele männipuistutele avaldasid tugevalt positiivset mõju talve ja v arakevade madalad temperatuurid. Näitaastate analüüs tõi välja enamiku harilike mändide vähenenud radiaaljuurdekasvu aastatel 1940 ja 1985 kõigis uuritud piirkondades. Erakordselt väikest radiaaljuurdekasvu oli põhjustatud äärmuslikult külmadest talvedest ja neile järgnenud põuast kasvuperioodi ajal. Järeldused Kliimatingimusi on alati iseloomustanud suur ajalis-ruumiline varieeruvus. Ilmastik koos lokaalsete kasvukohta iseloomustavate tingimustega on peamised tegurid, mis määravad puude võimaliku kasvu kindlas geograafilises piirkonnas. Vaatamata kliimanäitajate olulisusele ei ole neid Eestis metsade empiirilistes kasvumudelites seni sisendina kasutatud. Doktoritöös saadud tulemused (I, II, III, IV) näitavad, et ilmastikutingimustel on otsene mõju puude kasvule konkreetsel ajal ja konkreetses kohas. Erinevused kõrguskasvu dünaamikas (I) näitavad, et puude aastast kõrguskasvu soodustavate kliimatingimuste mõju võib pikema aja jooksul kumuleeruda, andes tulemuseks puudele varasemaga võrreldes suuremad mõõtmed ja puistutele kõrgema tootlikkuse ning seda isegi suhteliselt lühikese aja jooksul. Käesolevas doktoritöös seostati hariliku männi radiaaljuurdekasv kliimanäitajate varieeruvusega aastaste perioodide ulatuses ning leiti, et kliimanäitajate mõju harilikule männile sõltub kohalikest ilmastikutingimustest ja kasvukoha omadustest (II, III, IV), kuid puude kasvu võivad mõjutada ka konkreetse puistu tunnused (puistu tihedus, vanus jt.; IV). Uudseid teadmisi hariliku männi kasvu reageerimisest kliimanäitajate muutumisele saab kasutada hariliku männi kasvu edasisel modelleerimisel. Ilmastikunäitajate lisamine doktoritöös kasutatud kasvumudelisse parandab mudeli prognoose vaid vähesel määral, mistõttu tuleb metsa kasvumudelit edasi arendada ning ka täiendavalt valideerida. Koostada oleks vaja ka uus mudel metsamaal kasvava hari- 94

96 liku männi radiaalkasvu prognoosimiseks. Doktoritöö tulemused näitavad, et kliimamuutuste ajalist ja ruumilist mõju puude kasvule ei tohiks alahinnata, seda eriti erinevaid Eesti geograafilisi piirkondi võrreldes või analüüsides puude kasvu varasematel perioodidel. 95

97 ACKNOWLEDGEMENTS Firstly, I would like to thank my supervisors Professor Andres Kiviste and Associate Professor Ahto Kangur for guidance through the research studies, continuous support and help in my PhD studies. I appreciate your patience, encouragement and valuable comments received during the time of research and while preparing this thesis. I would also like to thank Prof. Henn Korjus for the opportunity to be part of Forest Management Department team and for his advice. I thank all the co-authors of the research articles for cooperation and contributions while preparing the articles for publication. Special thanks to Dr. John A. Stanturf and Dr. Lee E. Frelich for their insightful comments and language improvements during research article preparation. My sincere thanks go to Associate Professor Maris Hordo, who introduced me to tree rings and dendrochronology, and with whom I explored the forests of Estonia while conducting fieldwork and collecting samples. I also thank Dr. Regino Kask and Alar Padari and the students who assisted me during fieldwork, Dr. Allan Sims for helping with data analysis. I am grateful to Dr. Mait Lang for valuable conversations and assistance, and the rest of my colleagues from the Forest Management Department of Forest Management who supported me during this thesis preparation. I especially thank my family-husband Marek, who encouraged me during the hard times and helped me with practical issues, and my sons Lukas and Jonas who has been an inspiration in my life. I am grateful to my parents, relatives and friends for their continuous care, concern and encouragement during the years of this thesis preparation. The studies of this thesis were financially supported by the Estonian Science Foundation (Grant No. 8890), by the Estonian Ministry of Education and Research (SF s08), and by the ESTEA (MUU 8-2/T8047MIM). The ESTEA financed increment core collection (MUU 8-2/T7103MIMI) and provided meteorological data. The European Regional Development Fund represented by the Archimedes Foundation covered the costs related to participation in courses taken outside Estonia. 96

98 I ORIGINAL PUBLICATIONS 97

99 Metslaid, S., Sims A., Kangur, A., Hordo, M., Jõgiste, K., Kiviste, A., Hari, P Growth patterns from different forest generations of Scots pine in Estonia. Journal of Forest Research, 16 (3):

100 J For Res (2011) 16: DOI /s SPECIAL FEATURE: ORIGINAL ARTICLE Approaches for forest disturbances studies: natural variability and tree regeneration Growth patterns from different forest generations of Scots pine in Estonia Sandra Metslaid Allan Sims Ahto Kangur Maris Hordo Kalev Jõgiste Andres Kiviste Pertti Hari Received: 21 July 2010 / Accepted: 19 March 2011 / Published online: 15 May 2011 Ó The Japanese Forest Society and Springer 2011 Abstract There is strong evidence that climate change alters tree growth in boreal forests. In Estonia, the analysis of ring measurements has been a common method to study growth changes. In this study, annual height growth data from dominant or co-dominant Scots pine (Pinus sylvestris L.) trees were used to develop a growth model for three forest generations. Stem analysis was applied and annual height growth was measured as the distance between whorls, as detected by branch knots of whorls on the split stem surface. Retrospective analysis of height growth produced comparative trends for three different age groups. Statistical analyses were used to estimate the impact of different factors on growth. Annual height growth was considered the best indicator for detecting possible trends in the growth potential of trees. Study results indicate that three generations (separated by time periods of years) showed significant differences in growth patterns caused by shifts in climatic factors and management regimes (anthropogenic and natural disturbances). Keywords Annual growth Forest growth change Growth modelling Stem analysis S. Metslaid (&) A. Sims A. Kangur M. Hordo K. Jõgiste A. Kiviste Institute of Forestry and Rural Engineering, Estonian University of Life Sciences, Kreutzwaldi 5, Tartu, Estonia sandra.metslaid@emu.ee P. Hari Department of Forest Sciences, University of Helsinki, P.O. Box 27, Helsinki, Finland Introduction Forest stands are open systems (Pretzsch 2009, pp. 6 7) and they tend to respond to changes in environmental conditions, which may arise as a result of natural events or can be induced by human activities. Global climate change, characterized by rising temperatures, altered precipitation patterns and chemical composition of air, affects tree phenology, evapotranspiration demands, and at the same time influences soil moisture and nutrient supply. These changes may cause temporary or long-term alterations in forest growth (e.g., Kahle et al. 2008). The number of forest research studies concerning growth trends of forests has increased since the beginning of the 1980s, but the results have been controversial with high spatial variation (e.g., Spiecker et al. 1996). Growth increases have been found in boreal and hemi-boreal forests (Hari et al. 1984; Nilson and Kiviste 1986; Mielikäinen and Timonen 1996; Lopatin 2007) and in temperate forests of Europe (Eriksson and Johansson 1993; Wenk and Vogel 1996; Lebourgeois et al. 2000; Mellert et al. 2004; Bontemps et al. 2009), while no long-term growth changes have been detected in some studies (e.g., Tveite 1994, as cited in Elfving et al. 1996; Zetterberg et al. 1996; Mielikäinen and Sennov 1996; Sinkevich and Lindholm 1996). Analysis of tree-ring chronologies from a circumpolar network showed tree growth decline since the middle 1900s (Lloyd and Bunn 2007). Spatial and temporal growth variability was observed for trees growing at the cold margins of the boreal forest in three regions in Alaska (Lloyd and Fastie 2002). The signs of changing climate have been observed in Estonia. A warming trend in spring and delayed beginning of winter resulting in a longer growing season have been reported by Ahas and Aasa (2006) for the period

101 238 J For Res (2011) 16: Additionally, more intense solar radiation has been recorded since the 1990s (Wild et al. 2005). The forests of Estonia belong to hemi-boreal ecosystems (Kangur 2009), but changes in forest growth have not been studied extensively, nor covered by larger scale projects over the last century in the Baltic countries. Several attempts to detect changes in forest growth in Estonia have been done during recent decades. Kiviste (1999) and Nilson et al. (1999) used forest inventory data and observed changes in site index. Pärn (2008) investigated the changes in radial growth in two consecutive generations of Scots pine stands, using the constant cambial age method. All these studies reported positive trends. However, a better understanding of how single-tree growth on different forest site-types is affected by changing climate conditions is needed in Estonia for the development of growth models of trees and stands on different sites for efficient (sustainable) forest management planning. The aim of this work was to study the effects of climate change on height and radial growth in Estonian Scots pine (Pinus sylvestris L.). It is a dominant and one of the most economically important tree species in Estonia, growing on 35% of forest land, with a growing stock of million cu m (MMK 2008). Significant changes in growth of Scots pine stands may have a substantial impact on timber markets due to altered availability and supply of industrial round wood (Solberg et al. 2003). Dominant height is assumed to be a more advantageous growth variable than basal area growth, because it is relatively unaffected by stand density and management (Mäkinen and Isomäki 2004) and at the same time is a good proxy for site productivity (Clutter et al. 1983). Achievement of the main goal of this study requires development of a novel method to detect growth changes in annual height increments of trees. To evaluate long-term changes in forest growth at the site level, we compared the growth of tree groups with different germination years. We hypothesize that stands with three different germination years (separated by a time period of years) will show significant differences in growth patterns, caused by shifts in climate regime and due to evolution of site fertility during the last 100 years. Materials and methods Study area Four sampling sites were located in even-aged Scots pinedominated forest stands, at the Järvselja Training and Experimental Forest Centre (JTEFC) ( N, E), situated in south-east Estonia (Fig. 1). The forests managed by JTEFC belong to the hemi-boreal zone with a moderately cool and moist climate. Average annual temperature for the reference period of was 4.1 C and average annual precipitation amounted to 600 mm (EMHI 2011). The vegetation growth period, described by growing degree days (calculated as average of minimum and maximum temperatures and compared to a base temperature of 5 C) on average lasts for 170 days, starting in the beginning of April and ending in mid-october (Rautiainen et al. 2009). Stand selection and field sampling for sample trees selection Stands from a contiguous area, with the same site conditions and similar stand characteristics, but of different germination year, were selected for the study. Altogether four Scots pine-dominated stands, growing on the fertile Oxalis-Rhodococcum site type, were chosen. As a reference for past growth, two pine stands with average tree age of 115 years were selected. For growth comparison, one middle-age stand (average tree age 56 years) and one young stand (average tree age 40 years) were sampled. The oldest stands include 80 years of measurements on permanent sample plots, while the two younger stands were sampled here for the first time. All selected stands grow on previously cultivated agricultural land (oldest stands are first generation and younger stands second generation), where the regeneration was accomplished by seeding. The oldest stands were established with direct seeding on a ploughed field, the middleage stand was sowed on scarified patches and the young stand was sowed on a deeply ploughed clear-cut area. Stand development history is known for the two oldest stands (Sims et al. 2009), as they belong to one of the oldest longterm forest growth and yield experimental plot series still in existence, and their location together with measurement information is published in the Northern European Database of Long-Term Forest Experiments (NOLTFOX) where they can be found under experiment number M046. A circular sample plot was established within each selected stand. The radii of sample plots varied from 15 m in young stand to 20 m in middle-age and old stands. The radius of plots varied because of the fixed minimum size of sampled subpopulations: each plot was required to encompass at least 100 main canopy trees. All trees on the plots with DBH [4 cm were recorded and DBH and tree location measured. In addition, tree height was measured for every fifth tree. To describe the social rank of trees within stands, the sampling methodology described by Havimo et al. (2008) was used. The trees were assigned to one of five dominance classes by DBH: class 1 (top 5% of the trees by DBH), class 2 (top 6 20%), class 3 (top 21 50%), class

102 J For Res (2011) 16: Fig. 1 Study sites in Järvselja TEFC (black circle) and the location of Tõravere meteorological station (black triangle) Table 1 Main characteristics of the sampled trees Age group Number of trees Age range (years) Growth period DBH (cm) Height (m) Mean SD Mean SD Old Middle-age Young (51 80%), and class 5 (the remaining 20%). Depending on the stand average age and tree dominance class, we selected 6 11 trees for sectioning per sample plot. This resulted in 34 sample trees: 10 trees in class 1, 10 trees in class 2, 8 trees in class 3, 3 trees in class 4, and 3 trees in class 5. Five trees were felled in 2006 and 29 trees in For the analyses of current study, only dominant or codominant trees were taken into account (e.g., trees from the first three dominancy classes, 28 of the 34 trees felled). Detailed information about the selected trees is presented in Table 1. Stem analysis method for radial and apical increment The stem analysis technique was applied for retrospective assessment of annual growth, to obtain long-term radial and apical increment series. Selected trees were cut into 2.50-m-long sections. Disks were cut from the bottom of every trunk section; number and widths of annual rings were later recorded in the laboratory. Two additional disks were obtained from every tree: at breast height and at the live crown base. The stem sections were split in half, north south direction, and a board was cut from the central

103 240 J For Res (2011) 16: part of the tree along each section of trunk. The disks and panels were labeled and delivered to the laboratory for measurements. The surfaces of the disks were sanded and the number of rings as well as the annual ring-widths were measured to the nearest 0.01 mm in two perpendicular directions. A LINTAB tree-ring measuring table and the computer program TSAPWin Scientific Version 0.59 (Rinn 2003) were used. Measured series were cross-dated (Pilcher 1990) visually by comparing ring-width graphs. The quality of visual cross-dating was also checked statistically using COFECHA (Holmes 1983; Grissino-Mayer 2001). Total tree age was estimated as the number of tree-rings from the basal disk. The annual tree-ring measurements from breast height were used for analysis in this study. Annual height growth of each tree was determined by measuring distances between consecutive whorls with accuracy of 0.1 cm; whorls were detected as bud scars on the split stem surface. The number of height increments was cross-checked with the number of tree rings, counted on every cross-section (disc). Assessment of differences in forest growth A number of growth functions can be used to model tree height development (e.g., Zeide 1993; Kiviste et al. 2002). In this study, each tree height age series was approximated with the three-parameter Richards (1959) growth function h ¼ a ð1 expð b tþþ c ð1þ where h is tree height, t is tree age, and a, b, and c are parameters. Nonlinear regression analysis was used for parameter estimation. To characterize height growth rate of each individual tree, height predictions with Eq. 1 at age of 40 years were calculated. We tested different reference ages for height predictions (data not presented); however, the most similar results appeared at the age of 40 years, which was the oldest measured age for the young age group. Considering that the three-parameter Richards function (Eq. 1) is not flexible enough to describe single tree height development trends, another smoothing method the non-parametric generalized additive model (GAM model) (Wood 2006) was applied. Age group effect on individual tree growth prediction at the age of 40 years was analyzed with one-way ANOVA. Statistical analysis was performed and graphs were drawn with R (R Development Core Team 2009). For a comparison of long-term growth trends, Kiviste s mean stand height development model (Kiviste 1997; Kangur et al. 2007; Kiviste and Kiviste 2009) was used as a reference to describe typical Estonian forest growth on particular site. The model was created based on Estonian forest inventory data from and represents a mean growth rate for that time period. For long-term height development prediction with Kiviste s model, the GAM model estimates of individual tree growth prediction were used. Results Analysis of height measurement data from 28 trees showed different height development growth rates (estimated with the GAM model) for the three age groups (Fig. 2a). Higher height growth rates were observed in both younger age groups. Deviation trends of differences between long-term height development prediction (with Kiviste s model) and age group mean height development predictions (with the GAM model) for each age group are presented in Fig. 2b. The comparison of deviation trends showed growth rate differences over the observation period. The almost horizontal solid curve in Fig. 2b shows that height growth of the middle-age group behaves similarly to Kiviste s model. This is expected, because the data collected are from same time period as the data used in calibrating Kiviste s model. The old age group deviation trend shows that over a long period height growth rate has increased. Comparisons of age group mean annual height increment predictions with cumulative height (Fig. 3a) and age group mean annual radial increment with cumulative diameters (estimated with the GAM model) (Fig. 3a, b) show that younger age groups have higher height and radial increment rates than the old group for the same cumulative height and diameter. The effect of the age group was highly significant (p \ ) in case of both height and diameter. The adjusted R 2 of the GAM model for height was R 2 = while adjusted R 2 of the GAM model for diameter was R 2 = Therefore, the annual height increment variation is better described than the annual diameter increment variation for the whole observation period with the GAM model. Figure 4 presents tree height predictions at age 40 years, both with the Richards function (Rch) and with the nonparametric GAM model (GAM). One-way ANOVA revealed significant (p \ ) difference of individual tree predictions at the age of 40 years among age groups (see mean and 95% confidence intervals at the bottom of Fig. 4). The average predictions were 24.0, 21.8 and 19.5 m for the young, middle and old age groups, respectively. Discussion To detect changes in forest growth, data from long-term observations during stand development are needed, but

104 J For Res (2011) 16: Fig. 2 Height development (estimated with GAM model) of different age groups, in bold in (a); sampled trees data series are presented in the background (in gray) and deviation trends of difference between age group mean height and the long-term prediction in (b) Fig. 3 The relationships between age group mean annual height increment and cumulative height (a), and between age group mean annual radial increment and cumulative diameter (b). The relationships are presented in bold together with the 95% confidence intervals, sampled trees data series are presented in the background (in gray) repeated measurements of tree diameter and height usually cover only several decades. To overcome these problems, retrospective techniques, such as stem analysis and treering analysis, can be successfully applied (Spiecker 1999). This method provides annual data on height, diameter and volume growth for each individual tree. In previous studies where the cohort comparison approach (e.g., Untheim 1996; Salminen et al. 2009; Bontemps et al. 2010) was applied, substantial differences between generations were detected, showing that growth of young trees is accelerating. The results of our study support the growth acceleration findings in earlier studies by showing increased growth with consecutively younger age groups. These changes are observable for both height and diameter increment and for cumulative height and diameter. Studies by Spiecker et al. (1996); Karjalainen et al. (1999) and Kahle et al. (2008) suggested that increased growth of Scots pine (Pinus sylvestris L.) and Norway spruce (Picea abies L. Karst) stands in Central Europe is mainly due to higher supply of nitrogen, whereas temperature and elevated concentrations of CO 2 may have significant contributions in the future (Kahle et al. 2008). Curtis and Wang (1998), Ainsworth and Long (2005) and Hari and Nikinmaa (2008) state that the positive growth response of trees in recent years is due to elevated CO 2 concentration. In addition to global climate variables (elevated CO 2 concentration and nitrogen deposition) effects on tree growth, we consider that silvicultural methods (site preparation for regeneration, selective cuttings not recorded), site conversion from agricultural to forest land, and the unknown origin of second generation trees must equally be considered as causes of increases in site index. The single tree height prediction at the age of 40 years in our study is a good indicator of site-specific growth potential, and the comparison between the different age groups indicates a difference of approximately 2 m in height at the age of 40 years. Our results coincide with the results observed by Nilson et al. (1999), where the average increase in the site index (H50) of Estonian Scots pine stands for the period was estimated at 2 m. Traditionally, site indexes were used to describe the growth potential of the forest site. The changes in site index imply improvement of site productivity for woody vegetation

105 242 J For Res (2011) 16: local climate change better than older stands that were already growing in the previous climate conditions. Acknowledgments This study was supported by Järvselja Training and Experimental Forest Centre, Estonian Environmental Investment Centre, State Forest Management Centre, Metsämiesten Säätiö and The Ministry of Education and Research (project SF s08 and grant no ETF8890). We are very grateful to Lee E. Frelich, University of Minnesota, for discussion and language correction. References Fig. 4 Sample tree height prediction at age 40 years with Richards (1959) function (Rch) versus the prediction with GAM model (GAM). In the lower part, mean prediction values (Rch) by age groups with 95% confidence intervals are presented For vegetation growth, the most relevant limiting factors are availability of light, nutrients, water, and temperature (e.g., Hari et al. 2008). Ecological boundaries are affected by complex environmental factors and their interactions (Danz et al. 2010; Frelich and Reich 2010). The response of growth to increasing temperature can be divergent within one generation of trees, especially for species growing near the edge of the range. Temperature rise can create direct positive or negative effects, or negative effects mediated by water deficit (Lloyd and Bunn 2007). Tree ring growth happens throughout the whole growing season. Tree carbohydrate allocation to height growth occurs during a shorter period. The allocation of growth appears to be age dependent (Konôpka et al. 2010). Based on our results (Fig. 3), we suggest that tree height is a more appropriate proxy than diameter for studying long-term growth changes. This is because diameter growth is more influenced by stand density, whereas height growth is more affected by long-term changes in site conditions. All of the stands studied were established on previous agricultural land; thereafter, the old stands represent the first generation, but younger stands represent the second forest generation on agricultural land. Eriksson and Johansson (1993) found a 40% increase of stand volume in second generation spruce stands in south-western Sweden. One explanation for the higher growth on younger stands can also be the fact that the stands established more recently are able to adapt to site improvements triggered by Ahas R, Aasa A (2006) The effects of climate change on the phenology of selected Estonian plant, bird and fish populations. Int J Biometeorol 51:17 26 Ainsworth EA, Long SP (2005) What have we learned from 15 years of free-air CO 2 enrichment (FACE)? A meta-analytic review of responses of photosynthesis, canopy properties and plant production to rising CO2. New Phytol 165: Bontemps JD, Herve JC, Dhote JF (2009) Long-term changes in forest productivity: a consistent assessment in even-aged stands. For Sci 55: Bontemps JD, Herve JC, Dhote JF (2010) Dominant radial and height growth reveal comparable historical variations for common beech in north-eastern France. For Ecol Manag 259: Clutter JL, Fortson JC, Pienaar LV, Brister GH, Bailey RL (1983) Timber management: a quantitative approach. Wiley, New York Curtis PS, Wang XZ (1998) A meta-analysis of elevated CO2 effects on woody plant mass, form and physiology. Oecologia 113: Danz NP, Reich PB, Frelich LE, Niemi GJ (2010) Vegetation controls vary across space and spatial scale in a historic grassland-forest biome boundary. Ecography. doi: /j x Elfving B, Tegnhammar L, Tveite B (1996) Studies on growth trends of forests in Sweden and Norway. In: Spiecker H, Mielikäinen K, Kohl M, Skovsgaard J (eds) Growth trends in European forests: studies from 12 countries. Springer, Berlin, pp EMHI (2011) Estonian climate. Estonian Meteorological and Hydrological Institute. Accessed 2 Feb 2010 Eriksson H, Johansson U (1993) Yields of Norway spruce (Picea abies (L.) Karst.) in two consecutive rotations in southwestern Sweden. Plant Soil 154: Frelich LE, Reich PB (2010) Will environmental changes reinforce the impact of global warming on the prairie-forest border of central North America? Front Ecol Environ 8: doi: / Grissino-Mayer HD (2001) Evaluating crossdating accuracy: manual and tutorial for the computer program COFECHA. Tree-Ring Res 57: Hari P, Nikinmaa E (2008) Response of boreal forest to climate change. In: Hari P, Kulmala L (eds) Boreal forest and climate change. Springer, Berlin, pp Hari P, Arovaara H, Raunemaa T, Hautojärvi A (1984) Forest growth and the effect of energy production: a method for detecting trends in the growth potential of trees. Can J For Res 14: Hari P, Räisänen J, Nikinmaa E, Vesala T, Kulmala M (2008) Evaluation of the connections between boreal forest and climate change. In: Hari P, Kulmala L (eds) Boreal forest and climate change. Springer, Berlin, pp Havimo M, Rikala J, Sirviö J, Sipi M (2008) Distributions of tracheid cross-sectional dimensions in different parts of Norway spruce stems. Silva Fenn 42:

106 105 II

107 Hordo, M., Metslaid, S., Kiviste, A Response of Scots pine (Pinus sylvestris L.) radial growth to climate factors in Estonia. Baltic Forestry, 15 (2):

108 RESPONSE OF SCOTS PINE /.../ RADIAL GROWTH TO CLIMATE FACTORS IN ESTONIA BALTIC FORESTRY M. HORDO ET AL. Response of Scots Pine (Pinus sylvestris L.) Radial Growth to Climate Factors in Estonia MARIS HORDO, SANDRA METSLAID AND ANDRES KIVISTE Department of Forest Management, Institute of Forestry and Rural Engineering, Estonian University of Life Sciences, Kreutzwaldi 5, 51014, Tartu, Estonia; * maris.hordo@emu.ee Hordo, M., Metslaid, S. and Kiviste, A Response of Scots Pine (Pinus sylvestris L.) Radial Growth to Climate Factors in Estonia. Baltic Forestry, 15 (2): Abstract The following research paper analyzes Scots pine (Pinus sylvestris L.) radial growth responses to climatic factors in mesotrophic and heath forest site types in Estonia. Increment cores from 889 trees from 119 plots of the network of research plots were used and chronologies for mesotrophic and heath forest site types of Scots pine were constructed. The relationship between climatic factors and the radial growth of Scots pine was characterized by correlation coefficient; also pointer year analysis, Cropper method was applied to single tree series. Cropper values were calculated; extreme negative and positive pointer years were identified. According to analyses, 1940 and 1985 were the most significant negative pointer years among different sites; and significant positive years were 1945, 1946, 1989, and Extreme Cropper values indicated significant positive correlation with the monthly mean temperature in winter (January, February) and early spring (March, April) before a growing season; also with the mean annual temperature and the mean temperature of the vegetation period (from April to September). Significant negative correlation was found between the extreme Cropper value and the precipitation of the previous year August. Therefore, temperature can be considered as the most important single factor of growth activity. Pointer year analyses confirmed that severe winters, cool springs and dry summer conditions are the main causes for the sharp decrease in the radial growth. Key words: climate variables, pointer years, radial growth, Scots pine, tree-ring chronology The dynamics of the annual radial growth of a tree is closely related to fluctuations of climate and other ecological factors which encompass dendroclimatology (Fritts 1976, Cook and Kairiukstis 1990, Schweingruber 1996). Dendroclimatological methods are widely applied in modern studies on forest dynamics, assessing damages, forecasting climate changes, and productivity (Spiecker et al. 1996, Worbes 2004). Changes in average climatic conditions, such as air temperature, affect the length of the growing season and influence site productivity (Fabian and Menzel 1999). Tree rings are an excellent material for studying climate-growth relationships, as they reflect environmental conditions and changes, and store the reaction pattern over time, which can later serve as an archive (Spiecker 2002). Growth response to climatic influences varies with species, provenance, competitive status, and site conditions (Fitts 1976, Sweingruber 1996, Spiecker 2002). Estonia became engaged in dendrochronological research in the early 1970s, with Kalvi Aluve (Läänelaid 1997, 2002) measuring and dating tree-ring widths in historical buildings, which was continued by Alar Läänelaid and Dieter Eckstein (2003), who constructed the long-term chronology for Scots pine (Pinus sylvestris L.), covering the period of (482 years). The use of radial tree-ring data in forest research was introduced in the 1980s, when Erich Lõhmus (1992a), a researcher at the Estonian Forest Institute, developed a generalized chronology for Scots pine in Estonia for the period of (203 years). Lõhmus also compiled separate chronologies for three soil moisture classes (for arid, moderate, and humid sites), to enhance the sensitivity and informativeness of general chronology by minimizing typological diversity. Considerable input on studying the radial growth of Scots pine in park forests and regions of different cement dust loads has been provided by Henn Pärn (2003, 2004, 2006). Scots pine is the species in which tree rings are one of the main sources of chronologies used in climate reconstruction (Lõhmus 1992a, Läänelaid 1997, Helama and Lindholm 2003, Vitas 2008) and have been used successfully in dendroclimatological research. It has the widest geographical distribution of all pines, and due to long rotation it is not problematic to find old trees that are considered suitable for dendrochronological studies. Different forces control its growth in 2009, Vol. 15, No. 2 (29) ISSN

109 BALTIC FORESTRY RESPONSE OF SCOTS PINE /.../ RADIAL GROWTH TO CLIMATE FACTORS IN ESTONIA M. HORDO ET AL. climatically different regions (Cedro 2001, Helama and Lindholm 2003), any one of which may limit or stimulate their growth (Cedro 2001). Scots pine is one of the most comprehensively investigated tree species in Estonia and Baltic countries by using dendroclimatological techniques (Vitas and Erlickytë 2007). In Estonia, pine is the most common and economically important tree species, with Scots pine dominating forests covering 29% (757,100 ha) of total forest area (MMK 2008). Therefore, due to the changing environmental conditions and evolving forest management objectives, updated information about trees and their growth is needed (Nilson 2002). For growth and yield studies in Estonia, the network of forest growth research plots (Kiviste and Hordo 2002) has been established since For modelling individual diameter growth, it would be necessary to consider the influence of climate factors on diameter increment during past decades by using chronological data (Mielikäinen 1985, Zahner 1988, Hynynen 1995, Gaucherela et al. 2008). Examining past relationships between tree growth and climate in Estonia will help us to understand how tree growth (productivity) in pine forests might be affected in the future. The aim of our study is to analyze which season s climatic variables affect the annual growth of Scots pine the most, with particular attention to extreme climatic factors, like severe winters, cold springs and summer droughts, which can be the main causes for an abrupt decrease in the tree s radial growth. Material collection Increment cores from living Scots pine trees were collected in Estonia during the summer/autumn of For sample tree selection, we used growth and yield research plots established in mesotrophic and heath forest site types. Mesotrophic forests (Rhodococcum and Myrtillus site types) are dry and well-lighted pine stands, growing on moderately humid and temporarily moist sandy soils (Etverk et al. 1995, Lõhmus 2004). This forest type is the most widespread in Estonia, comprising 40.8% (MMK 2008) of pine dominated forest land. Heath forests (Cladonia and Calluna site types) are sparsely stocked pine stands, with slow growth, located on poor dry sandy soils. Heath forests can be found primarily in northern Estonia and on the islands and to a lesser extent in northeast, southeast and southwest Estonia. Heath forests are of great importance to coastal dunes where they provide soil protection. Heath forests comprise 1.5% (MMK 2008) of all pine dominated forests. An Estonian network of forest growth research plots (Kiviste and Hordo 2002) similar to a Finnish INKA system (Gustavsen et al. 1988) has been designed to provide empirical data for forest growth and yield modelling (Kiviste et al. 2003, Hordo 2005, Kiviste et al. 2005). The network of 756 research plots with more than 100,000 mapped trees was established and re-measured in For this study, a subset of 119 pine dominated sample plots was selected, which comprised 81 plots from the mesotrophic forest site type and 38 plots from the heath forest site type. According to the geographical location, the sample plots were grouped by regions: islands, northeast, southeast and southwest (Figure 1). Ristna Islands Figure 1. Layout of permanent sample plots, used for core collection and location of EMHI stations that provided metereological data Increment cores were collected from trees outside the research plots from the North, South, East and West directions, which were determined from the center of a plot (Figure 2). For each research plot, up to 8 dominant trees without visible damages were sampled. Two increment cores in perpendicular radii were taken using an increment borer from each tree at 1.30 m above the ground. For this study, altogether 889 trees were cored from 119 sample plots, 602 trees from the mesotrophic and 287 trees from the heath forest site type. Figure 2. Location of sample trees (yellow spots) on the research plot. Sample trees of the research plot were taken from the basic compass points, outside the circular plot South-West W Kunda North-East Tõravere South-East N S E 2009, Vol. 15, No. 2 (29) 196 ISSN

110 RESPONSE OF SCOTS PINE /.../ RADIAL GROWTH TO CLIMATE FACTORS IN ESTONIA BALTIC FORESTRY M. HORDO ET AL. Tree-ring measurements and cross-dating Annual ring-widths were measured with an accuracy of 0.01 mm using LINTAB tree-ring measuring table with the computer program TSAP-Win Scientific Version 0.59 (Rinn 2003). The measured series were cross-dated (Pilcher 1990) visually by comparing the graphs of ring-widths. Cross-dating and data quality were assessed using the computer program COFECHA (Holmes 1983, Grissino-Mayer 2001). As a result of the cross-dating of tree ring time series, possible errors were eliminated and the series were verified among each other. Mean time series by trees and plots were built up with the software TSAP-Win. Descriptive statistics of a ring-width series for mesotrophic and heath forest site types were calculated using the TSAP-Win program. Statistical paramaters of the tree ring data like mean sensitivity (MS), standard deviation (Std), tendency changes (TC), auto correlation (AC), Gleichläufigkeit (Glk), t-value (TBP), and Cross-Date Index (CDI) were calculated. The mean sensitivity (MS) is the mean percentage change from each measured yearly ring value to the next (Douglass 1936). Standard deviation (Std) is the measure of high-frequency variations (Fritts 1976). Tendency changes (TC) were calculated over the points of the running window, while creating a new time series and this new time series shows the variations of the calculated parameter along the original series (Rinn 2003). The first order autocorrelation AR(1) was calculated to estimate serial correlation (Fritts 1976). To express the quality of accordance between time series, Gleichläufigkeit (Glk), Baillie-Pilcher value (TBP) (1973), and Cross-Date Index (CDI) were used. These parameters are characterized by different sensitivities to tree-ring patterns. Gleichläufigkeit (Glk) represents the overall accordance of two series, while Baillie-Pilcher value (TBP) presents the correlation significance, and Cross-Date Index (CDI) is the combination of these two parameters (Glk and TBP), which is a date index of possible series matches (Rinn 2003). Values of Glk greater than 60% and values of TBP greater than 3.0 and CDI = 10 were considered significant. Standardization Measured tree-ring series were standardized using the program ARSTAN (Cook 1985). All series were detrended using a negative exponential curve. Index values were calculated as ratios between the actual and fitted values. Index values were then prewhitened using an autoregressive model selected on the basis of the minimum Akaike (1974) criterion and combined across all series using biweight robust estimation of the mean to exclude the influence of the outliers (Cook 1985). As recommended by Henderson and Grissino- Mayer (2009), we used the residual chronology because it lowers the bias inherent in tree-ring indices and therefore, provides a more rigorous assessment of climatic influences. As a result of standardization, chronologies for the mesotrophic and the heath forest site type were compiled. As a reference chronology, we used the chronology for Scots pine compiled by Erich Lõhmus (1992a). For his chronology, Erich Lõhmus used the detrending method of the moving average of 21 years. Climate data Sums of monthly precipitation (mm) and air temperature means ( C) were obtained from Kunda (for North-East, location N E; period ), Ristna (for islands and South-West, location N E; period ) and Tõravere (for South-East, location N E; period ) stations of the Estonian Meteorological and Hydrological Institute (EMHI). To illustrate climate data, average monthly temperatures, the mean temperature of the vegetation period (from April to September) and mean annual temperatures as well as monthly sums of precipitation, the sum of precipitation in the vegetation period and the annual sum of precipitation were calculated over the periods (Figure 3). In general, Estonia has a temperate climate, with warm summers and severe winters. The average annual temperature is 4 6 C. The annual sum of precipitation is between 500 mm and 750 mm, about mm of which falls down as snow. The active period of vegetation growth (daily air temperature above 5 C) mostly lasts between 170 and 180 days per year. Analysis of climate-tree growth relationships Relationships between climate variables and the tree radial increment were evaluated using the pointer year analysis and correlations. The pointer year analysis is an accepted method of showing annual growth reactions due to abrupt changes in environmental conditions (Cropper 1979, Schweingruber 1990), especially those due to climate variations (Rolland 1993, Kroupova 2002, Neuwirth et al. 2004, Karpavièius and Vitas 2006, Elferts 2007). To calculate pointer years, we used Cropper (1979) method, where ratios among the raw annual measurements for single tree series and their 13-year moving average were calculated: where: x i tree-ring width in year i; mean[window] and stdev[window] arithmetic mean and standard deviation of ring widths in the moving window x i-6, x i-5, x i-4, x i-3 x i-2, x i-1, x i, x i+1, x i+2, x i+3, x i+4, x i+5, x i+6. Years with a 2009, Vol. 15, No. 2 (29) ISSN

111 BALTIC FORESTRY RESPONSE OF SCOTS PINE /.../ RADIAL GROWTH TO CLIMATE FACTORS IN ESTONIA M. HORDO ET AL. Mean temperature C Mean temperature C Jan Jan Mean temperature Sum of precipitation Feb Mar Apr May Mean temperature Sum of precipitation Feb Mar Apr Jun North-East Jul Aug Mo nths Sep Oct Islands and South-West May Jun Jul Aug Months Sep Oct Nov Nov Dec Dec Vegper(A-S) Vegper(A-S) Annual Annual Sum of precipitation m m Sum of precipitation mm by regions was calculated and the t-test for the comparison of the means of climate data between normal years and strong/extreme pointer years was performed using SAS software (SAS 1996). Tree-ring chronologies While building up the general dendrochronology it is important to retain growth fluctuations by climate factors within the chronology when larger areas are being summarized, with different growing conditions. In this study, 889 trees from 119 plots were sampled from the Estonian network of forest growth and yield research plots and separate chronologies were built for mesotrophic and heath forest site types (Figure 4) for different regions of Estonia (Figure 5, Figure 6). The chronology for the mesotrophic forest site type is 145 years long, and based on 602 trees, covering the period ; whereas the chronology for the heath forest site type is 221 years long and based on 287 trees, covering the years from 1786 to Mean temperature C Jan Mean temperature Sum of precipitation Feb Mar Apr May Jun South-East Jul Aug Months Sep Oct Nov Dec Vegper(A-S) Annual Sum of precipitation mm Index Heath forest site type Mesotrophic forest site type Scots pine general chronology by E. Lõhmus Year Figure 4. Indexed chronologies of Scots pine for heath and mesotrophic forest types and the reference chronology by Lõhmus (1992a) Figure 3. Mean monthly temperatures and sums of precipitation averages from Kunda (period ), Ristna (period ), and Tõravere (period ) weather stations of EMHI South-West South-East North-East Islands value of Z i that was higher or lower than 1 or -1 (Neuwirth et al. 2007, Pourtahmasi et al. 2007) were defined as positive or negative pointer years, respectively. Those positive and negative Cropper values were divided into three classes by intensity: weak for, strong for and extreme for (Neuwirth et al. 2007). Pearson s correlation coefficient between Cropper values Z i and climate variables (the mean temperature and the sum of precipitation) for both forest site types Index Year Figure 5. Dendrochronological series for mesotrophic forest types of Scots pine by regions 2009, Vol. 15, No. 2 (29) 198 ISSN

112 RESPONSE OF SCOTS PINE /.../ RADIAL GROWTH TO CLIMATE FACTORS IN ESTONIA BALTIC FORESTRY M. HORDO ET AL. Index South-East North-East Islands Year Figure 6. Dendrochronological series for heath forest types of Scots pine by regions Updated chronologies were compared to the reference chronology built by Lõhmus (1992a) and measures of accordance were calculated (Table 1). Glk between the mesotrophic forest type and the reference chronology ranged from 56.1 to 77.2%, and between the heath forest type and the reference chronology varied from 72.8 to 80.7%. In South-West region, Glk was not significant. The TBP showed significant conformity between new chronologies and then the reference chronology, 7.1 for the mesotrophic and 5.0 for the heath forest type. Consequently, the overall conformity between new and reference chronologies was statistically significant. Intercorrelation among regional chronologies within a forest site type was statistically significant (p < 0.05). Descriptive statistics for indexed tree-ring series were calculated. The mean sensitivity varied between 8 and 14% (Table 1), whereas the values for mesotrophic forests were lower than those for heath forests, which shows that heath forests are more sensi- tive to climatic factors. The low first-order autocorrelations indicate that the influence of climate on the inter-annual growth variability of pine is high (Wigley et al. 1984, Pourtahmasi et al. 2007). We found out that ring width growth was mostly influenced by the previous year weather conditions, in the heath forest site type in South-East (AR(1) = 0.78). At the same time, a low influence of the previous year was detected in North-East (the mesotrophic forest site type) and islands (the heath forest site type), with AR(1) = All these findings above confirm that the pine chronologies from our network of sample plot data are suitable for studying the effects of climate on the radial growth. Correlation analysis The results of correlation analyses between the radial growth indices and climate data, which include monthly temperatures and precipitation from the previous growing season to this year, are presented in Figure 7. On both forest sites, correlation analyses revealed that tree-width growth is positively correlated with winters prior to the growing season temperature and the temperature during the vegetation period. Additionally, tree growth is significantly negatively correlated with the previous year August temperature and positively correlated with sum of the precipitation of this month. This indicates that high temperature and low amount of precipitation at that time is limiting the increment growth of pine. The analysis showed that in southwest Estonia in mesotrophic forest site types, precipitation in February has a significant positive influence on the radial growth. Considering that February is one of the coldest months in Estonia, with the average temperature from 3.3 to 7.4 C, higher Table 1. Internal statistical properties of time series of Scots pine forest types by regions (generalized chronology by Lõhmus was compared to sample plots chronologies). Std = standard deviaton; AR = first-order autocorrelation (lag = 1); MS = mean sensitivity; TC = tendency changes; Glk = Gleichläufigkeit; TBP = T-value with Baillie-Pilcher-Standardization; CDI = Cross Date Index; * 95% significance for the Glk value 2009, Vol. 15, No. 2 (29) ISSN

113 BALTIC FORESTRY RESPONSE OF SCOTS PINE /.../ RADIAL GROWTH TO CLIMATE FACTORS IN ESTONIA M. HORDO ET AL Temperature - Mesotrophic forest site type South-West South-East North-East Islands Correlation coefficient Jun Jul Aug Sept Oct Nov Dec Jan Feb Mar Apr Mai Jun Jul Aug Sept Oct Nov Dec VegPer AnnMean Months Precipitation - Mesotrophic forest site type South-West South-East North-East Islands Correlation coefficient Jun Jul Aug Sept Oct Nov Dec Jan Feb Mar Apr Mai Jun Jul Aug Sept Oct Nov Dec VegPer AnnSum Months Temperature - Heath forest site type South-East North-East Islands 0.3 Correlation coefficient Jul Jun Aug Sept Oct Nov Dec Jan Feb Mar Apr Mai Jun Jul Aug Sept Oct Nov Dec VegPer AnnMean Months Precipitation - Heath forest site type South-East North-East Islands Figure 7. Correlation coefficients for Scots pine chronologies (mesotrophic and heath forest site types) between ring-width indices and total precipitation and the mean monthly temperature. Dashed lines mark the significant level (r = or r = 0.273) at p = 0.05 Correlation coefficient Jun Jul Aug Sept Oct Nov Dec Jan Feb Mai Apr Mar Months Jun Jul Aug Sept Oct Nov Dec VegPer AnnMean 2009, Vol. 15, No. 2 (29) 200 ISSN

114 RESPONSE OF SCOTS PINE /.../ RADIAL GROWTH TO CLIMATE FACTORS IN ESTONIA BALTIC FORESTRY M. HORDO ET AL. amount of precipitation (snow) would stimulate tree growth in spring. On the islands, precipitation in April has a significant negative impact on the radial increment. The annual mean temperature is significantly positively correlated with tree-ring indices of pine on both sites. All the findings above suggest that the mean annual temperature significantly depends on winter climate, and when severe winter was followed by a long and cool spring, such factors limit tree growth. Pointer year analysis The Cropper values were calculated from single tree curves for mesotrophic and heath forest sites in different regions as growing conditions varied. Extreme negative and positive pointer years were identified when the Cropper value was lower than or higher than and strong pointer years were below 1.28 and above 1.28, respectively. On mesotrophic forest sites significant extreme negative pointer years of 1929 and 1940 in the southwest region were detected and 1934, a positive year; in the southeast region, an extreme negative pointer year was 1940 and a positive event year was 1945; in the northeast region, considerably negative pointer years were 1937 and 1985 and there were no significant positive years; on islands, significant negative years were 1901, 1940, 1985 and the extreme positive pointer year was On heath forest sites significant negative extreme years were detected 1906, 1931, 1932 and 1942 in the southeast region, and positive years were 1921, 1923, 1945, 1946, 1988, and 1997; in the northeast region, important positive extreme years were 1921, 1967, 1980, 1981, 1989, 1990 and negative years were 1920, 1940, 1941 and 1985; on islands, significant positive years were 1910, 1938, 1945 and negative years were 1940, 1941, and The results of the pointer year analysis are presented in Table 2. The analysis of pointer years identified 19 positive and 19 negative years on mesotrophic forest site types, and 34 positive and 27 negative years on heath forest site types. According to analyses, 1940 and 1985 were the most significant negative years. The records of EMHI prove that the most severe winters over the past century were in 1939/40, 1940/41, 1941/ 42 and cold winters were in 1984/85, 1986/87, 1995/96. This indicates that the cold winter prior to the growing season and late spring (mean T Mar May +2. C) affected the radial increment of pine tree growth and it was the main cause of the sharp decrease in the radial growth. Similar results were obtained in Estonia by Läänelaid and Eckstein (2003) and in Lithuania by Vitas (2008). Extreme events usually last only a few months and the organism can survive these conditions Table 2. Negative ( ) and positive (+) pointer years regions (extreme event years are highlighted) of the radial growth of Scots pine by 2009, Vol. 15, No. 2 (29) ISSN

115 D N S A J J A F J D N S A J J A F J BALTIC FORESTRY RESPONSE OF SCOTS PINE /.../ RADIAL GROWTH TO CLIMATE FACTORS IN ESTONIA M. HORDO ET AL. (Ahas et al. 2000), but the influence of extreme events may last for several years (Jaagus et al. 2003). The winter severity determines directly when the spring begins. In most cases, a severe winter is followed by a late, cool spring and an early and warm spring followed by a mild winter (Jaagus et al. 2003). The pointer year analysis revealed that significant positive pointer years were 1945, 1946, 1989 and According to the EMHI, warm winters appeared in 1945 (mean T Dec- Feb 3.9 seems that the warm winter of the previous year had a positive effect on increment growth. Additionally, growth-climate relationships were tested with the correlation of strong/extreme Cropper values and climate data. The results of the analysis are presented in Figure 8. The results indicate a positive correlation of Cropper value with the mean temperature in winter, particularly in January and February, and with the temperature in the beginning of the growing season (March, April). However, no negative correlation with average winter temperatures was found, but in an earlier study, Lõhmus (1992b) detected a high negative correlation of the radial growth with winter minimum temperature. A statistically significant positive correlation with temperatures in June and July was detected on the mesotrophic forest site type, while the temperature was significant for growth on heath forest site types in September and October. Scots pine showed a statistically significant positive response to the mean annual temperature and the temperature of the growing season. Analysis revealed that the high temperature of the previous year August had a negative effect on the radial growth, but the precipitation in this month had a positive influence on the radial increment. This negative effect of the temperature on the radial growth could be explained by the promotion of bud differentiation during this time. The late summer temperature affects the amount of nutrient storage which encourages sprouting in spring. This in turn affects the next growing year ratio of sprouts and diameter increment (Lõhmus 1992b). A similar effect is mentioned in Finland by Henttonen (1984) and in Sweden by Jonnson (Lõhmus 1992b). In mesotrophic forest site types, annual precipitation had a negative impact, while on heath forest site types, the sum of precipitation was important during the vegetation period. The results of the analysis showed that the impact of precipitation on the radial growth of trees in comparison to temperatures was not so significant. Temperature may be considered as the most important single factor initiating growth activity (Vaganov et al. 2006); however, low humidity can cause an earlier termination of growth in a season (Fritts 1976), in our case on heath forest site types. A combination of temperature and humidity changes in particular intervals of a season produces acceleration or deceleration of growth processes (Schweingruber 1996). It is confirmed by Jaagus et al. (2003, 2006) that beside temperature, precipitation has the most profound effect on the Mesotrophic forest type Temperature Precipitation Correlation coefficient Ann prec Vegp prec D N O S A J J M A M F J pd pn po ps pa pj pj Ann temp Vegp temp D N O S A J J M A M F J pd pn po ps pa pj pj Month Figure 8. Correlation between monthly and seasonal climate data (temperature and precipitation) and Cropper-values in heath and mesotrophic forest sites; responses were similar in all regions. indicates statistical significance (p<0.05) Correlation Coefficient pj pj pa ps po pn pd M M O Heath forest type Vegp temp Ann temp pj Month pj pa ps po pn pd M M O Temperature Precipitation Ann prec Vegp prec 2009, Vol. 15, No. 2 (29) 202 ISSN

116 RESPONSE OF SCOTS PINE /.../ RADIAL GROWTH TO CLIMATE FACTORS IN ESTONIA BALTIC FORESTRY M. HORDO ET AL. growth; however, precipitation is an extremely unstable climate variable which makes long-term changes almost impossible to predict. Finally, we used the t-test on climate data to analyze the differences between normal years and strong/ extreme years. The results of the test confirmed our previous results that positive and negative pointer year had a significant effect on the mean annual temperature and the temperature of the growing season, as well as on the temperature and the sum of precipitation in August of the previous year. This means that extreme climate events are influencing the radial increment growth of pine trees. In this study we compiled new chronologies for mesotrophic and heath sites for Scots pine and analyzed the responses of the radial increment to climatic factors. Measures of conformity Glk between the mesotrophic forest type and the reference chronology ranged from 56.1 to 77.2% and between the heath forest type and the reference chronology varied from 72.8 to 80.7%. Therefore, the overall conformity between new and reference chronologies was statistically significant (p < 0.05). The results of correlation analyses between radial growth indices and climate data, on both forest sites revealed that tree-width growth is positively correlated with temperatures in winter time and prior growing season, and temperatures during growth season (r = 0.273; p > 0.05). This means that the mean annual temperature significantly depends on winter climate, and when severe winter was followed by a long and cool spring, such factors limited tree growth. The pointer year analysis identified 19 positive and 19 negative years on mesotrophic forest site types, and 34 positive and 27 negative event years on heath forest site types. The analysis revealed that 1940 and 1985 were the most significant negative years, while significant positive pointer years were 1945, 1946, 1989 and Additionally, growth-climate relationships from the correlation analysis revealed the statistically most significant positive response to the mean annual temperature and the temperature of growing season (p < 0.05) in both forest site types, while the monthly consequence of the correlation varied. 2009, Vol. 15, No. 2 (29) The collection of the increment core data from the network of Estonian forest growth and yield research plots was supported by the Environmental Investment Centre. Estonian Meteorological and Hydrological Institute provided climate data. This study was supported by The Ministry of Education and Research (project SF s08). Ahas, R., Jaagus, J. and Aasa, A The phenological calendar of Estonia and its correlation with mean air temperature. International Journal of Biometeorology 44 (4): Akaike, H A new look at the statistical model identification. IEEE Transactions on Automatic Contro 19 (6): Baillie M.G.L. and Pilcher, J.R A sample crossdating program for tree-ring research. Tree-Ring Bulletin 33: Carrer, M. and Urbinati, C Age-dependent tree-ring growth responses to climate in Larix decidua and Pinus cembra. Ecology 85 (3): Cedro, A Dependence of radial growth of Pinus sylvetris L. from western Pomerania on the rainfall and temperature conditions. Geochronometria 20: Cook, E.R A time series analysis approach to tree ring standardization. Dissertation, University of Arizon, 171 p. Cook, E.R. and Kairiukstis, L.A Methods of Dendrochronology: Applications in the Environmental Sciences. Kluwer Academic Publischers, 394 p. Cropper, J.P Tree-ring skeleton plotting by computer. Tree-Ring Bulletin 39: Douglass, A.E Climatic Cycles and Tree Growth, Volume III. A Study of Cycles. Carnegie Institution of Washington Publication 289: 171 p. Elferts, D Scots pine pointer-years in northwestern Latvia and their relationship with climatic factors. Biology, Acta Universitatis Latviensis 723: Etverk, I., Karoles, K., Lõhmus, E., Meikar, T., Männi, R., Nurk, T., Pikk, J., Randveer, T., Tamm, Ü., Veibri, U. and Örd, A Estonian Forests and Forestry. Estonian Forest Department, Tallinn, 128 p. Fabian, P. and Menzel, A Change in phenology of trees in Europe. In: Karjalainen, T., Spiecker, H. and Laroussinie, O. (eds.), Causes and Consequences of Accelerating Tree Growth in Europe. European Forest Institute Proceedings, 27: Fritts, H.C Tree Rings and Climate. Academic Press, London, 567 p. Gaucherela, C., Campillob, F., Missonc, L., Guiota, J. and Boreuxd, J.-J Parameterization of a process-based tree-growth model: Comparison of optimization, MCMC and Particle Filtering algorithms. Environmental Modelling & Software 23 (10 11): Grissino-Mayer, H.D Evaluating crossdating accuracy: a manual and tutorial for the computer program COFECHA. Tree-Ring Research 57: Gustavsen, H.G., Roiko-Jokela, P. and Varmola, M Kivennäismaiden talousmetsien pysyvat (INKA ja TIN- KA) kokeet. Suunnitelmat, mittausmenetelmät ja aineistojen rakenteet. Metsäntutkimuslaitoksen tiedonantoja 292: 212 p. (in Finnish). Helama, S. and Lindholm, M Droughts and rainfall in south-eastern Finnland since AD 874, inferred from Scots pine ring-widths. Boreal Environment Research 8: Henderson, J.P. and Grissino-Mayer, H.D Climatetree growth relationships of longleaf pine (Pinus palus- ISSN

117 BALTIC FORESTRY RESPONSE OF SCOTS PINE /.../ RADIAL GROWTH TO CLIMATE FACTORS IN ESTONIA M. HORDO ET AL. tris Mill.) in the Southeastern Coastal Plain, USA. Dendrochronologia 27(1): Henttonen, H The dependence of annual ring indices on some climatic factors. Acta Forestalia Fennica 186: Holmes, R.L Computer-assisted quality control in treering dating and measurement. Research report. Tree-Ring Bulletin 43: Hordo, M Erindite ja / või mõõtmisvigade avastamise meetoditest puistu kasvukäigu püsiproovitükkide andmestikul. [Outlier and/or measurement errors on the permanent sample plot data]. Metsanduslikud uurimused / Forestry Studies 43: 9 23 (in Estonian). Hynynen, J Predicting the growth response to thinning for Scots pine stands using individual-tree growth models. Silva Fennica 29 (3): Jaagus, J Climatic changes in Estonia during the second half of the 20th century in relationship with changes in large-scale atmospheric circulation. Theoretical and Applied Climatology 83 (1 4): Jaagus, J., Truu, J., Ahas, R. and Aasa, A Spatial and temporal variability of climatic seasons on the East European Plain in relation to large-scale atmospheric circulation. Climate Research 23 (2): Karpavièius, J. and Vitas, A Influence of environmental and climatic factors on the radial growth of European ash (Fraxinus excelsior L.). Ecologija 1: 1 9. Kiviste, A. and Hordo, M Eesti metsa kasvukäigu püsiproovitükkide võrgustik. [Network of permanent forest growth plots in Estonia]. Metsanduslikud uurimused / Forestry Studies 37: (in Estonian). Kiviste, A., Aiaots, J., Hordo, M. and Sims, A Puude väljalangevust mõjutavad faktorid puistu kasvukäigu püsiproovitükkide andmete põhjal [Factors influencing tree mortality on the basis of permanent forest sample plot data]. Metsanduslikud Uurimused / Forestry Studies 43: (in Estonian). Kiviste, A., Nilson, A., Hordo, M. and Merenäkk, M Diameter distribution models and height-diameter equations for Estonian forest. In: Amario, A., Reed, D. and Soares, P. (eds.), Modelling Forest Systems. CABI Publishing, London, p Kroupova, M Dendroecological study of spruce growth in regions under long-term air pollution load. Journal of Forest Science 48: Läänelaid, A Dendrochronological dating of the Uppsala house in Tartu, Estonia. Dendrochronologia 15: Läänelaid, A Tree-ring dating in Estonia. Doctoral dissertation, University of Helsinki, Helsinki, 98 p. Läänelaid, A. and Eckstein, D Development of a Tree-ring Chronology of Scots pine (Pinus sylvestris L.) for Estonia as a Dating Tool and Climatic Proxy. Baltic Forestry 9 (2): Lõhmus, E. 1992a. Eesti männikute dendrokronoloogiline üldskaala [Scots pine generalized chronology in Estonia]. Metsanduslikud Uurimused / Forestry Studies 24: (in Estonian). Lõhmus, E. 1992b. Hariliku männi radiaalse juurdekasvu seosest meteoroloogiliste teguritega [The dependence of Scots pine annual ring indices on some climatic factors]. Metsanduslikud Uurimused / Forestry Studies 25: (in Estonian). Lõhmus, E Eesti metsakasvukohatüübid. [Estonian forest site types]. EPMÜ Metsanduslik Uurimisinstituut, 80 p. (in Estonian). Mielikäinen, K Koivusekoituksen vaikutus kuusikon rakenteeseen ja kehitykseen [Effect of an admixture of birch on the structure and development of Norway spruce stands]. Communicationes Instituti forestalis, Metsäntutkimuslaitos, 133: 79 p. (in Finnish). MMK, Aastaraamat Mets 2007 [Yearbook Forest 2007]. Tartu, 217 p. (in Estonian). Neuwirth, B., Esper, J., Schweingruber, F.H. and Winiger, M Site ecological differences to the climatic forcing of spruce pointer years from Lötschental, Switzerland. Dendrochronologia 21 (2): Neuwirth, B., Schweingruber, F.H. and Winiger, M Spatial patterns of central European pointer years from Dendrochronologia 24: Nilson, A Fragmente puistu kasvu ja ehituse mudelitest [Some fragments of stand growth and structure models]. Metsanduslikud Uurimused / Forestry Studies 37: 9 20 (in Estonian). Pärn, H Radial Growth Response of Scots Pine to Climate Under Dust Pollution in Northeast Estonia. Water Air and Soil Pollution 144 (1): Pärn, H Hariliku männi puistute radiaalkasvu ja kliimategurite vaheliste seoste ajalisest varieeruvusest [Temporal variability in the relationships between the radial growth of Scots pine stands and the climate]. Metsanduslikud Uurimused / Forestry Studies 40: (in Estonian). Pärn, H Radial growth of conifers in regions of different cement dust loads. Ökoloogia 55 (2): Pilcher, J.R Sample Preparation, Cross-dating, and Measurement. In: Cook, E. and Kairiukstis, L. (ed.), Methods of Dendrochronology: applications in the Environmental Sciences. Kluwer Academic Publishers, p Pourtahmasi, K., Parsapajouh, D., Bräuning, A., Esper, J. and, Schweingruber, F.H Climatic analysis of pointer years in tree-ring chronologies from Northern Iran and neighboring high mountain areas. Geoöko 28: Rinn, F TSAP-Win. Time Series Analysis and Presentation for Dendrochronology and Related Applications. User Reference, Heidelberg, 91 p. Rolland, C Tree-ring and climate relationships for Abies alba in the internal Alps. Tree-ring Bulletin 53: Sachs, L Applied statistics. Springer Series in Statistics, New York, 706 p. SAS Institute Inc SAS/STAT Software: changes and enhancements through release Cary, North Carolina, SAS Institute Inc., 1104 p. Schweingruber, F.H Dendroecological Information in Pointer Years and Abrupt Growth Changes. In: Cook, E. and Kairiukstis, L. (eds.), Methods of Dendrochronology: applications in the Environmental Sciences. Kluwer Academic Publishers, p Schweingruber, F.H Tree Rings and Environment. Dendroecology. Vienna, 609 p. Spiecker, H Tree rings and forest management in Europe. Dendrochronologia, 20 (1 2): Spiecker, H., Mielikäinen, K., Köhl, M. and Skovsgaard, J.P Growth Trends in European Forests. European Forest Institute Research Report, 5: 372 p. Vaganov, E.A., Hughes, M.K. and Shashkin A.V Growth dynamics of conifer tree rings: images of past and future environments. Springer, Berlin, 354 p. Vitas, A Tree-ring chronology of Scots pine (Pinus sylvestris L.) for Lithuania. Baltic Forestry 14 (2): Vitas, A. and Erlickytë, R Influenceof droughts. Influence of Droughts to the Radial Growth of Scots Pine 2009, Vol. 15, No. 2 (29) 204 ISSN

118 RESPONSE OF SCOTS PINE /.../ RADIAL GROWTH TO CLIMATE FACTORS IN ESTONIA BALTIC FORESTRY M. HORDO ET AL. (Pinus sylvestris L.) at Different Site Conditions. Baltic Forestry 13 (1): Wigley, T.M.L., Briffa, K.R. and Jones, P.D On the Average Value of Correlated Time Series, with Applications in Dendroclimatology and Hydrometeorology. Journal of Climate and Applied Meteorology 23: Worbes, M Mensuration. Tree-Ring Analysis. Encyclopedia of Forest Sciences, Zahner, R A model for tree-ring time series to detect regional growth changes in young, evenaged forest stands. Tree-Ring Bulletin 48: ÂËÈßÍÈÅ ÊËÈÌÀÒÈÅÑÊÈÕ ÔÀÊÒÎÐΠÍÀ ÐÀÄÈÀËÜÍÛÉ ÏÐÈÐÎÑÒ ÑÎÑÍÛ ÎÁÛÊÍÎÂÅÍÍÎÉ (PINUS SYLVESTRIS L.)  ÝÑÒÎÍÈÈ Ì. Õîðäî, Ñ. Ìåòñëàéä è À. Kèâèñòå Ðåçþìå  äàííîé ñòàòüå àíàëèçèðóåòñÿ âëèÿíèå êëèìàòè åñêèõ ôàêòîðîâ íà ðàäèàëüíûé ïðèðîñò ñîñíû îáûêíîâåííîé (Pinus sylvestris L.) â ìåçîòðîôíûõ è âåðåùàòíèêîâûõ ëåñàõ Ýñòîíèè. Õðîíîëîãèÿ ñîñíû îáûêíîâåííîé, ðàñòóùåé â ìåçîòðîôíûõ è âåðåùàòíèêîâûõ ëåñàõ, áûëà ñîñòàâëåíà íà îñíîâàíèè àíàëèçà ãîäè íûõ êîëåö 889 äåðåâüåâ, ïîëó åííûõ â ñåòè 119 ïðîáíûõ ó àñòêîâ. Oòíîøåíèÿ ìåæäó êëèìàòè åêèìè ôàêòîðàìè è ðàäèàëüíûì ïðèðîcòîì áûëè îöåíåíû, èñïîëüçóÿ êîýôôèöèåíò êîððåëÿöèè Ïèðñîíà, à òàêæå ïðè ïîìîùè àíàëèçà ðåïåðíûõ ëåò. Åäèíè íûå äåðåâüÿ áûëè îöåíåíû ïðè ïîìîùè ìåòîäà Êðîïïåðà. Ñîãëàñíî ðåçóëüòàòàì àíàëèçà, 1940 è 1985 áûëè íåãàòèâíûìè ðåïåðíûìè ãîäàìè, òîãäà êàê 1945, 1946, 1989 è 1990 ïîçèòèâíûìè. Ýòà òåíäåíöèÿ ñîõðàíÿåòñÿ äëÿ èçó åííûõ ìåñò ïðîèçðîñòàíèÿ. Ýêñòðåìàëüíûå çíà åíèÿ Êðîïïåðà èìåþò ïîçèòèâíóþ êîððåëÿöèþ ñî ñðåäíåé ìåñÿ íîé òåìïåðàòóðîé çèìîé è ðàííåé âåñíîé. Íåãàòèâíàÿ êîððåëÿöèÿ íàáëþäàåòñÿ ñ êîëè åñòâîì îñàäêîâ â àâãóñòå ïðîøëîãî ãîäà. Òåìïåðàòóðà ÿâëÿåòñÿ íàèáîëåå âàæíûì ôàêòîðîì, âëèÿþùåì íà èíòåíñèâíîñòü ðàäèàëüíîãî ïðèðîñòà. Àíàëèç ðåïåðíûõ ëåò ïîäòâåðäèë, òî ñóðîâàÿ çèìà, ïîçäíÿÿ âåñíà, à òàêæå ñóõîå ëåòî, ÿâëÿþòñÿ îñíîâíûìè ïðè èíàìè ðåçêîãî ñîêðàùåíèÿ ðàäèàëüíîãî ïðèðîñòà. Êëþ åâûå ñëîâà: ðàäèàëüíûé ðîñò, êëèìàò, ðåïåðíûé ãîä, ñîñíà îáûêíîâåííàÿ, õðîíîëîãèÿ ãîäè íûõ êîëåö 2009, Vol. 15, No. 2 (29) ISSN

119 118 III

120 Metslaid, S., Stanturf, J.A., Hordo, M., Korjus, H., Laarmann, D., Kiviste, A Growth responses of Scots pine to climatic factors on reclaimed oil shale mined land. Environmental Science and Pollution Research, 23:

121

122 Your article is protected by copyright and all rights are held exclusively by Springer- Verlag Berlin Heidelberg. This e-offprint is for personal use only and shall not be selfarchived in electronic repositories. If you wish to self-archive your article, please use the accepted manuscript version for posting on your own website. You may further deposit the accepted manuscript version in any repository, provided it is only made publicly available 12 months after official publication or later and provided acknowledgement is given to the original source of publication and a link is inserted to the published article on Springer's website. The link must be accompanied by the following text: "The final publication is available at link.springer.com

123 Author's personal copy Environ Sci Pollut Res DOI /s HOW CAN WE RESTORE THE BIODIVERSITY AND ECOSYSTEM SERVICES IN MINING AND INDUSTRIAL SITES? Growth responses of Scots pine to climatic factors on reclaimed oil shale mined land Sandra Metslaid 1 & John A. Stanturf 2 & Maris Hordo 1 & Henn Korjus 1 & Diana Laarmann 1 & Andres Kiviste 1 Received: 15 April 2015 /Accepted: 19 October 2015 # Springer-Verlag Berlin Heidelberg 2015 Abstract Afforestation on reclaimed mining areas has high ecological and economic importance. However, ecosystems established on post-mining substrate can become vulnerable due to climate variability. We used tree-ring data and dendrochronological techniques to study the relationship between climate variables and annual growth of Scots pine (Pinus sylvestris L.) growing on reclaimed open cast oil shale mining areas in Northeast Estonia. Chronologies for trees of different age classes (50, 40, 30) were developed. Pearson s correlation analysis between radial growth indices and monthly climate variables revealed that precipitation in June July and higher mean temperatures in spring season enhanced radial growth of pine plantations, while higher than average temperatures in summer months inhibited wood production. Sensitivity of radial increment to climatic factors on post-mining soils was not homogenous among the studied populations. Older trees growing on more developed soils were more sensitive to precipitation deficit in summer, while growth indices of two other stand groups (young and middle-aged) were highly correlated to temperature. High mean temperatures in August were negatively related to annual wood production in all trees, while trees in the youngest stands benefited from warmer temperatures in January. As a response to thinning, mean annual basal area increment increased up to 50 %. By managing tree Responsible editor: Philippe Garrigues * Sandra Metslaid Sandra.Metslaid@emu.ee 1 2 Department of Forest Management, Institute of Forestry and Rural Engineering, Estonian University of Life Sciences, Kreutzwaldi 5, Tartu, Estonia Center for Forest Disturbance Science, USDA Forest Service, 320 Green Street, Athens, GA 30602, USA competition in the closed-canopy stands, through the thinning activities, tree sensitivity and response to climate could be manipulated. Keywords Reclamation. Post-mining area. Afforestation. Tree-ring. Scots pine. Forest management Introduction Afforestation on reclaimed mining areas has high ecological and economic importance. However, ecosystems established on post-mining substrate can become vulnerable due to climate variability. Reclamation of oil shale open surface-mined lands in Estonia is mainly done by afforestation with the native Scots pine (Pinus sylvestris L). Since the 1960s, when systematic reclamation of post-mining areas started (Korjus et al. 2007), more than 16,000 ha have been afforested (Yearbook Forest 2014). To a large extent, the growth of forest plantations established on such reclaimed areas depends on the physical (e.g., Torbet et al. 1990) and chemical properties (e.g., Hüttl and Weber 2001) of the growth medium where tree roots are concentrated. In general, substrate conditions for plant growth on reclaimed post-mining areas are harsh due to shallow soil, high ( %) content of coarse unweathered rocks (Reintam 2004; Roberts et al. 1988), and low organic matter and nutrient content (especially N). Alkaline soil reaction persists in such substrates for decades (Reintam et al. 2002) and may inhibit natural tree establishment (Pensa et al. 2004) or reduce the survival of planted trees (Kaar 2002). These factors restrict tree species selection and as a result, single-species forest ecosystems of low stability are being established on large areas of degraded land. 122

124 Author's personal copy Environ Sci Pollut Res Scots pine is able to grow on infertile soils and has been widely used for reclamation of mining areas in Estonia (Kaar 2002) and throughout Europe (Hüttl and Weber 2001; Baumann et al. 2006; Pietrzykowski and Socha 2011). Despite limiting soil conditions, forest plantations on mined sites in Estonia show high productivity (Reintam and Kaar 2002)and are comparable to pine growth on fertile forest sites (Korjus et al. 2007; Kiviste et al. 2010). Plantations on post-mining areas are important for carbon sequestration (Karu et al. 2009) and play a significant ecological role in promoting soil development (Reintam 2001). Besides, these afforested areas are managed generally for timber production. Previous studies carried out on the reclaimed areas of oil shale excavation have investigated tree species selection (Kaar 2002), tree survival, and biomass production (Kuznetsova et al. 2011) as well as ecosystem restoration possibilities (Laarmann et al. 2015). Growth of the plantations has been extensively studied for young trees (Vaus 1970; Kuznetsova et al. 2011), but only a few studies on growth dynamics and productivity (e.g., Korjus et al. 2007; Kiviste et al. 2010) are based on long-term observations. Recently, changing climatic conditions have been suggested to have a pronounced effect on forest growth and productivity. Knowledge about climate-growth relationships is essential for the evaluation of possible impact of climate change on the productivity and vitality of pine plantations of reclaimed areas. Several previous studies (Linderholm 2001; Weber et al. 2007; Hökkä et al. 2012) showed that climatic sensitivity of Scots pine greatly depends on local soil properties, while the influence of climate variables (temperature or precipitation) differs along the latitudinal gradient (Helama et al. 2005; Hordoetal. 2011; Henttonen et al. 2014). Pine trees are more sensitive to variation in temperature at high latitude sites, while this relationship changes, moving toward the South, where stronger relationships between growth and precipitation are usually detected. In Estonia, several studies have investigated Scots pine radial growth response to climate variables (e.g., Läänelaid and Eckstein 2003; Hordo et al. 2009) but mainly considered trees growing on forest sites. Some attempts to relate growth of young trees, growing on post-mining areas to precipitation in spring were described by Vaus (1970), but no thorough dendroclimatic assessment of older trees growing on highly degraded soils in Northeast Estonia has ever been conducted. The aim of the present study was to investigate radial growth variability and climate influence on the growth of pine monocultures on reclaimed areas across several combinations of substrate and stand structural conditions. The specific aims were (a) to identify the main climatic factors driving annual diameter growth of Scots pine established on reclaimed oil shale mining areas, (b) to examine how the relationship to climate variables is affected by soil and forest stand characteristics, and (c) to assess possible effects of climate, site productivity, and changes in stand structure on annual basal area increment. Material and methods Study area The study was carried out in Northeast Estonia (Fig. 1), in Narva (Viivikonna section) oil shale quarry ( N, E, elevation m a.s.l.) that is a part of the largest commercially exploited oil shale mining area in Estonia (Raukas and Punning 2009). Lowlands prevail in the surrounding undisturbed landscape, while the topography of post-mining areas has been drastically altered by large-scale surface mining activity and mainly consists of artificial hills (leveled spoil heaps) covered with forest plantations (Sepp et al. 2010). Although the machinery used in oil shale extraction industry has changed over time, there were no major differences in oil shale extraction and reclamation methods in our sites. The access to commercially valuable layers of oil shale was gained by removing natural ecosystems, and topsoil was stockpiled and used during reclamation. The average oil shale extraction depth was 7.4 m and did not differ significantly among the sites. The remaining surface and rock overburden was removed and underwent mechanical reduction. The oil-bearing shale was excavated and hauled to the processing plant. After oil shale extraction, crushed waste rocks, mainly consisting of limestone (Reintam et al. 2002), were dumped at the bottom of excavated areas, covered with previously removed overburden, and bulldozers leveled the spoils (Pensa et al. 2004). Pine plantations were subsequently established by planting 2 3-year-old Scots pine seedlings with initial planting densities of stems per hectare (Kaar 2002; Korjus et al. 2007). Due to drastic land disturbance, post-mining substrates are characterized by locally high spatial variability in physical properties (primarily texture), fertility, and soil water availability. Physical soil conditions restrict tree-rooting depth and water holding in the substrate is low. Additionally, ground water levels are artificially lowered by pumping out the water to avoid flooding in the excavated areas. Furthermore, due to interactions with vegetation, soil-maturing processes produce gradual changes due to rock weathering and organic matter accumulation, resulting in decreased soil reaction (ph) and profile development over the time (Reintam 2001; Reintam2004). Sampling design and stand characteristics We selected 12 circular permanent sample plots, previously established across the reclaimed area of Narva (Viivikonna section) quarry. The plots were located in smallgroupsatthreesites(fig.1) and represent varying soil and stand structure conditions of reclaimed areas. Mean stand age decreased southwards, following the progression of excavation activities; stand age in 2010 varied from 44 to 24 years, with around 10 years difference 123

125 Author's personal copy Environ Sci Pollut Res Fig. 1 Location of the study area and sampling sites (dots) between consecutive sites. If our investigated stands grew on forest sites, they would belong to a single age group, but since continuous substrate formation occurs on postmining areas, we divided trees into three age groups (50 year old, 40-year middle-aged, 30-year young). The oldest trees grew on site 1; the middle-aged trees came from site 2 and the youngest ones were sampled on site 3. Stand dendrometric characteristics were calculated from inventory data in 2010, gathered in circular plots ( ha) (Kiviste and Hordo 2002). Site index at base age of 50 years (SI 50 ) and relative density were calculated according to equations proposed by Nilson (2014), where space for tree growth is calculated based on average tree size in the stand and mean distances between the trees. Since plantations on reclaimed areas are grown for wood production, management interventions occurred on two of the sites. Pre-commercial thinning was conducted in 2003 in the young pine populations (site 3) and first commercial thinning in 2008 in the oldest stands (site 1). Middle-aged stands (site 2) were unmanaged, with high tree numbers in the stands. Other dendrometric measurements for this site (Table 1) indicate higher competition levels (exceeding the self-thinning line; RD 1), as compared to other sites. The oldest stands were the sparsest, with the lowest tree densities and greatest stand basal area. However, quadratic mean diameter in the oldest stands was not significantly different from the youngest ones, where site productivity was higher as indicated by site index (SI 50 ). 124

126 Author's personal copy Environ Sci Pollut Res Table 1 Dendrometric characteristics of the studied Scots pine stands Site Sample plot Age 2010 (years) Tree density (stems ha 1 ) RD SI 50 (m) D q (cm) H(m) G(m 2 ha 1 ) V (m 3 ha 1 ) Site 1 VK VK VK VK VK VK Mean±SD 43± ± ± ± ± ± ± ±37.2 Site 2 VK VK VK Mean±SD 35± ± ± ± ± ± ± ±62.5 Site 3 VK VK VK Mean±SD 24± ± ± ± ± ± ± ±24.4 Site 1 50-year-old stands, site 2 40-year-old stands, site 3 30-year-old stands, Age 2010 mean stand age in 2010, RD relative stand density, SI 50 site index at base age of 50 years, D q mean quadratic diameter, H mean stand height, G mean basal area, V mean standing volume Soil characteristics To characterize the growth substrates, we sampled soils at each permanent sample plot following procedures described in Laarmann et al. (2015). Skeletal fraction content was assessed using the model developed by Laarmann et al. (2011). According to the results of soil analysis, physical and chemical properties differed among the sites (Table 2). The thickness of the organic layer (O-hor) varied from 2.7 to 6.4 cm and was significantly (p<0.05) thicker under the oldest stands as compared to the youngest stands. Depth of the fine soil fraction ranged from 7.1 to 18.0 cm (mean 10.5 cm), but no significant trend were detected along the age gradient. The content of the skeletal fraction did not differ across all sites and averaged 64 %. Clay content increased significantly (p<0.01) from the youngest site (site 3) toward the oldest (site 1), whereas soil alkalinity (ph) slightly decreased with stand age and ranged from 7.4 to 7.7 (p<0.01) across the sites. Amount of potassium (K) increased with stand age (from 41.1 to 84.9 mg kg 1 ; p<0.05); while lower amounts of phosphorus (P) were detected under old stands (mean range among the three sites mg kg 1 ; p<0.05). Amounts of P depended on fine soil content (r=0.84; p<0.001), while amounts of K were related to clay (r=0.87; p<0.001). Soil ph decreased with increasing amount of clay (r= 0.75; p<0.05). Climatic conditions The Estonian Weather Service provided daily weather data from the Jõhvi ( N, E) meteorological station, located 15 km from the study area. Mean annual air temperature for the period of was +4.9 C (Fig. 2). The warmest month at Jõhvi was July (+16.9 C), while the coldest month was February ( 6.5 C).Meanannualrainfallis around 720 mm. Melting snow is the main source of water in spring, since long-term mean monthly rainfall in early spring months (February April) is around 30 mm, increasing in June with a peak in the middle of summer (August, 97 mm month 1 ). The changes in long-term climate conditions were examined via linear regression analysis. Dendrochronological methods We sampled dominant and co-dominant Scots pine trees outside each permanent sample plot at the end of June Only trees without visible stem damage or top leader change were selected. We extracted increment cores from two radii at breast height (1.3 m above the ground). Cores from 18 trees with signs of healed moose damage inside the tree stem were excluded from further analysis because disturbance impulses are undesirable in assessing growthclimate relationships (Cook and Kairiukstis 1990). The remaining increment cores were prepared following standard dendrochronological techniques (Pilcher 1990). Tree-ring widths were measured with the precision of 0.01 mm using a LINTAB tree-ring measuring table equipped with TSAP-Win Scientific Version 0.59 software (Rinn 2003) for graphical and statistical growth pattern matching among trees of the same site. Cross-dating accuracy was examined with the program COFECHA (Holmes 1983) that calculates the correlation between individual ring-width 125

127 Author's personal copy Environ Sci Pollut Res Table 2 Edaphic characteristics of the studied sites Site Sample plot ph O-hor (cm) Skeletal fraction (%) Fine soil depth (cm) Sand (%) Clay (%) Silt (%) C org (%) C (%) N (%) P (mg kg 1 ) K (mg kg 1 ) Site 1 VK VK VK VK VK VK Mean±SD 7.4± ± ± ± ± ± ± ± ± ± ± ±23.9 Site 2 VK VK VK Mean±SD 7.6± ± ± ± ± ± ± ± ± ± ± ±19.2 Site 3 VK VK VK Mean±SD 7.7± ± ± ± ± ± ± ± ± ± ± ±5.4 ph soil reaction, O-hor thickness of organic layer, Skeletal fraction content of soil, with particles larger than 2 cm, Fine soil mean fine soil thickness, Sand content of soil with particle size mm, Clay content of soil with particle size <0.002 mm, Silt content of soil with particle size mm, Corg content of organic carbon, C content of total carbon, N content of nitrogen, P content of phosphorus, K content of potassium 126

128 Author's personal copy Environ Sci Pollut Res Fig. 2 Mean monthly air temperature ( C) and monthly sum of precipitation (mm), at Jõhvi meteorological station over the period of The upper edge of the box indicates upper and lower quartiles; line in the box depicts the median and whiskers show limits of inter-quartile range; points are outliers series and master series (Grissino-Mayer 2001) and identifies mismatching segments. Ring-width series of 21 trees had low correlation with other tree patterns, so they were eliminated. The age-related trend, usually present in ring-width measurement sequences, was removed through the standardization process. We fitted 67 %n smoothing splines with 50 % frequency cut-off on each ring-width measurement series (Cook and Peters 1981; Cook et al. 1990), where n is the series length, and calculated dimensionless tree-ring indices for each tree by dividing the measured radial increment with values of the fitted trend. Detrended ring-width series were prewhitened to remove autocorrelation and averaged into mean plot-wise and site-wise tree-ring-width chronologies using Tukey s biweight robust mean (Mosteller and Tukey 1977). Of the 167 trees we sampled, ring-width series from 128 trees were used for chronology building (Table 3). Three mean site tree-ring-width chronologies covered different periods (Table 3; Fig. 3), but shared a common period of (21 years). The residual chronology for the oldest stands was based on 64 trees and covered the period of The chronology of growth indices for the middle-aged group was compiled from annual growth increments of 31 trees and spanned the period The shortest residual chronology, covering the period of , was built for the 30- year-old trees from site 3. It included data from 33 trees. Several statistics used in dendrochronology studies (Briffa and Jones 1990) were calculated for un-standardized tree-ring series and for each site chronology to assess tree responsiveness to climate and to evaluate the suitability of the chronologies for climate-growth quantification. Mean sensitivity, standard deviation in tree-ring widths, and first-order autocorrelation were calculated for raw ring-width series (Table 3). Mean sensitivity is a measure of mean relative change between adjacent ring widths (Biondi and Qeadan 2008) and standard deviation accounts for variability. First-order autocorrelation quantifies the influence of the previous year growth conditions on the current year s growth. Signal strength for each chronology was estimated by signal-tonoise ratio (SNR) and by inter-correlation of indices series. The confidence of the chronologies was estimated with expressed population signal (EPS). Table 3 Dendrochronological statistics (mean values) of tree-ring measurements and residual chronologies for three sites Site 1 Site 2 Site 3 Ring measurements Mean tree age in 2014 (years) Number of trees/cores 64/128 31/62 33/66 Mean TRW (mm) Mean BAI (cm 2 year 1 ) Standard deviation First-order autocorrelation Mean sensitivity Residual chronologies Time span Chronology length [years] EPS SNR Series inter-correlation TRW tree-ring width, BAI basal area increment, EPS expressed population signal, SNR signal-to-noise ratio 127

129 Author's personal copy Environ Sci Pollut Res Fig. 3 Tree-ring-width residual chronologies for three sites Growth index Site 1 Site 2 Site 3 Year The rate of annual increment, expressed as annual basal area increments (BAIs), were used to model effects of site, competition, and climate on growth rates. To avoid bias when growth rates of different size trees are compared (Biondi 1999), tree size has to be considered; thus we calculated annual BAI from mean tree-wise ring-width measurements, assuming that rings were concentric (Biondi and Qeadan 2008). Annual BAI was calculated from the bark to the pith for each tree, according to the formula: BAI t = π (R 2 2 t R t 1 ), where R is tree radius and t is the year of ring formation. Subsequently, mean arithmetic BAI series were computed for each of the three studied sites. Statistical data analysis Dendroclimatic data analyses were performed using the R software (R Core Team 2014) and the packages dplr (Bunn 2008; Bunn 2010), detrender (Campelo et al. 2012), and bootres (Zang and Biondi 2012). Climate-growth relationships were considered significant at the 95 % confidence level. Growth-climate relationships were established between indices and climate factors and BAI in the nonlinear model. We used bootstrapped correlation analysis (Biondi 1999; Guiot 1991) to study the relationship between climate and radial growth. Residual chronologies were compared against climate variables. Considering that ring formation may take place from May through August (Jyske et al. 2014), the calculations were performed using a time window from the previous year s September to current year s August. Climatic conditions of the previous autumn and winter seasons were considered, since in autumn of the prior year, trees allocate carbohydrates that will be used in the spring to initiate the growth flush. Restricting analysis to the period relevant for ring formation reduces the risk of statistically significant correlations arising by chance, when numerous correlations are calculated. Climate variables used were monthly mean temperature and monthly rainfall sums, seasonal (autumn: September October November; winter: December January February; spring: March April May; summer: June July August) and mean annual climate variables. Mean annual temperature and total precipitation were calculated for the time window (previous September to current August) and vegetation period, defined from April to September. Climate-growth correlations were calculated for the common growth period of (21 years) in order to maintain the same weather conditions. Correlation relationships were further validated via linear regression analysis. To investigate how a particular climate-growth relationship was influenced by stand or soil characteristics, we also calculated climate-growth relationships for the common growth period between 12 plot-wise chronologies and climate variables. Furthermore, the obtained correlation coefficients were used in Pearson s correlation analysis, where relations strength and significance were established between correlation coefficients and stand/soil data available for each studied plot. Modeling response of basal area increment We used nonlinear modeling to investigate the factors affecting basal area growth of individual trees. Nonlinear models are suitable to test the relationships among the variables, when data are unbalanced (number of observations are not equally distributed among the groups) or when the relationship cannot be linearized (Crawley 2007). Annual basal area increment (iba) was used as the dependent variable. According to Burkhart (2003), growth models should be as parsimonious as possible, to avoid over parameterization. At the same time, important predictors for basal area growth should be considered. Thus, three components were considered: basic growth function, effect of thinning, and weather effect. Each component was described with a sub-model and compiled into one general prediction equation of multiplicative structure expressed as follows: iba ¼ f 1 ðtþ f 2 ðthþ f 3 ðavþþε ð1þ where iba is the annual basal area increment (cm 2 /year) of dominant and co-dominant trees, f 1 (t) is a basic growth function showing basal area growth dynamics in 128

130 Author's personal copy Environ Sci Pollut Res unmanaged stands, f 2 (th) is a function describing individual tree response to changes in competition (stand density reduction), and f 3 (av) is a function expressing annual growth variation, depending on weather conditions during the growing period; ε is the error term. Basic basal area increment model f 1 (t) was compiled considering that basal area growth of individual trees follows asymptotic pattern. We chose Weber growth function (cf. Kiviste et al. 2002) to describe basal area growth trend in unmanaged stands: f 1 ðþ t ¼ p tree ð1 expð p site AÞÞ ð2þ where p tree is the parameter indicating potential growth of the tree and also defining the magnitude of asymptote. This parameter was related to current tree diameter at breast height (cm); p site is the parameter indicating site fertility; A is the tree breast height age (years). Response to thinning of individual trees was modeled in a similar manner as proposed by Hynynen (1995). Since we did not have information about stand basal area before the thinning, years since thinning were used to describe changes in stand densities. Studies have shown that thinning effect decreases with increasing time since thinning. We assumed that the thinning effect would last for the same period and response would have an asymptotic shape as suggested by Hynynen (1995). We adjusted the equation to the available variables as follows: th c f 2 ðthþ ¼ 1 þ p th exp thc ð3þ b b where th is the time elapsed since thinning; b and c are the estimates of parameters, b=13.8 and c=1.58, as proposed by Hynynen (1995); and p th is the thinning intensity parameter to be estimated. Annual growth variation f 3 (av) was accounted for by including growth indices. Stepwise inclusion of model components was used to estimate the parameters and to investigate the effects of possible radial growth predictors. Root of the mean squared error (RMSE) and Akaike s Information Criterion (AIC) (Burnham and Anderson 2002) wereusedtoevaluate model improvements. We also calculated the model efficiency or pseudo-r 2, which describes the proportion of variance explained by factors included in the model. Several candidate models were compiled (Table 5) and iteratively compared with the previously compiled model for significant improvement. First, we fitted basic growth function with one parameter p tree (M 1 ) on our data set, and proceeded with inclusion of a second parameter p site (M 2 ). In the following steps, either maximum tree diameter (M 3 )or SI 50 (M 4 ) was related to the asymptote. The model with the lowest AIC and RMSE was used further in the analysis. Since maximum tree diameters (D max ) better described the associated parameter p tree, we used it further in the model and tested variables of site productivity SI 50 (M 5 ) and thickness of fine soil (M 6 ). Thinning effect (M 7 ) was evaluated after defining the best parameters to describe basic BAI growth functions. Finally, radial increment indices (M 8 M 10 ) were added to assess the climate effect on BAI growth. We tested if there were any effects of using more generalized indices to avoid more laborious population-level sampling. For that purpose, first, we included general growth indices into model (M 8 ), then replaced them by site mean chronologies (M 9 ) and finally tested if model fit improved by including plot-level mean chronologies (M 10 ). Statistics of each of the extended models were compared, and the model with the lowest RMSE and AIC and the greatest explained variance (pseudo-r 2 ) was selected as the final model. After comparing observed and predicted mean BAI patterns by sites, we observed that the model was not able to predict a smaller growth peak in growth pattern of site 3. We hypothesized that it could be caused by one earlier management intervention (spacing), which was not recorded. To test that hypothesis, we included a hypothetical thinning in 1996 (M 11 M 14 ). Nonlinear modeling was performed using nonlinear least squares (nls) function in the R environment (R Core Team 2014). Results Changes in long-term climate conditions Linear regression results revealed several temporal trends in long-term mean climate conditions for the period of Total precipitation increased for May (8.7 mm decade 1 ; p<0.05). Mean monthly temperature has steadily increased each decade: in April (0.5 C; p<0.05), July (0.8 C; p<0.001), August (0.5 C; p<0.05), and September (0.4 C; p<0.05). Characteristics of tree-rings and chronologies The widest tree-rings (TRW=3.86 mm year 1 )wereformedin the trees growing on the youngest site and decreased with stand age. Overall, year-to-year variation was not high in studied stands ( ), which is characteristic for complacent growth (Table 3). First-order autocorrelation ranged from 0.68 to 0.77, and was slightly higher in the oldest trees (site 1), suggesting that the weather of the preceding year had a pronounced effect on the diameter growth for these trees. The most homogeneous tree-ring patterns were present in the youngest trees (site 3) where series inter-correlation among indices was high (0.768). Considerably, lower similarities among patterns were found in the middle-aged tree group on site 2 (0.632), and the lowest (0.600) were among the oldest trees (site 1). Signal- 129

131 Author's personal copy Environ Sci Pollut Res to-noise ratio ranged from 16.8 to 78.1 and was more than two times higher in the young trees than in old trees and more than four times higher than in the middle-aged tree growth patterns. Expressed population signal in all chronologies was high, surpassing the recommended confidence threshold of 0.85 (Wigley et al. 1984), suggesting that chronologies are reliable for examining tree growth-climate relationships. Climate-growth relationships Correlation analysis indicated that variation in both precipitation and temperature had a strong relationship to Scots pine radial growth on reclaimed mining areas (Fig. 4). High precipitation in the June July period was significantly positively related to higher radial growth in all studied age groups, while there was opposite relationship between high mean temperatures in summer, especially in August and annual wood production. Importance of precipitation in June July period decreased with decreasing age (p<0.01). The strongest significant relationship between radial growth and rainfall over June July period was detected for the oldest age group (r=0.66), with precipitation in June being more important (r=0.63) than in July (r=0.55). This age group also benefited from higher rainfall in October of the previous year (r=0.45). Weaker correlation coefficients were found between chronology of the oldest trees and monthly temperatures. Higher mean temperatures in spring season (r=0.50) were positively related to tree-ring width, while high mean temperatures in June (r= 0.38) inhibited radial growth of the oldest age group. Similar relationship patterns (as for site 1) between precipitation in June July and radial growth were found for middle-aged group (site 2): r=0.55 (June July), r=0.55 (June), and r=0.50 (July). However, growth of this age group seemed to be controlled more by variation in temperature, as indicated by a greater number of significant correlations between indices and temperature variables. Wood production on this site was related mostly to higher mean temperatures in spring (r=0.69), whereas temperatures in March had a pronounced effect (r=0.49). Warm winters (r=0.38), and especially warmer weather conditions in January (r=0.39), were positively associated to pine radial growth on site 2. Growth indices of middle-aged group were also positively correlated with mean annual temperature (r=0.40). Mean summer temperatures were not significant for wood production on site 2, contrary to the other two age groups. Negative relationship between growth indices and current August temperatures was slightly stronger for the middle-aged site (r= 0.43) than for other two groups (r= 0.35). The least number of significant correlations was found for the youngest age group (site 3). Radial growth relationship to precipitation in June July was significant but moderate (r=0.39), and there was no significant association with monthly rainfall in June or July. The dominant climatic factor controlling Scots pine radial growth in this age group was January temperature (r=0.56). High mean temperatures in summer (r= 0.45) were negatively related to radial growth on site 3, with mean temperatures in July (r= 0.39) and August (r= 0.35) being the most influential. The greatest amount of variance (R 2 =0.28; p<0.05) present in radial increment of the oldest age group was associated Fig. 4 Pearson s correlation coefficients between tree-ringwidth indices of three sites and mean monthly, seasonal, annual, and vegetation period (Veg.p.) temperatures and precipitation; pr. denotes months previous to vegetation period, bars show medians of correlation coefficients, and whiskers indicate 95 % confidence intervals. Statistically significant correlations (p<0.05) are marked with dot Correlation coefficient Correlation coefficient Temperature Site 1 Site 2 Site 3 Precipitation pr.sep pr.oct pr.nov pr.dec Jan Feb Mar Apr May Jun pr.sep pr.oct pr.nov pr.dec Jan Feb Mar Apr May Jun Jul Aug Jul Aug Fall Winter Spring Summer Jun+Jul Veg.p. Annual Fall Winter Spring Summer Jun+Jul Veg.p. Annual 130

132 Author's personal copy Environ Sci Pollut Res Fig. 5 Observed and predicted basal increment growth patterns of pine trees on three sites: a prediction without hypothetical thinning; b prediction with included hypothetical thinning. Arrows indicate time of with the variability in June July precipitation, while fluctuations in mean spring temperature were responsible for 47 % (p<0.001) in annual ring-width variation at site 2. Temperatures in January accounted for 29 % (p<0.05) of variability in growth patterns from site 3. Growth patterns Mean annual basal area increment was greatest in the youngest age group trees (site 3, BAI=9.60 cm 2 year 1 ) and the smallest in unmanaged stands (site 2, BAI=6.01 cm 2 year 1 ). Sudden increases in BAI followed the pre-commercial thinning at site 3 in 2004, and first commercial thinning in the oldest stands (site 1) in 2008 demonstrated the effect of lowered competition (Fig. 5). After the thinning, mean annual BAI of dominant trees increased 58 and 3 % in old and young trees, respectively, considering 5-year periods before and after the intervention. Standardized growth patterns were quite synchronous among all chronologies for the period of , but diverged in the youngest age group for the period of , just after the pre-commercial thinning (Fig. 5). A sharp decrease in radial growth in the youngest age group (site 3) is evident for , while growth depression in the same period for the two other sites was considerably lower. Two-year growth depression in is present in the growth patterns of sites 1 and 2, but not for site 3. Very low growth indices for site 2 were observed in 1984 and for site 1 in Correlation analysis suggested that the patterns of indices differed slightly. The strongest correlation (r=0.74; p<0.001) for common pair-wise length was between chronologies of sites 1 and 2. Coefficients of the other two pairwise correlations between site 1 vs. site 3 and site 2 vs. site 3 were not statistically significant. management activity, PT pre-commercial thinning, FT first thinning, and HT hypothetical thinning Effects of soil and stand characteristics on tree response to climate Tree responsiveness to monthly temperature regimes to a large extent was influenced by stand characteristics, while sensitivity to precipitation seemed to increase with growing tree dimensions but also depended on soil characteristics (Table 4). Moderately strong relationships (r= ) between sensitivity to spring temperatures (particularly in March) and stand density (expressed in stand volume, relative density, and basal area) were established. Thicker layers of O-hor tended to mitigate negative influence of temperatures in August, as indicated by significant negative correlation. Growth sensitivity to mean temperatures in January was related to stand age and increased with site productivity (SI 50 ). The decrease in ph inversely influenced radial growth response to precipitation in June. Thinning, site and climate effects on radial growth An individual tree basal area increment model (Eq. 1) for dominant and co-dominant Scots pine trees on reclaimed oil shale areas was compiled and fitted on the data set, consisting of more than 3600 annual basal increment records. The fit statistics from nonlinear procedure used to estimate annual growth rate with alternative models (M 1 M 14 ) are presented in Table 5. The compiled growth model was logically consistent. Predicted tree basal area increment patterns followed the observed ones (Fig. 5a, b). Model prediction accuracy improved each time, after new variable was included, as indicated by lower RMSE and decreasing AIC (Table 5). At the same time, model prediction ability remained 131

133 Author's personal copy Environ Sci Pollut Res Table 4 Correlation coefficients between growth-climate relationships and soil/stand characteristics. Statistically significant (p<0.05) correlations are presented in italic Growth-climate relationship Stand characteristics Soil characteristics Age H D q SI 50 G RD V ph O-hor Fine soil Clay C C org N K T T T Tspring Tsummer Tannual P P P T temperature, P precipitation, numbers 1 10 denote calendar months, 1 January, 2 February etc.; Age mean stand age (years), H mean stand height (m), D q mean quadratic diameter (cm), SI 50 site index at base age of 50 years (m), G mean stand basal area (m 2 ha 1 ), RD relative density related to average tree size and distances between trees, V mean standing volume (m 3 ha 1 ), O-hor thickness of organic layer (cm), ph soil reaction, Fine soil mean fine soil thickness (cm), Clay content of clay (%), C content of carbon (%), C org content of organic carbon (%), N content of nitrogen (%), P content of phosphorus (mg kg 1 ), K content of potassium (mg kg 1 ) Table 5 Parameter estimates and fit statistics from the nonlinear regressions of annual basal area increment Model Components Parameters RMSE AIC pseudo-r 2 p tree p site p th Estimate SE Estimate SE Estimate SE M 1 f 1 (A),p tree M 2 f 1 (A),p tree, p site M 3 f 1 (A),p tree -D max, p site M 4 f 1 (A),p tree -SI 50, p site M 5 f 1 (A),p tree -D max, p site -SI M 6 f 1 (A),p tree -D max, p site -fine soil M 7 f 1 (A),p tree -D max, p site - fine soil * f 2 (th) M 8 f 1 (A),p tree -D max, p site -fine soil* f 2 (th)*f 3 (av) general chron M 9 f 1 (A),p tree -D max, p site -fine soil* f 2 (th)*f 3 (av) site chron M 10 f 1 (A),p tree -D max, p site -fine soil* f 2 (th)*f 3 (av)-plot chron M 11 f 1 (A),p tree -D max, p site -fine soil* f 2 (hth) M 12 f 1 (A),p tree -D max, p site -fine soil* f 2 (hth)*f 3 (av) general chron M 13 f 1 (A),p tree -D max, p site -fine soil* f 2 (hth)*f 3 (av) site chron M 14 f 1 (A),p tree -D max, p site -fine soil* f 2 (hth)*f 3 (av)- plot chron Parameters: p tree tree size parameter, depending on maximum tree diameter (diameter when trees were sampled); p th response to thinning parameter, described as years since thinning; p site site productivity parameter described by either SI 50 or fine soil thickness (cm); f 1 (A) basic grow function, f 2 (th) response to thinning function, f 3 (av) function describing annual variation related to climate; f 2 (hth) response to thinning function with included hypothetical thinning. The best models are marked in italic: M 10 without hypothetical thinning and M 14 with hypothetical thinning considered; RMSE root mean square error, AIC Akaike s information criterion, pseudo-r 2 shows total explained variance 132

134 Author's personal copy Environ Sci Pollut Res considerably stable, with low variation in parameter standard errors. All the parameter estimates were logical and highly significant (p<0.001). The basic models (M 3 M 6 ), where site fertility and tree size were used as predictor variables, explained % of the total variation in the observed basal area increments. Fine soil thickness was a better predictor than SI 50. Inclusion of response function to thinning (M 7 ) increased pseudo-r 2 up to 55 %, with a reduction in RMSE and AIC. After including climate effects (M 8 M 10 ), the fit statistics (RMSE, AIC) showed improvement, and maximum explained common variance increased up to 59 %. The greatest effect on radial growth, according to our model, had site properties and thinning, while climatic variation was significant, but it improved model predictions only slightly. Mean plot chronologies were slightly better predictors of basal area increment than mean site or mean general chronology, proving that response to climatic factor among the sites are not the same. After inclusion of a hypothetical thinning into the data set, model predictions improved as indicated by lower RMSE and AIC. Percentage of maximum explained common variance (pseudo-r 2 ) increased up to 63 % (M 14 ). No evidence of dependencies was observed in bi-plots of residuals against predicted values (data not presented). Discussion Climate effects on radial growth Dendrochronological studies in Baltic countries report that Scots pine, growing on dry and infertile sites, is sensitive to cold winters (Elferts 2007; Vitas 2004; Vitas and Erlickyte 2007) and benefits from higher temperatures in winter and early spring (Hordo et al. 2009; Läänelaid and Eckstein 2003). Erlickyte and Vitas (2008) found in Lithuania that due to changes in long-term climate trends, the influence of winter cold on Scots pine annual diameter growth decreased, while sensitivity to rainfall in springtime and summer became more important. Our results show that Scots pine radial growth on the reclaimed areas is limited by water deficit in summer, apparently as a result of low precipitation and high temperatures in June to August. The main water source for tree growth is precipitation and extended dry and warm periods may cause water deficiency in the area (Vaus 1970). Läänelaid and Eckstein (2003) reported that rainfall in June is the main climatic factor driving radial growth in Norway spruce in Estonia. Spruce sensitivity to precipitation is usually explained by close-to-surface root system, which could be the same reason for pine trees growing on reclaimed areas where rooting depth is confined by the rocky substrate. Root penetration into deeper layers is restricted by physical soil conditions and water holding in the substrate is low. Additionally, ground water levels are artificially lowered by pumping out the water to avoid flooding in the excavated areas. Because of this drainage, the water table can be too deep to supply the moisture for roots by capillary action. Besides that, microbial activity is also dependent on soil water availability. Microbial activity decreases with lower soil water availability, and microbial metabolism may become totally inhibited during extended drought events (Borken and Matzner 2009), therefore suspending uptake of nutrients. In Estonia, severe meteorological summer droughts, when precipitation was absent or was low for longer than 20-day period, were observed in 2002 and 2006, with exceptionally dry summers in 1983, 1992, and 1999 (Tarand et al. 2013). These dates coincide with years of extremely low radial increment in our studied plots. In general, Scots pine is tolerant to drought, and its sensitivity to precipitation and water deficit is more characteristic of dry regions (Allen and Breshears 1998). However, Scots pine trees become more sensitive to climate dryness, in the areas where water deficit becomes more frequent or lasts for an extended time, eventually leading to tree mortality (Hereş et al. 2012). Temperature governs physiological processes, including photosynthesis and carbon sequestration, in the tree and determines the rate at which water evaporates from the soil (Fritts 1976). Formation of a permanent forest floor enhances moisture retention, and may also influence tree response to climate (Drobyshev et al. 2010). Our results suggest that thicker layers of the organic horizon, which develops on reclaimed sites as vegetation develops, reduced the negative influence of temperature in August. Pine sensitivity to precipitation in October could be explained as due to formation of new roots (Ostonen et al. 2006) during this time. The presence of higher amounts of carbon and organic carbon increased tree sensitivity to precipitation in October (Table 4). With a better-developed root system, water and nutrient uptake could have been enhanced, which might be advantageous during the ring-formation time. Mean temperatures in July and August are steadily increasing, suggesting that evaporative demand will be increasing and growth stress induced by high temperatures is likely to increase in the future. This could be especially significant for older trees, as we found a positive relationship between precipitation in June and July and increasing tree size. Tree density reduction through thinning has been suggested as a way to mitigate water stress effects (Giuggiola et al. 2013). Basal area increment patterns did not show growth decline, but rather was flat for dense stands and increased sharply as a response to reduced competition. We also found that trees of different age, growing in close vicinity and under the same climatic conditions, responded to 133

135 Author's personal copy Environ Sci Pollut Res climate differently. Radial growth patterns were similar within the sites (Table 3), but less coherent between the sites, confirming spatial variability of growth response to environmental factors. Variation in tree responsiveness to climatic variables has been previously observed between the tree species on the same site and in the trees of a species between the sites (Fritts 1976). Furthermore, correlation coefficients, established during our study between growth indices and climatic variables, were higher than the ones reported for Scots pine, growing on forest sites (e.g., Läänelaid and Eckstein 2003; Hordo et al. 2009, 2011) in the region. Apparently, Scots pine growing on reclaimed areas is more sensitive to weather variability than pine trees growing on forest sites. Greater growth rates were detected on the youngest site as compared to other two groups. Similar results were observed for pine growing on fertile forest sites in Estonia, where younger tree cohorts had enhanced growth as compared to the older ones (Metslaid et al. 2011). For pine trees growing on the mined areas, the greater BAI growth could be a result of more fertile soil conditions, as defined by thicker fine soil layer and more intensive management. The youngest Scots pine generation was more sensitive to temperature, suggesting that their water needs were satisfied, possibly due to greater amount of fine soil present in the substrate. The fine soil had a higher water holding capacity, which means that moisture could have been stored for longer periods. Growth sensitivity to mean temperatures in spring was related to stand density. Soil under forest cover tends to be cooler in the summer and warmer in the winter, than in the clear-felling areas (Aussenac 2000). Already during the sampling occasion, we noticed that ground vegetation under theses stands was sparse, indicating lack of sunlight in these stands. It was darker, because of full canopy shading and the temperature felt also cooler, than at the other two sites. Due to delayed snowmelt in spring, onset of the growing season in denser stands could have started later, having considerable influence on tree annual productivity (Vaganov et al. 1999; Helama et al. 2013). The greater canopy cover in the middle-aged stands could explain the positive correlation to annual temperature (longer time span) and lower sensitivity to temperatures in summer months as compared to managed stands. However, relation to temperatures in August for middle-age group was stronger as compared to the other two groups, suggesting that there could be delay in ring formation processes due to later onset in spring. Ring formation can be continuing in these trees, while growth processes are about to cease in more sparse stands. Modeling basal area increment The basal area increment model developed during this study included three main components, which described site properties and accounted for climatic variation and tree response to changes in competition. These variables are commonly used in basal area models (Wykoff 1990; Keyser and Brown 2014; Mehtätalo et al. 2014). By applying a nonlinear modeling approach, we avoided many assumptions that are necessary for using linear modeling (Zuur et al. 2009; Eastaugh et al. 2013). As a result, our model was able to predict basal area increment in managed and unmanaged Scots pine plantations growing on reclaimed areas. For simplicity, and because we were not interested in tree-level effects, we used tree maximum diameter to describe potential tree growth (asymptote) on particular sites. According to our model, site properties and thinning had the greatest effect on the radial growth (Table 5). Our results (Fig. 5) show that thinning on these sites reduced competition between trees resulting in a positive effect on radial growth, with trees responding immediately after treatment. The first thinning increased the basal area increment of dominant trees up to 50 %. This corresponds to the results of Mäkinen and Hynynen (2014) who found that intensive thinning on mineral soils can increase the basal area increment of remaining trees 50 %. Thinning on reclaimed sites has positive effect on radial growth of dominant trees reducing the competition between trees after canopy closure. Individual tree response to changes in stand density was modeled as time elapsed since thinning, because stand basal area before and after thinning was not available. In practice, changes in basal area are rarely recorded, and inclusion of a variable describing time since thinning is more convenient. It was a surprising finding that with a model assistance, we were able to detect one more thinning (on site 3), presumably a cleaning (Fig. 5b), since it was carried out at an early stand development stage. We refer to this management activity as hypothetical, because management activities are not well recorded for these areas, and we could not detect any proof for this activity. However, model prediction was improved significantly after including a thinning in 1996 for that site. Thus, our finding suggests that trees on reclaimed areas are highly responsive even to minor changes in stand structure. It should be noted that only dominant and co-dominant trees were investigated in this study, since these trees are expected to be crop trees, growing for many years on that site. Trees with other social status (intermediate or suppressed) might have been responding differently to increased space, since radial growth is size-mediated (e.g., Gavin et al. 2008); therefore, all the findings in this study apply only for the largest trees in the stand (trees with crowns extending above general level and from the general level). To evaluate the effect of site productivity on radial growth, we tested traditionally used site index. Fine soil depth was included because it showed high correlation to site index (SI 50 ) and because previously similar variables have been used (e.g., Schröder et al. 2002) as a direct indicator of site productivity. For our data set, fine soil depth was a better predictor than site index. Besides, it is easy to measure in 134

136 Author's personal copy Environ Sci Pollut Res the field. Site productivity is an important variable in basal area growth modeling, describing differences in trees growing under the same climate conditions, but on different soils (Subedi and Sharma 2013). Annual basal area increment data, expanding over the period of (37 years), was used to fit the model. Therefore, growth predictions may be used to predict long-term growth on reclaimed areas and predictions are climate sensitive. Mitigating effects of climate change Dry and warmer than average climatic conditions in summer, especially in June and July, were negatively related to diameter growth of the Scots pine trees on reclaimed oil shale mined areas in Northeast Estonia. This relationship grew stronger with increasing tree dimensions, i.e., as stands developed. Water stress, caused by lack of rainfall in summer, when active ring formation takes place, was expressed among the trees unevenly and was associated to physical characteristics of the substrate (e.g., soil texture and the amount of fine soil fraction, and the depth of organic matter accumulation) and stand density that influenced microclimate in the understory due to shading. Thicker organic matter reduced the negative effects of temperature especially in August. In general, organic matter accumulation on reclaimed mined sites is correlated with litter production, meaning that quickly attaining full site occupancy and high productivity are necessary for rapid accumulation (Mandre and Kuznetsova 2004). Positive relationship between growth and mean spring temperatures indicated that greater annual wood production could be expected during early and warm springs. Trees growing in dense stands, however, are subject to competition-induced mortality (Laarmann et al. 2009) and older managed stands are sensitive to moisture stress. Thinning in young stands may retard soil formation and organic layer accumulation as well as stability on mined sites and first thinning often is delayed. The relationships among climate variables and soil and stand characteristics established in this study can be used to guide management of reclaimed oil shale mined sites under altered future climate. As suggested by Giuggiola et al. (2013) and D Amato et al. (2013), managing tree competition in closed-canopy stands by thinning can be an adaptation to climate change-induced drought stress. Thinning young and old stands on reclaimed sites positively influenced growth rates, with immediate response of the dominant and co-dominant trees. Temperatures in January controlled radial growth in the youngest stands. Responsiveness to lower temperatures in January grew weaker with age, although higher productivity sites (expressed as SI 50 ) retained sensitivity to temperature. As climate changes toward warmer winter temperatures, increased growth rates on reclaimed mining sites may accelerate stand development and necessitate earlier thinning to avoid competition-induced mortality and moisture stress. Acknowledgments The study was conducted at the Estonian University of Life Sciences. The study was supported by the Institutional Research Funding IUT21-4 of the Estonian Ministry of Education and Research, and SOO project ( ) and the Environmental Investment Center supported data collection from the network of Estonian research plots. Estonian Weather Service at Estonian Environmental Agency provided climate data. We would also like to thank Kalmer Sokman for provided information and reviewers for the valuable comments, which helped to improve manuscript considerably. References Allen CD, Breshears D (1998) Drought-induced shift of a forestwoodland ecotone: rapid landscape response to climate variation. Proc Natl Acad Sci U S A 95(25): Aussenac G (2000) Interactions between forest stands and microclimate: ecophysiological aspects and consequences for silviculture. Ann For Sci 57: Baumann K, Rumpelt A, Schneider BU, Marschner P, Hüttl RF (2006) Seedling biomass and element content of Pinus sylvestris and Pinus nigra grown in sandy substrates with lignite. Geoderma 136: Biondi F (1999) Comparing tree-ring chronologies and repeated timber inventories as forest monitoring tools. Ecol Appl 9(1): Biondi F, Qeadan F (2008) A theory-driven approach to tree-ring standardization: defining the biological trend from expected basal area increment. Tree-Ring Res 64(2):81 96 Borken W, Matzner E (2009) Reappraisal of drying and wetting effects on C and N mineralization and fluxes in soils. Glob Chang Biol 15(4): Briffa KR, Jones PD (1990) Basic chronology statistics and assessment. In: Cook ER, Kairiūkštis LA (eds) Methods of dendrochronology: applications in the environmental sciences. Kluwer, Dordrecht, pp Bunn AG (2008) A dendrochronology program library in R (dplr). Dendrochronologia 26: Bunn AG (2010) Statistical and visual crossdating in R using the dplr library. Dendrochronologia 28(4): Burkhart HE (2003) Suggestions for choosing an appropriate level for modelling forest stands. In: Amaro A, Reed D, Soares P (eds) Modelling Forest Systems. Papers from a workshop held in Sesimbra, Portugal, 2-5 June CABI Publishing, UK, pp 3 10 Burnham KP, Anderson DR (2002) Model selection and multi-model inference: a practical information-theoretic approach, 488 p Campelo F, Garcia-Gonzalez I, Nabais C (2012) detrender - A Graphical User Interface to process and visualize tree-ring data using R. Dendrochronologia 30(1):57 60 Cook ER, Kairiūkštis LA (1990) Methods of Dendrochronology: Applications in the Environmental Sciences. Kluwer Academic Publishers, Dordrecht Cook ER, Peters K (1981) The smoothing spline: a new approach to standardizing forest interior tree-ring width series for dendroclimatic studies. Tree-Ring Bull 41:45 53 Cook ER, Briffa K, Shiyatov S, Mazepa V (1990) Tree-ring standardization and growth-trend estimation. In: Cook ER, Kairiukstis LA (eds) Methods of Dendrochronology: Applications in the Environmental Sciences. Kluwer Academic Publishers, Dordrecht, pp D Amato AW, Bradford JB, Fraver S, Palik BJ (2013) Effects of thinning on drought vulnerability and climate response in north temperate forest ecosystems. Ecol Appl 23(8): Drobyshev I, Simard M, Bergeron Y, Hofgaard A (2010) Does soil organic layer thickness affect climate- growth relationships in the Black Spruce Boreal Ecosystems? Ecosystems 13:

137 Author's personal copy Environ Sci Pollut Res Eastaugh CS, Kangur A, Korjus H, Kiviste A, Zlatanov T, Velichkov I, Srdjevic B, Srdjevic Z, Hasenauer H (2013) Scaling issues and constraints in modelling of forest ecosystems: a review with special focus on user needs. Balt For 19(2): Elferts D (2007) Scots pine pointer-years in north western Latvia and their relationship with climatic factors. Biol Acta Universitatis Latviensis 723: Erlickyte R, Vitas A (2008) Influence of climatic and anthropogenic factors on the radial growth of Scots pine (Pinus sylvestris L.). Balt For 14(2): Fritts HC (1976) Tree Rings and Climate. Academic, New York Gavin DG, Beckage B, Osborne B (2008) Forest dynamics and the growth decline of red spruce and sugar maple on Bolton Mountain, Vermont: a comparison of modeling methods. Can J For Res 38(10): Giuggiola A, Bugmann H, Zingg A, Dobbertin M, Rigling A (2013) Reduction of stand density increases drought resistance in xeric Scots pine forests. For Ecol Manag 310: Grissino-Mayer HD (2001) Evaluating crossdating accuracy: a manual and tutorial for the computer program COFECHA. Tree-Ring Res 57: Guiot J (1991) The bootstrapped response function. Tree-Ring Bull 51: Helama S, Lindholm M, Meriläinen J, Timonen M, Eronen M (2005) Multicentennial ring-width chronologies of Scots pine along northsouth gradient across Finland. Tree-Ring Res 61:21 32 Helama S, Arentoft BW, Collin-Haubensak O, Hyslop MD, Brandstrup CK, Mäkelä HM, Tian QH, Wilson R (2013) Dendroclimatic signals deduced from riparian versus upland forest interior pines in North Karelia, Finland. Ecol Res 28(6): Henttonen HM, Mäkinen H, Peltoniemi M, Laurén A, Hordo M (2014) Response of radial increment variation of Scots pine to temperature, precipitation and soil water content along a latitudinal gradient across Finland and Estonia. Agric For Meteorol : Hereş AM, Martínez-Vilalta J, López BC (2012) Growth patterns in relation to drought-induced mortality at two Scots pine (Pinus Sylvestris L.) sites in NE Iberian Peninsula. Trees 26(2): Hökkä H, Salminen H, Ahti E (2012) Effect of temperature and precipitation on the annual diameter growth of Scots pine on drained peatlands and adjacent mineral soil sites in Finland. Dendrochronologia 30(2): Holmes RL (1983) Computer-assisted quality control in tree-ring dating and measurement. Research report. Tree-Ring Bull 43:69 78 Hordo M, Metslaid S, Kiviste A (2009) Response of Scots pine (Pinus sylvestris L.) radial growth to climate factors in Estonia. Balt For 15(2): Hordo M, Henttonen HM, Mäkinen H, Helama S, Kiviste A (2011) Annual growth variation of Scots pine in Estonia and Finland. Balt For 17(1):35 49 Hüttl RF, Weber E (2001) Forest ecosystem development in post-mining landscapes: a case study of the Lusatian lignite district. Naturwissenschaften 88: Hynynen J (1995) Predicting the growth response to thinning for Scots Pine stands using individual-tree growth models. Silva Fennica 29(2): Jyske T, Mäkinen H, Kalliokoski T, Nöjd P (2014) Intra-annual tracheid production of Norway spruce and Scots pine across a latitudinal gradient in Finland. Agric For Meteorol 194: Kaar E (2002) Coniferous trees on exhausted oil shale opencast mines. Forestry Studies 36: Karu H, Szava-Kovats R, Pensa M, Kull O (2009) Carbon sequestration in a chronosequence of Scots pine stands in a reclaimed opencast oil shale mine. Can J For Res 39(8): Keyser TL, Brown PM (2014) Long-term response of yellow-poplar to thinning in the southern Appalachian Mountains. For Ecol Manag 312: Kiviste A, Hordo M (2002) Network of permanent forest growth plots in Estonia. Forestry Studies 37:43 58 (in Estonian) Kiviste A, Alvarez Gonzales JG, Rojo Alboreca A, Ruiz Gonzales AD (2002) Funciones de crecimiento de aplocacion en el ambito forestal. Instituto Nacional de Investigacion y Tecnologia Agraria y Alimentaria Ministerio de Ciencia y Tecno, Madrid Kiviste A, Kaar E, Sims A (2010) Forest growth dynamics in abandoned oil shale quarries. In: Kaar E, Kiviste K (eds) Mining and rehabilitation in Estonia. Estonian University of Life Sciences, Tartu, pp (in Estonian) Korjus H, Sims A, Kangur A, Kaar E, Kiviste A (2007) Forest growth dynamics on abandoned oil shale quarries on the basis of permanent plot data. Miškininkystė 61(1):24 29 Kuznetsova T, Tilk M, Pärn H, Lukjanova A, Mandre M (2011) Growth, aboveground biomass and nutrient concentration of young Scots pine and lodge pole pine in oil shale post-mining landscapes in Estonia. Environ Monit Assess 183: Läänelaid A, Eckstein D (2003) Development of a tree-ring chronology of Scots pine (Pinus sylvestris L.) for Estonia as a dating tool and climate proxy. Balt For 9(2):76 82 Laarmann D, Korjus H, Sims A, Stanturf JA, Kiviste A, Köster K (2009) Analysis of forest naturalness and tree mortality patterns in Estonia. For Ecol Manag 258:S187 S195 Laarmann D, Sims A, Korjus H, Kiviste A, Kangur A (2011) Forest ecosystem restoration microhabitat pattern analysis on abandoned oil-shale mining areas in Estonia. SER Europe Knowledge Base on Ecological Restoration in Europe ed5-49b8-9b1a-28d3307ebc8b/EA_SER2010_168.pdf Accessed 01 April 2015 Laarmann D, Korjus H, Sims A, Kangur A, Kiviste A, Stanturf JA (2015) Evaluation of afforestation development and natural colonization on a reclaimed mine site. Restor Ecol 23(3): Linderholm H (2001) Climatic influence on Scots pine growth on dry and wet soils in the central Scandinavian mountains, interpreted from tree-ring widths. Silva Fennica 35: Mäkinen H, Hynynen J (2014) Wood density and tracheid properties of Scots pine: responses to the timing of first commercial thinning and fertilisation. Forestry 87: Mandre M, Kuznetsova T (2004) Conditions for mineral nutrition of conifers on reclaimed oil shale mine sites. Forestry Studies 41:17 26 Mehtätalo L, Peltola H, Kilpela A, Ikonen V (2014) The response of basal area growth of Scots pine to thinning: a longitudinal analysis of treespecific series using a nonlinear mixed-effects model. Ann For Sci 60(4): Metslaid S, Sims A, Kangur A, Hordo M, Jõgiste K, Kiviste A, Hari P (2011) Growth patterns from different forest generations of Scots pine in Estonia. J For Res 16(3): Mosteller F, Tukey JW (1977) Data Analysis and Regression. Addison- Wesley, Reading Nilson A (2014) On growth and structure of pine stands in Estonia. In: Kurm M (ed) Pine in Estonia. Estonian University of Life Sciences, Tartu, pp (in Estonian) Ostonen I, Lõhmus K, Alama S, Truu J, Kaar E, Vares A, Uri V, Kurvits V (2006) Morphological adaptations of fine roots in Scots pine (Pinus sylvestris L.), silver birch (Betula pendula Roth.) and black alder (Alnus glutinosa (L.) Gaertn.) stands in recultivated areas of oil shale mining and semicoke hills. Oil Shale 23(2): Pensa M, Sellin A, Luud A, Valgma I (2004) An analysis of vegetation restoration on opencast oil shale mines in Estonia. Restor Ecol 12(2): Pietrzykowski M, Socha J (2011) An estimation of Scots pine (Pinus sylvestris L.) ecosystem productivity on reclaimed post-mining sites in Poland (central Europe) using of allometric equations. Ecol Eng 37(2):

138 Author's personal copy Environ Sci Pollut Res Pilcher JR (1990) Sample preparation, cross-dating and measurement. In: Cook ER, Kairiukstis LA (eds) Methods of Dendrochronology: applications in the Environmental Sciences. Kluwer Academic Publishers, pp R Core Team (2014) R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Accessed 01 April 2015 Raukas A, Punning J-M (2009) Environmental problems in the Estonian oil shale industry. Energy Environ Sci 2: Reintam L (2001) Changes in the texture and exchange properties of skeletal quarry detritus under forest during thirty years. Proceedings of the Estonian Academy of Sciences. Biology. Ecology 48: Reintam L (2004) Rehabilitated quarry detritus as parent material for current pedogenesis. Oil Shale 21(3): Reintam L, Kaar E (2002) Natural and man-made afforestation of sandytextured quarry detritus of opencast oil-shale mining. Balt For 8(1): Reintam L, Kaar E, Rooma I (2002) Development of soil organic matter under pine on quarry detritus of open-cast oil-shale mining. For Ecol Manag 171: Rinn F (2003) TSAP-Win professional, Time series analysis and presentation for dendrochronology and related applications. Version 0.3. Heidelberg, Germany Roberts JA, Daniels WL, Bell JC, Burger JA (1988) Early stages of mine soil genesis in Southwest Virginia spoil lithosequence. Soil Sci Soc Am J 52: Schröder J, Rodríguez Soalleiro R, Vega Alonso G (2002) An ageindependent basal area increment model for maritime pine trees in northwestern Spain. For Ecol Manag 157:55 64 Sepp K, Metsaots K, Roose A (2010) Appreciating and redeveloping landscapes changed in the course of mining operations. In: Kaar E, Kiviste K (eds) Mining and rehabilitation in Estonia. Estonian University of Life Sciences, Tartu, pp (in Estonian) Subedi N, Sharma M (2013) Climate-diameter growth relationships of black spruce and jack pine trees in boreal Ontario, Canada. Glob Chang Biol 19: Tarand A, Jaagus J, Kallis A (2013) Climate in Estonia in the past and nowadays. Tartu University (in Estonian) Torbet JL, Burger JA, Daniels WL (1990) Pine growth variation associated with overburden rock type on reclaimed surface mine in Virginia. J Environ Qual 19:88 92 Vaganov EA, Hughes MK, Kirdyanov AV, Schweinguber FH, Silkin PP (1999) Influence of snowfall and melt timing on tree growth in subarctic Eurasia. Nature 400: Vaus M (1970) Silvicultural properties of the quarry detritus of opencast oil-shale mining in Estonia. Valgus, Tallinn (in Estonian) Vitas A (2004) Dendroclimatological research of Scots pine (Pinus sylvestris L.) in the Baltic coastal zone of Lithuania. Balt For 10(1):65 71 Vitas A, Erlickyte R (2007) Influence of droughts to the radial growth of Scots pine (Pinus sylvestris L.) in different site conditions. Balt For 13(1):10 16 Weber P, Bugmann H, Rigling A (2007) Radial growth responses to drought of Pinus sylvestris and Quercus pubescens in an inner- Alpine dry valley. J Veg Sci 18(6): Wigley TML, Briffa KR, Jones PD (1984) On the average value of correlated time series, with applications in dendroclimatology and hydrometeorology. J Appl Meteorol 23: Wykoff RW (1990) A basal area increment model for individual conifers in the northern Rocky Mountains. For Sci 36: Yearbook Forest 2013 (2014) Estonian Environmental Information Centre, Tartu (in Estonian) Zang C, Biondi F (2012) Dendroclimatic calibration in R: the bootres package for response and correlation function analysis. Dendrochronologia 31(1):68 74 Zuur AF, Leno EN, Walker N, Saveliev AA, Smith GM (2009) Mixed Effects Models and Extension in Ecology with R, Statistics for Biology and Health. Springer Science, LLC 137

139 138 IV

140 Metslaid, S., Hordo, M., Korjus, H., Kiviste, A., Kangur, A. 20xx Spatio-temporal variability in Scots pine response to annual climate fluctuations in hemiboreal forests of Estonia. Submitted to Agricultural and Forest Meteorology 139

141 Spatio-temporal variability in Scots pine response to annual climate fluctuations in hemiboreal forests of Estonia Sandra Metslaid 1, *, Maris Hordo 1, Henn Korjus 1, Andres Kiviste 1, Ahto Kangur 1 1 Department of Forest Management, Institute of Forestry and Rural Engineering, Estonian University of Life Sciences, Kreutzwaldi 5, Tartu, Estonia *corresponding author Abstract In this study we used a comprehensive tree-ring network from Estonia and investigated Scots pine (Pinus sylvestris L.) radial growth responses to changing climate conditions, considering differences in site conditions and local climates. To assess whether climate influences Scots pine radial growth consistently across the country, we developed thirteen tree-ring width chronologies for pine populations growing in four forest site types Cladonia, Calluna, Myrtillus, and Rhodococcum in four sub-regions of Estonia and compared these to climate data. A correlation analysis between ring-width indexes and monthly resolved temperature and precipitation applied over the period of revealed that in most of the cases, radial growth was related to several climatic variables, however, significant positive correlations were found with winter/early spring temperatures and the total precipitation of late summer in the year prior to growth. High mean temperatures in August, the year prior to growth were negatively related to pine growth, particularly on islands and in the Northeast of Estonia. Scots pine growth on mesic and medium fertile Myrtillus and Rhodococcum sites in the Southeast exhibited greater sensitivity to mean February April temperatures, while high temperatures and low precipitation at the end of the summer of the previous growing season limited radial growth of pine on the islands and in the Northeastern sub-region. A hierarchical cluster analysis performed on mean index chronologies and a principal component analysis conducted on bootstrapped correlation coefficients revealed that local climate is the main driver of common growth, followed by ecological site conditions. A moving correlation analysis, performed over the period of , using 30-year windows shifted by one year showed that climate-growth relationships are not stable in changing climatic conditions. Associations between Scots pine tree-ring width and winter temperatures are getting weaker for some of the forest site types, whereas relationships with summertime water availability are gaining significance. Keywords: dendrochronology; climate-growth relationships; PCA; ecological site conditions; moving correlation; summer water deficit. 1. Introduction Since the middle of the last century, the growth and productivity of hemiboreal forests in Northeastern Europe have been subjects to rapid changes in climatic conditions, arising as a result of global climate warming (Serreze et al., 2000; HELCOM, 2013; Rutgersson et al., 2015). According to Giorgi (2006), Northeastern Europe is among the top hot-spot regions in Europe 140 1

142 where the greatest shifts in thermal and precipitation regimes are expected by the end of this century. Considering that trees are long-living organisms, concerns arise regarding the possible impact of climate change on the functioning and productivity of forest ecosystems in the region. A productivity increase is projected for boreal forests (Karjalainen and Kellomäki, 1995), while the impact of climate warming in a hemiboreal vegetation zone, which is a transition zone between boreal and temperate zones, is suggested to change and depend on site conditions and change itself (Lindner et al., 2010). In the Baltic Sea region, increases in mean annual air temperature have been observed since the beginning of the meteorological measurements in 1871, with a temperature rise of 0.08 C per decade, which is greater than at the global scale (BACC Author Team, 2008). Due to a strong influence of large-scale circulations, particularly the North Atlantic Oscillation (NAO) during wintertime (Niedzẃiedz et al., 2015), which is characterized by interchanging cycles of cool (severe) and warm (mild) phases that last for several decades (Hagen and Feistel, 2005), the warming process has been fragmented, and had several cooling and warming periods (Rutgersson et al., 2015). For several recent decades all areas around the Baltic Sea experienced a distinct warming and changes in precipitation regimes, which in turn, have already altered the seasonal water balance and river run-off in the Baltic States (Kriauciuniene et al., 2012). The magnitude of the climatic change and seasonal timing in the region is not equally distributed and depend on the geographical location (Lind and Kjellström, 2009; BACC II Author Team, 2015). In Estonia, the most pronounced warming has been taking place in winter and early spring (Jaagus, 2006; Tarand et al., 2013). Winters have become milder and shorter, as the number of very cold days has decreased (HELCOM, 2013). The duration of snow cover has become shorter by days, with the greatest changes in coastal areas (Jaagus, 2006). Therefore, the growing season starts earlier (Ahas and Aasa, 2006). Projections of future climate change, based on regional climate models show that a similar warming tendency will continue during the 21 st century, however, multiple uncertainties regarding precipitation patterns remain (Jaagus and Mändla, 2014). Both a precipitation increase and decrease are projected, and the results of a simulated future climate seem to depend on the used model, emission scenario and may vary by the geographical location (Lind and Kjellström, 2009; Jaagus and Mändla, 2014). A precipitation increase during the cold season is expected in most of the Baltic Sea region, including Estonia, however, much drier spring and summer seasons are also possible under future climate scenarios, especially in the southern areas of the region (Kjellström and Ruosteenoja, 2007; Rutgersson et al., 2015). Higher temperatures together with changes in precipitation patterns have already been altering seasonal hydrological regimes (Graham, 2004; Kriauciuniene et al., 2014) which may affect forest growth (Tullus et al., 2012; Rosenvald et al., 2014; Sellin et al., 2017) and productivity in the region (Lindner et al., 2010). Because the growth and wood production of trees are closely related to temperature fluctuations (Schmitt et al., 2004, Rossi et al., 2008) the question arises as to what the growth response to altered annual weather fluctuations is and what consequences it will have on the growth dynamics and functioning of forest ecosystems in the region. Therefore, a better understanding of interactions between tree growth and shifting climatic conditions is greatly needed

143 Scots pine (Pinus sylvestris L.) is the most abundant (Navasaitis, 2004; Eckenwalder, 2013) and economically important (e.g. Kaimre, 2014) coniferous tree species in Northeastern Europe. In addition to wood and non-wood products, hemiboreal pine forests provide a number of other ecosystem services, including soil and water protection (Wisńiewski and Kistowski, 2015), carbon sequestration and storage (Karjalainen, 1996; Vucetich et al., 2000; Kolari et al., 2004), provision of habitats for wildlife (Marmor et al., 2013; Laarman et al., 2013), recreational opportunities (Riepšas, 2012), as well as aesthetic and cultural values. Because of great importance for the region and local communities, a profound understanding of interactions between tree growth and climate variability and responses to change is needed to promote sustainable forest development. The annual radial growth of trees, including that of Scots pine, have been extensively used to investigate growth responses to past climate (e.g. Kirchhefer, 2001; Linderholm, 2002; Friedrichs et al., 2008; Büntgen et al., 2012) and assess possible impacts of environmental change on the studied populations (e.g. Bauwe et al., 2013). An understanding has developed (based on previous spatially fragmented studies) that the radial growth of Scots pine populations in the northern latitudes (boreal forests) is limited by low temperatures during the growing season (Briffa et al., 1990; Babst et al., 2013) while the growth of pine populations at their southern range limit can be related to received precipitation. However, recent studies (e.g. Sánchez- Salguero et al., 2015; Hellmann et al., 2016) have used tree-ring networks and found that on the spatial scale responses are much more diverse. In the central part of tree species distribution, climatic responses are more complex and depend on both local climatic and local site conditions. In hemiboreal forests, cold winters (Bitvinskas, 1974; Lõhmus, 1992b) and low spring temperatures (Läänelaid and Eckstein, 2003; Hordo et al., 2009, 2011) have previously been identified as climatic factors that limit Scots pine growth. Most of the previous knowledge related to climate influence on Scots pine growth in Estonia originates from smaller-scale dendrochronological studies restricted to a certain location (Pärn, 2003, 2008; Metslaid et al., 2016) or based on general master chronologies (Lõhmus, 1992a, 1992b; Läänelaid and Eckstein, 2003). Lõhmus (1992b) used a large set of tree-ring data collected across Estonia and investigated climate influences on Scots pine growth considering the fertility gradient of sites but not focussing in more detail on regional differences in local climates. Thus, a comprehensive understanding of climate influence on Scots pine growth on the spatial scale in the country is still limited (e.g. Hordo et al., 2009, 2011). Several recently conducted spatially explicit dendrochronological studies in Estonia, which investigated Norway spruce (Helama et al., 2016), common oak (Sohar et al., 2014) and also Scots pine (Hordo et al., 2009) showed that speciesspecific responses to climate across the country differ, suggesting that local climatic conditions are diverse enough to result in contrasting responses, as reported for other geographical areas (e.g. Mazza et al., 2014). In addition to spatial variation in climatic conditions, the dominance of the climatic variables or seasonal timing may also shift, due to temporal changes in climatic conditions (Moir et al., 2011; Büntgen et al., 2012). To improve our understanding of climate influences on Scots pine growth on the spatial scale in Estonia, to define climatic variables which could be further used for future growth modelling, and to make models climate-sensitive, we used a tree-ring network distributed across the Estonia, which includes a range of climatic and site conditions (Kiviste et al., 2015). Since soil conditions are shown to be important in mediating climate influences (e.g. Moir et al., 2011) we studied Scots pine response to climate variability based on growing conditions as defined by classification into forest site types. A forest 142 3

144 site type in general terms describes edaphic and hydrological site conditions and is the main classification unit used in forest management in Estonia (Lõhmus, 2004). Therefore, the aims of this study were 1) to quantify climatic variables that are responsible for most of the annual variability in Scots pine tree-rings, considering ecological site conditions and local climates, 2) to define similarly responding populations and 3) to test if relationships remained stable over the last century when most of the warming has taken place. 2. Material and Methods 2.1. Study area The study was conducted in Estonia, the country in Northeastern Europe that spans between 59.5 N, 28 E and 57.5 N 22 E. The area belongs to the hemiboreal vegetation zone, which is a transition area between boreal and temperate forests (Ahti et al., 1968). Coniferous tree species, including Scots pine and Norway spruce, prevail in the forests and are the most relevant tree species for local forestry, followed by birch, aspen and alder species (Raudsaar et al., 2014). The topography of Estonia is relatively flat, with more hilly landscape (max elevation 318 m a.s.l.) in the Southeastern part of the country. The climate in Estonia is temperate (Hordo et al., 2011), characterized by cold snowy winters and warm summers. The variation in air temperatures between different sub-regions of Estonia mainly arises due to the proximity of the Baltic Sea (Jaagus, 2006; Tarand et al., 2013). Along the coast and on the islands, maritime climate prevails, gradually changing to semi-continental when moving towards the inland. As a result, winters tend to be milder in the coastal areas and colder further inward the country, with the opposite effect in early spring and summer. Mean annual temperature in Estonia (in ) ranged between 4.6 ºC and 6.8 ºC, being higher on the West Estonian Archipelago and lower in the eastern part of the country (Tarand et al., 2013). The coldest month is February (mean air temperature -2.5 to -6.5 ºC) and the warmest month is July (mean temperature to ºC). The number of sunshine hours is higher on the islands and in coastal areas than in the inland, ranging from 1,600 to 1,900 hours respectively. Accordingly, the growing season (average for Estonia days) is longer in the coastal areas. However, seasonal change is slightly delayed in the coastal areas and first approaches the Southeastern part of Estonia and then gradually spreads across the other sub-regions. The climate in the Northeast tends to be slightly cooler and the growing season is shorter than in the rest of the country. In general, the climate in Estonia is humid, with the annual average relative air humidity being around 80 percent. The mean annual precipitation varies between 570 and 750 mm (reference period ; Tarand et al., 2013). It is unevenly distributed across the country and is determined by landscape topography (Jaagus, 2003). Continental Estonia, particularly uplands in the Southeast and Southwest of the country receive more precipitation than the rest of the regions, especially in autumn. Precipitation is much lower in the coastal areas and on islands (Tammets and Jaagus, 2013), especially in spring and in the beginning of summer (Fig.1 and Fig. A.1)

145 Fig. 1 Summary of climatic conditions in different study sub-regions for The solid line in climographs refers to mean air temperature ( C) and the dashed line to mean precipitation sums (mm). T shows mean annual temperature ( C) and P indicates mean annual precipitation (mm) for the period of from meteorological stations. NE Northeast, ISL Islands, SE Southeast, SW Southwest Tree-ring network and ecological site conditions Sampling of increment cores of Scots pine trees was carried out in 2007 using a layout of the Estonian Network of Forest Research Plots (ENFRP; Hordo et al., 2009; Kiviste et al., 2015). The ENFRP is distributed across the country and covers a wide range of ecological and climatic conditions. Increment core sampling was restricted to plots established in Scots pine dominated forests growing on mineral soils, including the most common forest site types (Myrtillus, Rhodococcum, Calluna and Cladonia) for this tree species in Estonia (Fig. 2). For this study, we also used a set of annual radial increment data (1.3 m above ground cross-sections) gathered in 2008 (Metslaid et al., 2011) from Southeast Estonia. Since a forest site type in general terms describes edaphic and hydrological site conditions and is the main classification unit used in forest management in Estonia (Lõhmus, 2004), ecological site conditions in this study are described by the four above-mentioned forest site types. Cladonia is a nutrient-poor site type on soils consisting mainly of fine sand, containing very little clay. The organic layer is thin, decomposing slowly. The groundwater in most cases is deeper than 3 meters, therefore the site is very prone to droughts (Lõhmus, 2004; Pärn, 2009). This site type prevails in the North, Northwest, Southeast of Estonia and on islands. Calluna forest site type similarly to Myrtillus and Rhodococcum site type are mesic and more productive due to greater clay content. The level of groundwater is deeper 144 5

146 than 2 meters but can vary depending on topography. The top layer of the soil may dry out periodically (Lõhmus, 2004). Moisture to tree roots is better supplied on Myrtillus sites, besides that, clay content on these sites might be somewhat greater compared to Rhodococcum sites, which facilitates tree growth. Myrtillus forest site type is the most productive site among the investigated forest site types. Scots pine achieves somewhat lower productivity on Rhodococcum site type, followed by Calluna forest site type, whereas Cladonia is the least fertile forest site. All forest sites in the country are classified according to the same system, however, soils on the islands are much shallower than on mainland (Reintam and Kõlli, 2012). Pärnu Fig. 2 Geographical location of Scots pine increment sampling sites (dots) and meteorological stations (triangles) in Estonia. Abbreviations refer to sub-regions as follows: ISL Islands, NE Northeast, SE Southeast, SW Southwest Sampling, measurements and chronology development For tree-ring analysis, at least 8 dominant or co-dominant trees were sampled outside the selected ENFRPs, by extracting double-radii core at 1.30 m above the root collar. Pine trees were sampled at 126 sample plots, but only a subset of well inter-correlating ring-width measurement series was used in this study to build chronologies and examine growth-climate relationships with regard to forest site types. Depending on the geographical location, individual sampling sites were assigned to one of four sub-regions: Southeast (SE), Southwest (SW), Northeast (NE) and Islands (ISL; Fig. 2). Samples were prepared in the laboratory by applying dendrochronological techniques (Pilcher, 1990). Tree-ring widths of the increment cores were measured with 0.01 mm measuring precision, using the measurement table Lintab TM 5 equipped with a stereo microscope and TSAPWin TM software (RINNTECH, Heidelberg, Germany). Time 145 6

147 series of the ring-width measurements were cross-dated (Fritts, 1976) within the tree and among several other trees, to assign correctly the years of tree-ring formation. Tree-ring measurement series were visually and statistically cross-dated using TsapWin TM program (Rinn, 2003). Statistics, including Gleichläufigkeit (sign agreement; Eckstein and Bauch, 1969) and student s t- test (shows a similarity between series) were used during the synchronization procedure. Two radii from the same tree were averaged to produce a mean curve for individual trees. Crossdating accuracy was further checked using COFECHA software (Lamont-Doherty Earth Observatory, Palisades, USA; Grissino-Mayer, 2001) by using mean tree-ring width series for each tree. Synchronous series were further individually detrended with a spline (50% of frequency response and cut-off in 67% of series length) to remove the age trend and impulses related to forest stand structure changes from the growth patterns (Cook and Peters, 1981). The tree-ring width index was calculated by dividing the raw ring-width measurement by the fitted spline value. Autoregressive modelling was used to remove autocorrelation in the individual series (Box and Jenkins, 1970). The order of the autoregressive model was defined using the first minimum Akaike information criterion (Akaike, 1974). Pre-whitened index series were further grouped according to forest site type and sub-region and averaged into mean residual chronologies using Tukey s robust mean, which enables exclusion of outliers (Mosteller and Tukey, 1977). Finally, chronologies were truncated to the calendar year, replicated at least by five trees. To evaluate common variability and assess the chronology suitability for dendroclimatic analysis, we calculated several statistical measures, including mean sensitivity (MS), first-order autocorrelation (AC1), expressed population signal (EPS; Wigley et al., 1984) and variance captured by the first principal component (PC1; Fritts, 1976) on each set of detrended increment series, grouped by forest site type and sub-region. Standardization was achieved and dendrochronology statistics were calculated using R software (R Core Team, 2016) and package dplr (Bunn, 2008; Bunn et al., 2013) Meteorological data Mean monthly temperature and total precipitation data from four meteorological stations, including Kunda, Tõravere, Pärnu and Ristna, provided by the Estonian Weather Service under the Estonian Environmental Agency were used to characterize weather fluctuation in the selected sub-regions. The data from the nearest meteorological station to a sampling site were used in calculations. Therefore, data from the meteorological station in Tõravere were used for sites in the Southeast, data from Kunda meteorological station were utilized for determining growth-climate relationships in the Northeastern sub-region, data from Ristna station were used for the Islands and meteorological data from the station in Pärnu were used to describe weather variation in the Southwestern sub-region. The length of digitally available data records differed whereas a common period covered by all stations started in 1946 (Table 1) Statistical analysis To assess the influence of annual climate variability on Scots pine radial growth, relationships between forest site type residual chronologies and monthly resolved climatic variables (mean monthly temperature and precipitation sums) were used to calculate correlation coefficients 146 7

148 (Fritts, 1976). The analysis was restricted to a 52-year-long period ( ) which was common for all forest site type chronologies and meteorological data (Tables 1 and 2). Climatic variables used in the growth-climate analysis were mean monthly temperatures and monthly precipitation sums from the previous June to the current August. The significance of the relationships at the 95 th percentile was tested by applying the bootstrapping procedure (Guiot, 1991) by running 1000 iterations. The dendroclimatic analysis was conducted with R software (R Core Team, 2016), using the functions from the package treeclim (Zang and Biondi, 2015). Table 1. Summary of mean climatic characteristics over the period of for different sub-regions of Estonia. Sub-region Islands Southwest Southeast Northeast Station name Ristna Pärnu Tõravere Kunda Latitude N N N N Longitude E E E E Data available since Mean annual temperature ( C) Mean annual precipitation (mm) Growing period temperature ( C) Growing period precipitation (mm) Mean July temperature ( C) Mean July precipitation (mm) Mean February temperature ( C) Difference between mean July and February temperatures ( C) Previously, several approaches have been used to define similarly responding populations, including a comparison of growth patterns (e.g. Mäkinen et al., 2001; Sohar et al., 2014) or a comparison of responses to annual climate variability patterns (e.g. Nash and Kincaid, 1990; van der Maaten, 2012; Lebourgeois et al., 2013). In this study, to get a comprehensive understanding of pine responses to weather fluctuations on the spatial scale, first we compared indexed growth patterns using hierarchical cluster analysis (HCA), considering the period of that was common for all chronologies. In HCA, dis/similarities between growth patterns were assessed by estimating Euclidean distances, according to Ward s method. To examine the influence of ecological site conditions and local climate on growth-climate relationships, principal component analysis (e.g. Tardif et al., 2003; Mérian et al., 2011; Lebourgeois et al., 2013) was performed on previously calculated bootstrapped correlation coefficients (BCC). The input data for this analysis consisted of the BCC matrix, containing only those climate variables that were statistically significant (p<0.05) for at least one chronology. PCA analysis was conducted using the variance-covariance matrix. Temporal stability of the climate-growth relationships was examined using moving correlation analysis. Correlation coefficients between residual chronologies and climate variables over the period of were calculated in a 30-year window, shifted by one year. The 147 8

149 bootstrapping method was applied and 95% confidence intervals were calculated to test the statistical significance of the correlation coefficients. Moving correlation analysis was performed with R software (R Core Team, 2016), using the treeclim package (Zang and Biondi, 2015). 3. Results 3.1. Statistical characteristics of the chronologies Altogether, 672 Scots pine trees were used and 13 forest site chronologies for four sub-regions developed (Fig. 3). The chronologies extended between 52 to 192 years (mean 92 years) with the shortest chronologies in the Southwestern sub-region (Table 2). The mean annual tree-ring width over the period ranged from 0.95 to 2.06 mm (overall mean 1.3 mm) and on average was slightly greater in the Southeastern and Southwestern pine populations, compared to the pine trees originating from the Islands and the Northeast. The trees sampled on the Islands and in the north of the country were somewhat older, as indicated by the mean series length compared to inland sites in the southern part. The number of samples used to build a chronology varied greatly among the sites and ranged from 14 to 173. The mean series intercorrelation (IC) ranged from to and was slightly greater for the Islands and the Northeastern sub-region, indicating a stronger common signal in these pine populations. Sign accordance (Glk) was in agreement with IC spatial patterns and was in the range. The overall mean sensitivity (MS) was 0.18 and did not vary much among the chronologies ( ), being slightly greater in the Northeastern sub-region. Standard deviation (SD) with a range of followed the MS variation pattern. First-order autocorrelation (AC1) was high in the raw series (data not shown) but when calculated in the detrended series it was moderate, ranging from 0.35 to 0.55, being relatively greater for mesic sites in Northeastern and Southeastern sub-regions. The common variance, captured by the first principal component (PC1) that is assumed to show climate influence on the investigated growth series (Fritts, 2001) ranged between 27% and 54% and tended to be the lowest for sites on the Islands, with no trends observed across an ecological gradient. For most of the chronologies, except the Calluna site in the Northeast, replication of 5 trees was sufficient to reach robust signal strength, EPS 0.85, which is an accepted threshold for a reliable chronology (Wigley et al., 1985). Since replication of five series was not sufficient, for this site we truncated the chronology to the calendar year when it reached the 0.85 limit. The EPS was greater for well-replicated chronologies and for chronologies with a higher IC, however it tended to be lower for most of the chronologies of dry sites (Calluna, Cladonia). Descriptive statistics suggested that the common signal and thus climatic influences on Scots pine growth are greater on sites in the Northeast whereas annual growth variation was much lower, suggesting that growth is steady in two mesic sites (MyrtSW, RhodSW) in the Southwest Growth-climate relationships Most of the forest site type chronologies were positively correlated with winter (December January February) and early spring (March April) month temperatures (Fig. 4). The 148 9

150 strongest relationship between Scots pine growth and climate was observed for mean temperature in March April (r= ). The associations with early spring temperatures were stronger for moderately mesic-fertile forest site types (Myrtillus and Rhodococcum). The relationship with February temperatures was significant for nine out of thirteen chronologies, with the strongest linkage for Myrtillus and Rhodococcum sites on the Islands. A great number of chronologies, particularly those from the Islands and the Northeast, showed positive correlations with previous August precipitation (r= ) in combination with an inverse correlation to the same month temperature (r= ), implying that climatic moisture availability at the end of the growing season has a significant influence on the tree-ring width of the next year. All forest site types in the Southeast, Myrtillus forest site type in the Northeast and both mesic forest sites on the Islands were significantly positively related to current year June temperature. Previous year September precipitation correlated negatively with Scots pine growth in four forest site types, including the Rhodococcum site in the Southwest (r=-0.33), Southeast (r=-0.32) and both mesic sites (MyrtISL and RhodISL) on the Islands (r=-0.36; r=-0.40). Several chronologies, from Islands and Southeast (MyrtISl, RhodISL, CladSE) and both pine chronologies from Southwest showed not a very strong (r= ), but a significantly positive relationship to current March precipitation. Sites in the Northeast were positively related to the total rainfall in current August (r= ), whereas greater precipitation at the end of summer was inversely linked (r=-0.31) to pine radial increment for RhodSW chronology

151 Sample depth Cladonia NE RWI Sample depth Cladonia ISL Sample depth RWI Calluna NE RWI RWI Sample depth Rhodococcum NE Calluna ISL RWI Rhodococcum ISL Sample depth Sample depth Myrtillus NE Sample depth RWI Cladonia SE Sample depth RWI Myrtillus ISL Rhodococcum SW RWI Sample depth RWI Sample depth Myrtillus SE RWI RWI Sample depth Rhodococcum SE RWI Sample depth RWI Sample depth Myrtillus SW Fig. 3 Residual tree-ring-width chronologies (lines) and sample sizes (grey area) of Scots pine for each forest site type by geographical sub-regions in Estonia. Forest site types from Cladonia to Myrtillus describe moisture availability and the productivity gradient (from the driest and poorest to the most mesic and productive). Abbreviations of sub-regions: NE Northeast, ISL Islands, SE Southeast, SW Southwest

152 Table 2. Descriptive statistics of Scots pine residual chronologies by forest site type and sub-region. Abbreviations: RW ring width (mm), IC interseries correlation, Glk Gleichläufigkeit, MS mean sensitivity, SD standard deviation, AC1 first-order autocorrelation, SNR signal to noise ratio, EPS expressed population signal, PC1 variance (%) accounted by the first principal component (eigenvector). Chronology name Forest site type Region N of trees (cores) Series length (years) min/mean/max Chronology span (n>4 series) Chronology length (n>4 series) Treering width (mm) IC Glk MS SD AC1 SNR EPS PC1 CalISL Calluna Islands 19 (38) 17/73/ CladISL Cladonia Islands 29 (58) 41/85/ MyrtISL Myrtillus Islands 90 (180) 21/65/ RhodISL Rhodococcum Islands 77 (154) 20/74/ CalNE Calluna Northeast 21 (42) 38/89/ CladNE Cladonia Northeast 173 (346) 35/68/ MyrtNE Myrtillus Northeast 40 (80) 26/58/ RhodNE Rhodococcum Northeast 24 (48) 52/65/ CladSE Cladonia Southeast 14 (28) 27/74/ MyrtSE Myrtillus Southeast 24 (48) 39/67/ RhodSE Rhodococcum Southeast 97 (194) 20/67/ MyrtSW Myrtillus Southwest 31 (62) 20/40/ RhodSW Rhodococcum Southwest 48 (96) 34/45/

153 Fig. 4 Pearson s correlation coefficients between residual ring-width chronologies and mean monthly temperature and total precipitation in Dark bars refer to statistically significant (p<0.05) coefficients. Calendar months of the previous year (t-1) are indicated in lowercase letters and months of the current year (t) in capital letters. Climatic variables are arranged from June (j) of the previous year to August (A) of the current ring formation year. Chronology abbreviations are explained in Table

154 3.3. Similarities in growth patterns and climatic responses Hierarchical cluster analysis, conducted on 13 residual chronologies covering the period of , revealed a greater similarity between indexed growth patterns from the same subregion (Fig. 5), than between chronologies of the same forest site type. Three main clusters were produced by the HCA. Chronologies from the Southeast and Southwest were compiled into one cluster, implying a greater similarity between growth patterns. The second cluster consists of forest site type chronologies from the Islands, while the last cluster, bearing least similarity to the rest of the growth patterns, consisted of pine chronologies from the Northeast. A closer resemblance in all cases was observed between Calluna and Cladonia site types and Rhodococcum and Myrtillus sites, suggesting the importance of ecological site conditions in shaping annual growth variations. MyrtNE chronology was assigned into the Southeast-Southwest cluster and was the only chronology which was included in a cluster of a sub-region different from which it actually belonged to. Fig. 5 Dendrogram of HCA performed on 13 residual forest site chronologies, based on the period of Chronology abbreviations are explained in Table 2. Grey lines separate three clusters. The principal component analysis confirmed the results of the HCA and suggested that variation in Scots pine sensitivity to climate is related to special distribution and can be summarized by geographical sub-regions. Altogether the first two principal components captured 70.9% of the total variance (Fig. 6). The first principal component (PC1) resumed 48.3% of the total inertia and was associated with Scots pine sensitivity to climatic conditions of the current growing season. Temperature and precipitation in August (T8 and P8, respectively), previous year December temperature (t12), previous year September precipitation (p9) and current year March April temperatures had the greatest loading on this dimension. As observed from the score (chronology) distribution (Fig.6), climatic sensitivity captured by this component is characteristic to most of the chronologies in Southeast, Islands and Southwest, but contradictory to pine populations (CalNE, CladNE, RhodNE and CladISL) growing on dry and infertile sites under maritime climatic conditions. The second dimension (PC2) accumulated 22.6% of the

155 total variance and captured Scots pine sensitivity to water availability, mainly at the end of the previous growing season. Precipitation in July, August and November (p8, p7 and p11, respectively) and temperature in August (t8) of the previous year and current July precipitation (P7) were strongly related to this principal component. Chronology scores suggested that, sensitivity to previous summer water availability was most characteristic to pine chronologies from Islands but not limiting in Southwest sub-region. Fig. 6 PCA ordination diagram of Scots pine response to climate variability. PCA was performed on statistically significant BCC, calculated between residual chronologies and climate data. Abbreviations: ISL Islands, NE Northeast, SE Southeast, SW Southwest; Letter and number combination refers to climatic variables of a certain calendar month (1-January, 2-February, etc.). Lower case letters (p, t) refer to the previous year and capital letters to the current year climatic variables (p-precipitation and t-temperature, respectively). Abbreviations of chronologies are explained in Table Temporal stability of climate-growth relationships According to the temporal stability analysis, relationships between growth and climate variables in were not very stable (Fig. A.2). Statistically significant (p<0.05) linkage to winter (t12, T1, T2) and early spring (T3, T4) temperatures was strongest at the beginning of the analysis period, but the significance progressively vanished already at the end of the twentieth century (e.g. MyrtISL, MyrtSW, MyrtSE). Regarding seasonal timing, the significance of temperature in winter months for pine radial growth strongly diminished towards the end of the analysis period (e.g. CladISL, MyrtSW, RhodSW, RhodSE), with statistically significant linkage remaining only for March and April. Moving correlation analysis revealed that the radial growth of Scots pine populations in the Northeastern region was weakly linked to temperatures of winter and early spring already in the middle of the 20 th century, except in Myrtillus (MyrtNE) and Calluna (CalNE) forest site types, where linkage of thermal conditions at the end of the dormant season become progressively weaker at the end of last century. The correlation to climatic conditions of the previous growing season, particularly August mean temperature (t8) and the total precipitation (p8) for pine growth on the Islands and the Northeastern region are not statistically significant for the entire period of Positive correlation between growth

156 on Myrtillus (MyrtNE) and Rhodococcum (RhdNE) sites in the Northeast became stronger in recent decades. While a negative association between tree-ring width and mean August temperature of the previous year seems to get stronger for pine trees on Cladonia site in the Northeast and on the Islands. 4. Discussion Since the Baltic countries appear in the central part of Scots pine natural distribution (Laas, 1987), optimal climatic conditions for this tree species should exist in this part of Europe. In this study, we investigated climate influence on Scots pine radial growth on the spatial scale considering ecological site conditions and found that over the period of , which is characterized by rapid warming in the region (Jaagus 2006), a clear spatially explicit variation in Scots pine responses (Fig. 4). Differentiation between growth responses to climate observed in our study seemed to be preconditioned by variability in local climatic conditions over the period of (Table 1, Fig. 1). Thermal conditions on the coast and Islands over the studied period were warmer by around 1 C, particularly because of changes in winter. Seasonal moisture conditions in this part of Estonia also differ considerably and are characterized by very dry conditions at the beginning of the vegetation period. Likewise in other studies, investigating species-specific responses on the spatial scale (e.g. Lindholm et al., 2000), we found stronger regional coherence in growth patterns (Fig. 5). The results of PCA, conducted on the BCC matrix, confirmed that climatic variables responsible for most of the year-to-year variability in Scots pine tree-ring widths were more similar within the same study sub-region than for the same forest site type in different regions (Fig. 6). Ecological site conditions certainly played an important role in determining growth rates, however remained on the second place. Regional differentiation between pine responses to climate variability at forest type level in Estonia has been reported by Hordo et al. (2009), whereas Lõhmus (1992) studied climate influence on Scots pine growth in the country with respect to site conditions, and concluded that the main growth limiting factor in this part of Europe is low winter temperature. Recently contrasting responses to inter-annual variation have been observed for common oak (Sohar et al., 2014) and Norway spruce (Helama et al., 2016). The first two components of PCA, conducted on the BCC matrix, captured 70.9 % of the total variance, suggesting that there is a high similarity in Scots pine response to climate variability in the studied region. Nevertheless, Scots pine chronologies from the Northeast according to both HCA and PCA results, were separated from the rest of the sites (Fig. 5, Fig. 6), implying that growth patterns and responses to climate variability in these pine populations were somewhat different compared to the rest. We hypothesise that a sharp distinction between Northeastern chronologies from the rest of the growth patterns could be due to dust pollution (Pärn, 2003). Even if most of our sampling sites in this sub-region are concentrated further from pollution sources, it is known that emissions were much higher in the past (Pärn, 2003). Since tree-rings are good indicators and archives of environmental changes (Fritts, 1976; Cook and Kairiūkštis, 1990), this effect can be recorded in tree-ring series. In addition, descriptive statistics for Northeastern Scots pine chronologies indicated greater climatic sensitivity compared to other chronologies, suggesting a more pronounced influence of abiotic factors in this region. Pärn (2003) studied the impact of dust pollution on Scots pine growth and found that pine

157 populations growing close to a pollution source were more sensitive to macroclimate fluctuations. Considering that species-specific radial growth response to annual climate variability depends on a number of factors, including ecological site characteristics (e.g. Tardif et al., 2003; Moir et al., 2011; Hökka et al., 2012) and stand structure (Guillemot et al., 2015; Metslaid et al., 2016), we developed chronologies for forest site types rather than for a separate sampling location. By averaging growth patterns from several locations of the same forest site type, we expected to reduce the non-climatic signal related to management activities and stand structure dynamics inherent to a single stand. In that way, common variability, which is assumed to be caused by large-scale environmental factors like climate, characteristic of selected ecological conditions, would be enhanced (Fritts, 1976). Most of the forest site chronologies in our study correlated positively with February and March April temperatures and these were among the most important climatic variables throughout the country, with weaker and usually insignificant ties in the driest forest sites (Calluna and Cladonia) and populations in the Northeast. The strong linkage to winter/early spring temperatures implies that wider rings in Scots pine are produced in years with warm winters and springs. This is in accordance with the previous studies conducted in the Baltic region (Bitvinskas, 1974; Lõhmus, 1992; Stravinskienė, 2002; Läänelaid and Eckstein, 2003; Vitas, 2008; Hordo et al., 2009, 2011). Scots pine sensitivity to winter/early spring temperatures has been observed in Southwest Finland (Helama et al., 2014), in Poland (Koprowski et al., 2010), and in West and Central Germany (Bauwe et al., 2013). Several physiological explanations could be proposed for this relationship, including frost-induced damage to fine roots during very cold winters, especially when the insulating snow cover is very thin or absent (Tuovinen et al., 2005). Severe winters may induce coniferous tree defoliation in the following spring (Kuulman, 1991), leading to lower stem biomass production due to reduced photosynthetic capacity (Ericsson et al., 1980). According to Ensminger et al. (2004), the most stressful seasons for Scots pine forests in boreal climates are winters and early springs when low temperatures restrict photosynthesis. They found that the production of pigments that regulate the amount of chlorophyll and progress of photosynthetic recovery is closely related to thermal conditions in spring, and temporal lowtemperature spells during this period may even have a reversed effect on physiological recovery. Moreover, delayed snow melt and soil thaw after very cold winters may restrict water and nutrient uptake and flow in the tree and delay initiation of cambium activity (Vaganov et al., 1999). Therefore, warm springs also favour the recovery of transport capacity in xylem and phloem (Vanhatalo et al., 2015). In addition to the previously observed Scots pine sensitivity to winter/spring temperatures, a positive relationship to previous August precipitation and a negative correlation to the same month mean temperature suggested that water deficit at the end of the growing season is strongly negatively related to the next year s tree-ring width. This relationship is characteristic of coniferous trees growing in temperate forests of central Europe (Oberhuber et al., 1998; Weber et al., 2007, Lebourgeois et al., 2013). In Estonia, associations between Scots pine radial growth and rainfall have been observed in several cases in Scots pine plantations on reclaimed areas (Metslaid et al., 2016), pine populations growing in dust polluted areas (Pärn, 2003), but also in dry pine forests on Hiiumaa island (Hordo et al., 2009, 2011). In this study, the inverse correlations to previous year August precipitation and mean temperature were the strongest for pine populations growing on the Islands and in Northeastern sub-regions, disregarding the forest

158 site type. Scots pine is known as a drought-resistant tree species that during dry periods reduces water-use by closing stomata. Since carbon assimilation in pine is inhibited during such periods, dry and hot climatic conditions at the end of summer may impact root elongation (Konôpka et al., 2005), carbohydrate storage (Ericsson et al., 1980) and bud development, which may, in turn, reduce tree growth potential in the following year. As an alternative explanation for the lagged association between the growth rate and climatic conditions of the previous season, tree flowering (Sensuła et al., 2015) or seed production have been suggested (Hacket-Pain et al., 2015). Earlier research has reported that warm and dry climatic conditions at the end of the growing season promote the formation of generative buds in some tree species (la Bastide and van Vredenburch, 1970). For instance, in Scots pine, needle buds are replaced by male flower buds, which results in lower foliage in the following summer (Innes, 1994). Resource allocation during flowering, especially followed by subsequent seed production may restrict radial growth in masting tree species, including Pinus genera (Hacket-Pain et al., 2015). Moreover, radial growth in Calluna and Cladonia forest sites on the Islands were strongly related to spring precipitation, which could be explained by much dryer climatic conditions during spring-time in this part of the study area. It should be noted that common variance as described by dendrochronology statistics in these two forest site types was lower compared to more fertile and mesic Myrtillus and Rhodococcum sites; therefore, we established statistically less reliable relationships between growth and climate for these two forest sites (Fig. 5). Less coherent radial increment patterns could result from stand dynamics or pine defoliators who commonly attack pine stands on drier sites (Gedminas, 2003; Gedminas et al., 2004). Considering that climatic conditions on the Islands are warmer and dryer, pine populations growing in such conditions could have been already adapted to warmer and dryer growing conditions that could possibly emerge in the rest of the country by the end of the century (Jaagus and Mändla, 2014). Growth rates of such moisture-sensitive pine populations were very low, particularly in old trees, suggesting that trees are just maintaining basic survival functions, but little carbon is left for allocation into radial growth. Low growth rates could explain lower coherency between growth series in these sites because little variation is contained in such growth patterns. Based on previous research findings (Martin-Benito et al., 2011) less dense stands and density reduction through management activities could reduce growth decline in stands with water deficit stress and could even increase their productivity. Besides that, seeds originating from warmer and drier climates, when transferred into more favourable conditions have shown to produce wider rings and achieve greater basal area growth compared to local populations (Savva et al., 2007). Knowledge on Scots pine response to climate variability obtained in this research is recommended to be considered when preparing adaptation strategies for forest management. 5. Conclusions We used an extensive tree-ring network distributed across Estonia and based on dendrochronological methods, determined the most influential climatic variables and seasonal timing for Scots pine radial growth in hemiboreal ecosystems of Europe. Our results suggest that growth-climate relationships are greatly influenced by local climatic conditions, followed by ecological site characteristics. The radial growth over the period of in most studied

159 pine populations favoured mild winters and warm spring conditions and was restricted by warm and dry conditions at the end of the previous year summer. The relationship to water deficit at the end of summer was stronger for Scots pine stands growing under maritime climatic conditions, particularly on the Islands and Northeast coast. Pine trees growing under semicontinental climatic conditions (SE) and also in the high-precipitation sub-region (SW) showed stronger sensitivity to late winter/early spring temperatures and weaker correlations to precipitation. An analysis of temporal stability showed that relationships to temperatures of the dormant season are getting weaker, while associations with summer precipitations are becoming more significant. Temporal instability in growth-climate relationships and considerable variation on the spatial scale encumber assessment of climate warming effects on future Scots pine productivity in the region. Acknowledgments The study was conducted at the Estonian University of Life Sciences. The study was supported by the Institutional Research Funding IUT21-4 of the Estonian Ministry of Education and Research. The Environmental Investment Centre supported increment core sampling from the ENFRP and measurements. The Estonian Weather Service at the Estonian Environmental Agency provided meteorological data. References Ahas, R., Aasa, A., The effects of climate change on the phenology of selected Estonian plant, bird and fish populations. International Journal of Biometeorology. 51(1), Ahti, T., Hämet-Ahti, L., Jalas, J., Vegetation zones and their sections in northwestern Europe. Annales Botanici Fennici. 5(3), Akaike, H., A new look at the statistical model identification. IEEE Transections on Automatic Control. 19(6), Babst, F., Poulter, B., Trouet, V., Tan, K., Neuwirth, B., Wilson, R., Carrer, M., Grabner, M., Tegel, W., Levanič, T., Panayotov, M., Urbinati, C., Bouriaud, O., Ciais, P., Frank, D., Site- and species-specific responses of forest growth to climate across the European continent. Global Ecology and Biogeography. 22(6), BACC Author Team, Assessment of Climate Change for the Baltic Sea Basin. Springer. BACC II Author Team, Second Assessment of Climate Change for the Baltic Sea Basin. Springer-Verlag, Berlin. Bauwe, A., Koch, M., Kallweit, R., Konopatzky, A., Strohbach, B., Lennartz, B., Tree-ring growth response of Scots pine (Pinus sylvestris L.) to climate and soil water availability in the lowlands of northeastern Germany. Baltic Forestry. 19(2), Bauwe, A., Jurasinski, G., Scharnweber, T., Schrööder, C., Lennartz, B., Impact of climate change on tree-ring growth of Scots pine, Common beech and Pedunculate oak in northeastern Germany. IForest. 9(Feb 2016),

160 Box, G.E.P., Jenkins, G.M., Time Series Analysis: Forecasting and Control. Holden-Day, San Francisco. Briffa, K.R., Bartholin, T.S., Eckstein, D., Jones, P.D., Karlén, W., Schweingruber, F.H., Zetterberg, P.A., ,400-year tree-ring record of summer temperatures in Fennoscandia. Nature. 346, Bunn, A.G., A dendrochronology program library in R (dplr). Dendrochronologia. 26(2), Bunn, A., Korpela, M., Biondi, F., Campelo, F., Mérian, P., Qeadan, F., Zang, C., dplr: Dendrochronology Program Library in RR Package Version Büntgen, U., Kaczka, R. J., Trnka, M., Rigling, A., Ensemble estimates reveal a complex hydroclimatic sensitivity of pine growth at Carpathian cliff sites. Agricultural and Forest Meteorology. 160, Bitvinskas, T., Dendroklitatologicheskije isledovaniji [Denroclimatological Investigations]. Gidromeoisdat, Leningrad. Büntgen, U., Kaczka, R.J., Trnka, M., Rigling, A., Ensemble estimates reveal a complex hydroclimatic sensitivity of pine growth at Carpathian cliff sites. Agric. For. Meteorol. 160, doi: /j.agrformet Camarero, J.J., Gazol, A., Sancho-Benages, S., Sangüesa-Barreda, G., Know your limits? Climate extremes impact the range of Scots pine in unexpected places. Ann. Bot. 116, doi: /aob/mcv124 Cook, E. R., Kairiūkštis, L.A., Methods of Dendrochronology: Applications in the Environmental Sciences. Kluwer Academic Publishers, Dordrecht, The Netherlands. Cook, E. R., Peters, K., The smoothing spline: a new approach to standardizing forest interior tree-ring width series for dendroclimatic studies. Tree-Ring Bulletin. 41, Eckenwalder, J. E., Conifers of the World: The Complete Reference, Portland, London, Timber Press. Eckstein, D., Bauch, J., Beitrag zur Rationalisierung eines dendrochronologischen Verfahrens und zur Analyse seiner Aussagesicherheit. Forstwissenschaftliches Centralblatt. 88(1): Ensminger, I., Sveshnikov, D., Campbell, D. A., Funks, C., Jansson, S., Lloyd, J., Shibistova, O., Öquist, G., Intermittent low temperatures constrain spring recovery of photosynthesis in boreal Scots pine forests. Global Change Biology. 10, Ericsson, B. Y. A., Larsson, S., Tenow, O., Effects of Early and Late Season Defoliation on Growth and Carbohydrate Dynamics in Scots Pine. Journal of Applied Ecology. 17(3), Friedrichs, D.A., Büntgen, U., Frank, D.C., Esper, J., Neuwirth, B., Loffler, J., Complex climate controls on 20th century oak growth in Central-West Germany. Tree Physiol. 29, doi: /treephys/tpn003 Fritts, H.C., Tree Rings and Climate. Academic Press, New York. Gedminas, A., Outbreaks of Pine Defoliating Insects and Radial Growth, in: IUFRO Kanazawa Forest insect population dynamics and host influences, pp Gedminas, A., Vaivada, S., Spyglius graužiančių kenkėjų įtaka pušų radialiajam prieaugiui. [Effects of pine defoliating insects on radial growth]. Ekologija. (1),

161 Giorgi, F., Climate change hot-spots. Geophysical Research Letters. 33(8), Graham, L. P., Climate change effects on river flow to the Baltic Sea. AMBIO: A Journal of the Human Environment. 33(4), Grissino-Mayer, H. D., Evaluating crossdating accuracy: A manual and tutorial for the computer program COFECHA. Tree-Ring Research. 57(2), Guillemot, J., Klein, E. K., Davi, H., Courbet, F., The effects of thinning intensity and tree size on the growth response to annual climate in Cedrus atlantica: a linear mixed modeling approach. Annals of Forest Science. 72, y Guiot, J., 1991.The bootstrapped response function. Tree-Ring Bulletin. 51, Hacket-Pain, A.J., Friend, A.D., Lageard, J.G.A., Thomas, P.A., The influence of masting phenomenon on growth-climate relationships in trees: Explaining the influence of previous summers climate on ring width. Tree Physiology. 35(3), Hagen, E., Feistel, R., Climatic turning points and regime shifts in the Baltic Sea region: The Baltic winter index (WIBIX) Boreal Environment Research. 10(3), Helama, S., Holopainen, J., Timonen, M., Mielikainen, K., An 854-year tree-ring chronology of scots pine for South-West Finland. Studia Quaternaria. 31(1), Helama, S., Läänelaid, A., Bijak, S., Jaagus, J Contrasting tree-ring growth response of Picea abies to climate variability in western and eastern Estonia. Geografiska annaler: Series A, physical geography. 98(2), HELCOM, Climate change in the Baltic Sea Area: HELCOM thematic assessment in Baltic Sea Environment Proceedings No Hordo, M., Metslaid, S., Kiviste, A., Response of Scots pine (Pinus sylvestris L.) radial growth to climate factors in Estonia. Baltic Forestry. 15(2), Hordo, M., Henttonen, H., Mäkinen, H., Annual growth variation of Scots pine in Estonia and Finland. Baltic Forestry. 17(1), Hökkä, H., Salminen, H., Ahti, E., Effect of temperature and precipitation on the annual diameter growth of Scots pine on drained peatlands and adjacent mineral soil sites in Finland. Dendrochronologia. 30(2), Innes, J.L., The occurrence of flowering and fruiting on individual trees over 3 years and their effects on subsequent crown condition. Trees-Structure and Function. 8(3), Jaagus, J., Kliimamuutuse tendentsid Eestis 20. Sajandi teisel poolel seostatuna muutustega atmosfääri tsirkulatsioonis. [Climate change tendencies in Estonia on the second half of 20 th century in relation to shifts in atmospheric circulations]. Publicationes Instituti Geographici Universitatis Tartuensis, Tartu, 93, Jaagus, J., Climatic changes in Estonia during the second half of the 20th century in relationship with changes in large-scale atmospheric circulation. Theoretical and Applied Climatology. 83(1),

162 Jaagus, J., Mändla, K., Climate change scenarios for Estonia based on climate models from the IPCC Fourth Assessment Report. Estonian Journal of Earth Sciences. 63(3), Kaimre, P., Männipuistute väärtusest ja majandamise tasuvusest [The profitability of pine stand management], in: M. Kurm (Ed.) Mänd Eestis, Vali Press OÜ, Tartu, pp Karjalainen, T., Dynamics and potentials of carbon sequestration in managed stands and wood products in Finland under changing climatic conditions. Forest Ecology and Management. 80(1-3), Kirchhefer, A. J., Reconstruction of summer temperatures from tree-rings of Scots pine (Pinus sylvestris L.) in coastal northern Norway. Holocene. 11(1), / Kiviste, A., Hordo, M., Kangur, A., Kardakov, A., Laarmann, D., Lilleleht, A., Metslaid, S., Sims, A., Korjus, H., Monitoring and modeling of forest ecosystems: the Estonian Network of Forest Research Plots. Forestry Studies. 62, Kjellström, E., Ruosteenoja, K., Present-day and future precipitation in the Baltic Sea region as simulated in a suite of regional climate models. Climatic Change. 81, Kolari, P., Pumpanen, J., Rannik, U., Ilvesniemi, H., Hari, P., Berninger, F., Box, P. O., Carbon balance of different aged Scots pine forests in. Global Change Biology, 10(7), Konôpka, B., Yuste, J. C., Janssens, I. A., Ceulemans, R., Comparison of fine root dynamics in Scots pine and Pedunculate oak in sandy soil. Plant and Soil. 276(1 2), Koprowski, M., Przybylak, R., Zielski, A., Pospieszyńska, A., Tree rings of Scots pine (Pinus sylvestris L.) as a source of information about past climate in northern Poland. International Journal of Biometeorology. 56(1), Kriauciuniene, J., Meilutyte-Barauskiene, D., Reihan, A., Koltsova, T., Lizuma, L., Sarauskiene, D., Variability in temperature, precipitation and river discharge in the Baltic States. Boreal Environment Research. 17(2), Kullman, L., Ground frost restriction of subarctic Picea abies forest in northern Sweden, a dendroecological analysis. Geografiska Annaler. 73A, Laas, E., Dendroloogia [Dendrology]. Valgus, Tallinn. La Bastide, J.G.A., van Vredenburch, C.L.H., The influence of weather conditions on the seed production of some forest trees in the Netherlands. Mede. Bosbouwproefstation, nr Laarmann, D., Korjus, H., Sims, A., Kangur, A., Stanturf, J.A., Initial effects of restoring natural forest structures in Estonia. Forest Ecology and Management. 304, Lebourgeois, F., Gomez, N., Pinto, P., Mérian, P., Mixed stands reduce Abies alba treering sensitivity to summer drought in the Vosges mountains, western Europe. Forest Ecology and Management. 303, Linderholm, H. W., Twentieth-century Scots pine growth variations in the central Scandinavian Mountains related to climate change. Arctic, Antarctic, and Alpine Research. 34(4),

163 Lindholm, M., Lehtonen, H., Kolstrom, T., Merilainen, J., Eronen, M., Timonen, M., Climatic signals extracted from ring-with chronologies of Scots pines from the northern, middle and southern parts of the boreal forest belt in Finland. Silva Fennica. 34(4), Lindner, M., Maroschek, M., Netherer, S., Kremer, A., Barbati, A., Garcia-Gonzalo, J., Seidl, R., Delzon, S., Corona, P., Kolström, M., Lexer, M.J., Marchetti, M., Climate change impacts, adaptive capacity, and vulnerability of European forest ecosystems. Forest Ecology and Management. 259(4), Lind, P., Kjellström, E., Water budget in the Baltic Sea drainage basin: Evaluation of simulated fluxes in a regional climate model. Boreal Environment Research. 14(1), Lõhmus, E., The dependence of Scots pine annual ring indices on some climatic factors. Forestry Studies. 25, Lõhmus, E., Eesti metsa kasvukohatüübid [Estonian forest site types]. Eesti Loodusfoto, Tartu. Läänelaid, A., Eckstein, D., Development of a tree-ring chronology of Scots pine (Pinus sylvestris L.) for Estonia as a dating tool and climatic proxy. Baltic Forestry. 9, Marmor, L., Tõrra, T., Saag, L., Leppik, E., Randlane, T., Lichens on Picea abies and Pinus sylvestris from tree bottom to the top. The Lichenologist. 45(1), Mazza, G., Cutini, A., Manetti, M. C., Site-specific growth responses to climate drivers of Pinus pinea L. tree rings in Italian coastal stands. Annals of Forest Science. 71(8), Metslaid, S., Sims, A., Kangur, A., Hordo, M., Jõgiste, K., Kiviste, A., Hari, P., Growth patterns from different forest generations of Scots pine in Estonia. Journal of Forest Research. 16(3), Metslaid, S., Stanturf, J. A., Hordo, M., Korjus, H., Laarmann, D., Kiviste, A., Growth responses of Scots pine to climatic factors on reclaimed oil shale mined land. Environmental Science and Pollution Research. 23(14), Mäkinen, H., Nöjd, P., Mielikäinen, K., Climatic signal in annual growth variation in damaged and healthy stands of Norway spruce (Picea abies (L.) Karst.) in southern Finland. Trees. (15), Mosteller, F., Tukey, J.W., Data analysis and regression. A Second Course in Statistics. Addison-Wesley Publishing Company, Philippines. Moir, A. K., Leroy, S. A. G., Helama, S., Role of substrate on the dendroclimatic response of Scots pine from varying elevations in northern Scotland. Canadian Journal of Forest Research. 41, Nash, T.H., Kincaid, W.B., Statistical methods for detecting environmental changes, in: Kairiūkštis, L.A., Cook, E.D., (Eds.) Methods of Dendrochronology: Applications in the Environmental Sciences, International Institute for Applied Systems Analysis, Kluwer Academic Publishers, Boston, MA, pp Navasaitis, M., Dendrologija [Dendrology]. Margi Raštai, Vilnius. Niedzẃiedz, T., Glaser, R., Hansson, D., Helama, S., Klimenko, V., Łupikasza, E., Małarzewski, Ł., Nordli, Ø., Przybylak, R., Riemann, D., Solomina, O., A historical time frame

164 (past 1000 years), in: BACC II Author Team, Second assessment of climate change for the Baltic Sea Basin. Springer, Berlin, pp Oberhuber, W., Stumböck, M., Kofler, W., Climate-tree-growth relationships of Scots pine stands (Pinus sylvestris L.) exposed to soil dryness. Trees. 13, Pilcher, J.R., Sample preparation, cross-dating, and measurement, in: Kairiūkštis, L.A., Cook, E.D., (Eds.) Methods of Dendrochronology: Applications in the Environmental Sciences. International Institute for Applied Systems Analysis, Kluwer Academic Publishers, Boston. Pärn, H., Radial growth response of Scots pine to climate under dust pollution in Northeast Estonia. Water, Air, and Soil Pollution. 144(1 4), Pärn, H., Männipuistute radiaalkasvust muutuvates keskkonnatingimustes [The radial growth of pine stands in changing environmental conditions]. Forestry Studies. 48, Pärn, H., Temporal history of relationships between scots pine (Pinus sylvestris L.) radial growth and mean monthly temperatures. Baltic Forestry. 15(1), R Core Team, R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL Raudsaar, M., Pärt, E. Adermann, V., Forest resources, in: Yearbook Forest 2014, Keskkonnaagentuur, Tallinn, pp Reintam, E., Kõlli, R., Muldade klassifikatsioon ja leviku seaduspärasused, in: Astover, A., Kõlli, R., Roostalu, H., Reintam, E., Leedu, E. (Eds.) Mullateadus. Eesti Maaülikool, Tartu, pp Riepšas, E., Rekreacinė miškininkystė. [Recreational forestry], ASU Leidybos centras, Kaunas. Rinn, F., TSAP-Win professional, Time series analysis and presentation for dendrochronology and related applications. Version 0.3. Heidelberg, Germany. Rosenvald, K., Tullus, A., Ostonen, I., Uri, V., Kupper, P., Aosaar, J., Varik, M., Sõber, J., Niglas, A., Hansen, R., Rohula, G., Kukk, M., Sõber, A., Lõhmus, K., The effect of elevated air humidity on young silver birch and hybrid aspen biomass allocation and accumulation acclimation mechanisms and capacity. Forest Ecology and Management 330: Rossi, S., Deslauriers, A., Gričar, J., Seo, J.-W., Rathgeber, C. B., Anfodillo, T.,Morin, H., Levanič, P.O., Jalkanen, R., Critical temperatures for xylogenesis in conifers of cold climates. Global Ecology and Biogeography. 17(6), Rutgersson, A., Jaagus, J., Schenk, F., Stendel, M., Bärring, L., Briede, A., Claremar, B., Hanssen- Bauer, I., Holopainen, J., Moberg, A., Nordli, Ø., Rimkus, E., Wibig, J., Recent Climate Change (Past 200 Years), in: BACC II author team, Second assessment of climate change for the Baltic Sea Basin. Springer Berlin, pp Savva, Y., Denneler, B., Koubaa, A., Tremblay, F., Bergeron, Y., Tjoelker, M.G., Seed transfer and climate change effects on radial growth of jack pine populations in a common garden in Petawawa, Ontario, Canada. Forest Ecology and Management. 242(2 3),

165 Schmitt, U., Jalkanen, R., Eckstein, D., Cambium dynamics of Pinus sylvestris and Betula spp. in the northern boreal forest in Finland. Silva Fennica. 38(2), Sellin, A., Alber, M., Keinänen, M., Kupper, P., Lihavainen, J., Lõhmus, K., Oksanen, E., Sõber, A., Sõber, J., Tullus, A., Growth of northern deciduous trees under increasing atmospheric humidity: possible mechanisms behind the growth retardation. Regional Environmental Change, Sensuła, B., Wilczyński, S., Opała, M., Tree Growth and Climate Relationship: Dynamics of Scots Pine (Pinus Sylvestris L.) Growing in the Near-Source Region of the Combined Heat and Power Plant During the Development of the Pro-Ecological Strategy in Poland. Water, Air, and Soil Pollution. 226: Serreze, M.C., Walsh, J.E., Chapin III, F.S., Osterkamp, T., Dyugerov, M., Romanovsky, V., Oechel, W.C., Morison, J., Zhang, T., Barry, R.G., Observational evidence of recent change in the northern high-latitude environment. Climatic Change. 46(1 2), Sohar, K., Läänelaid, A., Eckstein, D., Helama, S., Jaagus, J., Dendroclimatic signals of pedunculate oak (Quercus robur L.) in Estonia. European Journal of Forest Research. 133(3), Stravinskienė, V., Dendrochronological indication of climatic factors and anthropogenic environmental trends. Lututė, Kaunas. Tarand, A., Jaagus, J., Kallis, A., Eesti kliima minevikus ja tänapäeval [Climate in Estonia in the past and nowadays]. Tartu Ülikooli kirjastus, Tartu. Tardif, J., Camarero, J. J., Ribas, M., Gutierrez, E., Spatiotemporal Variability in Tree Growth in the Central Pyrenees : Climatic and Site Influences. Ecological Monographs. 73(2), Tullus, A., Kupper, P., Sellin, A., Parts, L., Sõber, J., Tullus, T., Lõhmus, K., Sõber, A., Tullus, H., Climate change at northern latitudes: rising atmospheric humidity decreases transpiration, N-uptake and growth rate of hybrid aspen. PLoS ONE 7:e Tuovinen, M., Jalkanen, R., McCarroll, D., The effect of severe ground frost on Scots pine (Pinus sylvestris L.) trees in northern Finland and implications for palaeoclimate reconstruction. Fennia. 183(2) Vaganov, E.A, Hughes, M.K., Kirdyanov, A.V., Schweingruber, F.H., Silkin, P.P., Influence of snowfall and melt timing on tree growth in subarctic Eurasia. Nature. 400(6740), van der Maaten, E., Climate sensitivity of radial growth in European beech (Fagus sylvatica L.) at different aspects in southwestern Germany. Trees-Structure and Function. 26(3), Vanhatalo, A., Chan, T., Aalto, J., Korhonen, J.F., Kolari, P., Ho ltta, T., Nikinmaa, E., Ba ck, J., Tree water relations can trigger monoterpene emissions from Scots pine stems during spring recovery. Biogeosciences. 12, Vitas, A., Tree-Ring Chronology of Scots Pine (Pinus sylvestris L.) for Lithuania. Baltic Forestry. 14(2), Vucetich, J.A., Reed, D.D., Breymeyer, A., Degórski, M., Solon, J., Roo-Zielinska, E., Noble, R., Carbon pools and ecosystem properties along a latitudinal gradient in northern Scots

166 pine (Pinus sylvestris L.) forests. Forest Ecology and Management. 136, Weber, P., Bugmann, H., Rigling, A., Radial growth responses to drought of Pinus sylvestris and Quercus pubescens in an inner-alpine dry valley. Journal of Vegetation Science. 18(6), Wigley, T.M.L., Briffa, K.R., Jones, P.D., On the average value of correlated time series, with applications in dendroclimatology and hydrometeorology. Journal of Applied Meteorology. 23, Wiśniewski, P., Kistowski, M., Structure and importance of soil-protecting forests in the areas administered by the RDSF Toruń.Bulletin fo Geography. Physical Geography Series. 8(8), Zang, C., Biondi, F., Treeclim: An R package for the numerical calibration of proxyclimate relationships. Ecography. 38(4),

167 166

168 167