Needle and branch biomass turnover rates of Norway spruce (Picea abies)

Transcription

1 2517 Needle and branch biomass turnover rates of Norway spruce (Picea abies) P. Muukkonen and A. Lehtonen Abstract: Turnover rates of needle and branch biomass, number of needle cohorts, and needle-shed dynamics were modelled for Norway spruce (Picea abies (L.) Karst.) in southern Finland. Biomass turnover rates, vertical distribution, and biomass of the branches were modelled simultaneously. The rate of needle turnover was determined from needle-shed dynamics. The potential litterfall of branches was modelled by combining the vertical distribution of branch biomass and the annual change in height of the crown base. The mean annual turnover rates for needle and branch biomass are 0.10 and , respectively. At the age of 5.5 years, 50% of the needles in the needle cohort have been shed. In addition, at the age of 12 years, all needles of the needle cohort have been shed. Turnover of branch biomass was dependent on stand density and tree size. The modelled rates of biomass turnover agreed with measurements of needle and branch litterfall. Many process- or inventory-based models use a single turnover rate for branch litterfall based on literature, and some of the models are fully ignoring the litterfall of branches. Species-specific turnover rates or dynamic litterfall models should be applied when carbon flows in forest stands are modelled. Résumé : Le taux de renouvellement des aiguilles et de la biomasse des branches, le nombre de cohortes d aiguilles et la dynamique de la perte des aiguilles ont été modélisés chez l épicéa commun (Picea abies (L.) Karst.) dans le sud de la Finlande. Le taux de renouvellement de la biomasse, la distribution verticale et la biomasse des branches ont été modélisés simultanément. Le taux de renouvellement des aiguilles a été déterminé à partir de la dynamique de la perte des aiguilles. La litière de branches potentielle a été modélisée en combinant la distribution verticale de la biomasse des branches et la variation annuelle de la hauteur de la base de la cime. Les taux annuels moyens de renouvellement sont respectivement de 0,10 et 0,0125 pour la biomasse des aiguilles et celle des branches. À l âge de 5,5 ans, 50 % des aiguilles de la cohorte d aiguilles étaient disparues. De plus, à l âge de 12 ans, toutes les aiguilles de la cohorte d aiguilles étaient disparues. Le renouvellement de la biomasse des branches dépendait de la densité du peuplement et de la dimension des arbres. Les taux de renouvellement de la biomasse obtenus par modélisation correspondaient aux mesures de litière d aiguilles et de branches. Plusieurs modèles basés sur les processus ou sur des relevés utilisent un taux unique de renouvellement basé sur la littérature pour la litière de branches et quelques modèles ignorent complètement la litière de branches. Des taux de renouvellement propres à chaque espèce ou des modèles de la dynamique de la litière devraient être appliqués pour modéliser les flux de carbone dans les peuplements forestiers. [Traduit par la Rédaction] Muukkonen and Lehtonen 2527 Introduction Litterfall represents the most important source of element flux to the forest floor. Fluxes of litterfall depend on several ecological factors, for example, species, climate, site quality, stand increment, stand age, stand density, and thinning (Pedersen and Bille-Hansen 1999). Holstener-Jørgensen et al. (1979) made an extensive literature review and concluded that in boreal conditions the total amount of needle litter in Norway spruce stands varies from 1500 to 4500 kg ha 1 year 1. Amounts of branch litter reported for Norway spruce in boreal forests vary from 100 to 500 kg ha 1 year 1 (Viro 1955; Nilsson and Wiklund 1992). In addition to large spatial variation in litterfall, the annual variation is also large (Bille-Hansen and Hansen 2001). Received 21 January Accepted 11 August Published on the NRC Research Press Web site at on 14 January P. Muukkonen 1 and A. Lehtonen. Finnish Forest Research Institute, P.O. Box 18, FIN Vantaa, Finland. 1 Corresponding author ( petteri.muukkonen@metla.fi). The proportion of aboveground litter compartments of Norway spruce (Picea abies (L.) Karst.) is nearly 73% for needles, 13% for branches, 5% for cones, and 10% for other mixed litter (Viro 1955), which consists of seed, flowers, bud scales, epiphyte lichen, and small pieces of bark. Although the amount of branch litterfall is much lower than that of foliage litter, its contribution to the carbon stock of the soil is high, since it decomposes slowly; this should be taken into account when ecosystem models are built. Relatively few field measurements are available for large branch litter of Norway spruce, while twigs are more often reported with needle litter. Therefore estimations of branch litter fluxes are based on measured conifer stands reported in ecosystem study compilations like those of Reichle (1981) and Cannel (1982). The conifer crown is a population of shoots, each developed from a branch of the main stem of the tree (Schoettle and Fahey 1994). The needles in internodes of the same age form a needle cohort (Fig. 1), and a new needle cohort is produced annually on the apices of the stem leader and the leader shoots of the branches (Jalkanen 1998). The great variability in number of needles in the needle cohort indicates that the life-span of the needles is limited, Can. J. For. Res. 34: (2004) doi: /X04-133

2 2518 Can. J. For. Res. Vol. 34, 2004 Fig. 1. Needle cohorts. First-order needle cohorts are located on the main stem of the branch. that is, over time, the number of needles in the needle cohort decreases because of aging of the needles (Ross et al. 1986). During the first 2 years, the needle mortality is negligible. Thereafter, the number of needles from a particular year decreases drastically. It has been suggested that the retention of needles throughout the year is a mechanism for nutrient conservation and offers the potential for a gain in photosynthetic carbon gain during favourable periods in fall, winter, or early spring at times when deciduous species are leafless (Schoettle and Fahey 1994). In coniferous trees, where as much as 70% of the canopy can consist of old (>1 year) needles, needle lifespan is directly influenced by the growth environment and the distribution of resources within the canopy (Balster and Marshall 2000). Therefore, changes in needle longevity could affect the carbon balance, the rate of nutrient cycling, and ultimately, the net primary production of a forest ecosystem. Norway spruce does not shed needles one needle cohort at a time (Ross et al. 1986; Salemaa et al. 1993). This species sheds needles from several of the oldest needle cohorts simultaneously. However, the long-term mean amount of annual needle shed is about what it would be if one needle cohort were shed at a time, but in declining trees the annual needle shed is clearly greater than the amount of one needle cohort. In Norway spruce, needle shed is not concentrated at the end of the growing season; rather, the litterfall of Norway spruce needles seems to be distributed uniformly over the year (Viro 1955; Salemaa et al. 1993). As much as 60% of the total annual needle shed of Norway spruce occurs during the growing season (Mork 1942; Viro 1955; Holstener- Jørgensen et al. 1979). Owing to differences in weather from year to year, needles of Norway spruce fall at different times in different years (Viro 1955). Norway spruce is a shade-tolerant species, which can be seen from the canopy shape and the branch longevity. The life-span of the branches is longer in this species than in pioneer conifers like Scots pine (Pinus sylvestris L.) (Kärkkäinen 2003). This faster turnover of branch biomass can also be estimated based on the published data like crown ratios, branch biomass models, and height development. Firstly, the average crown ratio (crown length per tree length) of Norway spruce trees is 0.76, while it is 0.62 for Scots pine in Finland (Hynynen et al. 2002). Seconly, the branch biomass of Norway spruce is almost twice as much as branch biomass of equal-sized Scots pine (Marklund 1988). Thirdly, the average height increment is lower with Norway spruce (Hynynen et al. 2002). And fourthly, it can be assumed that in both of these species the crown base height follows equally the tree height. By summarizing these assumptions, it can be assumed that the annual turnover rate of the branch biomass of Norway spruce is clearly lower than that for Scots pine and other pioneer conifers. Because the branch mortality and litterfall of Norway spruce have not been studied as comprehensively, more information on their biomass turnover is needed. Because of limited information about the litter flux of spruce needles to the soil, and especially that of spruce branches, an alternative method was developed for assessing needle and branch litter flux. To estimate average turnover rate of needle biomass and potential branch litterfall, this study combined inventory variables and biomass measurements. The objective of this study was to estimate the litterfall of the needles and branches of Norway spruce. The aims are to estimate the rate of needle turnover determined from needleshed dynamics and to estimate the potential litterfall of branches modelled by combining the vertical distribution of branch biomass and the annual change in height of the crown base. To understand carbon cycle and flows of forests, information on litter production by compartment is needed. Material and methods Data In this study, two sources of data were used: (1) sample trees and branches from the national tree research Valtakunnallinen Puututkimus (VAPU) and (2) height to crown base data from the National Forest Inventory permanent plots. VAPU data and National Forest Inventory data are independent data sets. The VAPU data used in this study consisted of measurements of sample trees on sample plots established by the Finnish Forest Research Institute in southern Finland (south of 62 4 N) during (Korhonen and Maltamo 1990). Three to six sample trees (with dbh of more than 5 cm) from the dominant canopy layer closest to the plot centre were selected and felled (Fig. 2). From 94 sample plots, a total of 196 and 80 Norway spruce trees were analysed for estimation of branch litterfall and needle litterfall, respectively. Estimation of needle litterfall is based on needle cohort longevity (VAPU database). For these data, first-order needle cohorts (Fig. 1) were estimated visually from two branches in the 15th whorl from the top of the tree (Fig. 2a). The first branch pointed to the centre of the sample plot, and the second pointed in the opposite direction (Fig. 2b). Kendall s coefficient of concordance (Ranta et al. 1999) shows that there are statistically significant similarities between the needle cohorts of the two measured directions. Therefore to avoid

3 Muukkonen and Lehtonen 2519 Fig. 2. Needle cohorts were estimated visually from the two branches in the of 15th whorl from the tree top. The first branch pointed to the centre of the sample plot and the second one pointed in the opposite direction. Fig. 3. Estimation of potential litterfall of branches (B t /B tot ), based on vertical distribution of branch biomass and annual change in height of the crown base. measurements that are dependent on each other, it is reasonable to analyse the measurements of branches in only one direction. The percent survival of needles in each of the needle cohorts was estimated visually and classified into one of six classes: (1) 0% 5%, (2) 6% 25%, (3) 26% 50%, (4) 51% 75%, (5) 76% 95%, and (6) 96% 100%. The diameter of every branch on the sample trees was measured. The sample branches were selected randomly with probability proportional to diameter. The sum of diameters was divided by 10 (denote r), and then a random integer between 1 and r was selected, and that integer indicated the location of the first sample branch according to the cumulative diameter sum starting from tree top. Sample branches were then selected at an interval of r; and 10 branches per tree were chosen. Furthermore, the dry masses of the second, fifth, and eighth sample branches were determined in the laboratory, while others were measured only for fresh mass. To have a more precise model for branch biomass, all branches of the living crown that were less than 7.5 mm in diameter were excluded, and those branches constitute less than 1% of total branch biomass. The height to the crown base was measured on permanent sample plots established in and remeasured in 1995 by the National Forest Inventory. The subsample of permanent plots used in this study is located below 62 N and consists of 267 plots, including a total of 782 measured Norway spruces. In 1985 and 1995, all trees from the plots were measured for dbh, height to crown base, and tree height. The annual change in height to crown base was derived from the difference between measurements in 1985 and 1995 (Fig. 3). The crown base was defined as the lowest whorl with at least one living branch, separated from the other living whorls above by no more than one dead whorl. Statistical analyses and modelling Needles To study needle-shed dynamics and to estimate the turnover rate of needle biomass, ordinal regression (Bender and Benner 2000) was used to model the relationship between age of the needle cohort and the survival class (Table 1). The survival class (Y) is a categorical response variable with k + 1 (6) ordered categories. The Y is a discretized variable of an underlying latent continuous trait defined by cut-off points j. It is then natural to formulate a model by means of the cumulative probabilities γ j. Let π j (x) =P(Y = j X = x) bethe probability for realization of Y = j given X = x, j = 0, 1,, k. The class of grouped continuous models is based upon the cumulative probabilities [1] γ j (x) =P(Y j X = x) =π j (x), j = 1,, k The class of grouped continuous models is obtained by the generalized linear model [2] f[π(x)] = α + βx in which the cumulative probabilities are used instead of π. [3] f[γ j (x)] = α j + β j x, j = 1,, k where f is an appropriate function, α is the intercept, and β is the regression coefficient for X. The standard assumption in

4 2520 Can. J. For. Res. Vol. 34, 2004 Table 1. Number and proportion (%) of observations according to survival class and age of needle cohort. Age of needle cohort (years) Total Class boundaries (%) Survival class % 59% 36% 25% 13% 5% 0% 0% 0% 0% 0% 0% % 25% 36% 36% 39% 21% 11% 3% 1% 0% 0% 0% % 6% 18% 18% 16% 23% 20% 15% 5% 0% 0% 0% % 1% 9% 9% 10% 19% 28% 20% 11% 3% 1% 1% % 1% 5% 5% 10% 13% 13% 19% 15% 18% 4% 1% % 8% 8% 8% 13% 20% 29% 44% 68% 80% 95% 98% Total % 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% most applications is that the regression coefficient does not depend on j; that means the model [4] f[γ j (x)] = α j + βx, j = 1,, k is considered. Thus, it is assumed that for the considered link function f the corresponding regression coefficients are equal for each cut-off point j. Since the logit link [5] f(π) = ln[π/(1 π)] is used, the generalized linear model becomes γ () α j x j + e [6] ln = αj + βx γj() x = αj 1 γ () x 1 + e j ( βx) ( + βx) From the model, we determined where the cumulative probabilities of ordinal regression reached its 50% limit. Two approaches used to study survivorship are as follows: (1) the cohort approach and (2) the time-period approach (Fleming and Piene 1992a, 1992b). Fleming and Piene (1992b) concluded that, when survival rates are stationary, these approaches produce similar estimates. The cohort approach involves following a group of individuals born simultaneously through time (Fleming and Piene 1992b). The time-period approach uses data on the age structure of the population at a specific period of time to infer survival rates (Fleming and Piene 1992a). In this study, we applied the period approach to the data from national tree research (VAPU). We characterized the decrease in needle survival over the course of time by percent survival according to the age of the needle cohort. The needle percent survival indicates the proportion of original needles present in a needle cohort at a particular time. Since, the needle survival data are provided in classes, we converted those survival classes to percent survival using random numbers within the class boundaries. We made 1000 simulations. The dry mass of living needle increases during the first 4 years (Viro 1955). The mass of second-year needles, thirdyear needles, and needles older than that is 36%, 30%, and 40%, respectively, higher than that of first-year needles. Norway spruce sheds needles from all needle cohorts, and most of the needles will become yellow before they are shed (Salemaa et al. 1993). Upon yellowing, the spruce needles become lighter, and the absolute amounts of nutrients in them usually diminish, being transferred to the trunk (Viro 1955). In other words, a substantial amount of the nutrients required for construction of new needles each year can be supplied by the relocation of nutrients from aging needles (Schoettle and Fahey 1994). In this process spruce needles lose 13% 39% of their mass, depending on the age of the needle cohort (Viro 1955). The turnover rate of needle biomass (NL) in the timeperiod approach was calculated separately for each simulation and the model as follows [7] NL = n 1 i= 0 ( b b ) m d i i+ 1 i i n 1 i= 0 ( b m ) i i

5 Muukkonen and Lehtonen 2521 Table 2. Description of litterfall collection stands. Avg. annual branch litter (kg ha 1 ) Avg. annual needle litterfall (kg ha 1 ) Estimated branch biomass (kg ha 1 ) b Estimated needle biomass (kg ha 1 ) b Measurement year of stand variables Stocking (no. ha 1 ) Basal area (m 2 ha 1 ) No. of litter traps Beginning of litter collection Stand age in 1998 (years) Location a Stand Aulanko 163 SF Heinola 145 SF Kittilä 200 NF Kuorevesi 138 SF Rovaniemi 155 NF Siilinjärvi 99 SF a SF, southern; NF, northern Finland. b Biomass of needles and branches was estimated as a function of dbh and height of trees by applying the biomass equations of Marklund (1988). where b is the percent survival of the needle cohort, m is a mass factor indicating weighting of single needles over time, and d indicates loss of mass during yellowing of needles. The numerator indicates the total amount of needles removed annually, and the denominator indicates the total amount of needles on a tree or single branch. In this study n is 12, which indicates the number of age classes. Branches The potential branch litterfall was estimated according to the method described by Lehtonen et al. (2004b). The dry mass of each branch (excluding foliage) as a function of branch diameter was modelled with a mixed linear model based on the VAPU database. The mixed model approach was justified by the correlated observations; in the sample, branches from the same tree are correlated with each other. The tree and branch levels were taken into account in the random part of the model. The dry mass of branch i on tree k (m ki ) was modelled as the following function of branch diameter (d ki ) [8] ln m ki (d) =lna 0 + A 1 [ln(d ki )] lna 0k + a 1k [ln(d ki )] lnε ki where A 0 and A 1 are fixed population parameters, while a 0k and a 1k are random tree parameters with zero expectations, which were estimated by the restricted maximum likelihood method in a mixed procedure (SAS Institute Inc. 1999). Before fitting the mixed model, branch diameter (d ki ) were transformed to the power of This was done after examining residual figures of diameter mass relationship of data. The transformation was done by nlin procedure (SAS Institute Inc. 1999). The error term (lnε ki ) of the mixed model was assumed to be distributed normally. The vertical distribution of branch biomass in tree crowns was modelled based on estimates of branch biomass. The live tree crowns were divided into 10 segments of equal relative length from the base to the top of the crown (0% 10%, 11% 20%,, 91% 100%). Thereafter, the proportion of total branch biomass for each segment was estimated and used when the nonlinear model for vertical branch biomass distribution was fitted. Then, the distribution of branch biomass (s) as a function of relative height (h) and the crown ratio (cr) is [9] s(h,cr) =(a 0 + a 1 cr) (h 1) +(b 0 + b 1 cr) (h 2 1)+(c 0 + c 1 cr) (h 3 1)+ε where a 0, a 1, b 0, b 1, c 0, and c 1 are parameters, and ε is the error term. The relative height within the tree crown equals zero at the crown base and one at the top of a tree. The integral in eq. 9 was not constrained to be equal to one, but being close to 0.1 because of 10% segments. The parameters of the function were estimated by using the Gauss Newton method in the nlin procedure (SAS Institute Inc. 1999). The height of the crown base was measured from sample trees on permanent sample plots in 1985 and The annual change in the height of the crown base was estimated to be one-tenth the change between 1985 and The branch biomass gained by growth between crown base change mea-

6 2522 Can. J. For. Res. Vol. 34, 2004 Table 3. Description of published litterfall studies. Reference Location No. of sites Age of stand (years) Needle litter data available Branch litter data available Biomass data available Mork 1942 Norway Y Y Viro 1955 Finland Y Y Y Bonnevie-Svedsen and Gjems 1957 Norway Y Y Nilsson and Wiklund 1992 Sweden 6 Y Y Y Note: Y, data available. Stand volume data available Table 4. Cumulative probabilities of needle cohorts according to survival class and age of the needle cohort modelled by ordinal regression. Age of needle cohort (years) Survival class Class boundaries (%) a 0.54 a a 0.59 a a a a a Diagonal values indicate the 50% probability limit. surements was assumed to be minor and was therefore ignored. The distribution of vertical branch biomass in 1985 was estimated for each sample tree by applying the previous model. By using vertical biomass distribution, the amount of biomass lost because of annual increase in the crown height (h 1 ) was calculated (Fig. 3). The amount of annually lost biomass was proportioned to the total biomass distribution of the branches using [10] B r = h sh (, cr)dh sh (, cr)dh which gave an estimate of the proportion of potential branch litter (B r ) for each of the 782 trees on National Forest Inventory sample plots. With measurements of the height of the crown base, a nonlinear regression model was developed for the proportion of biomass lost annually as litter because of the rise of the crown base in each tree. Tree size and stand density were included as independent variables. Other variables like fertility, stand age, and basal area were tried to predict annual biomass loss, but stand density was found to be superior. Thereafter the proportion of potential branch litter in the branch biomass (B r ) was modelled as a function of tree diameter (dbh) and stand density (n) 2 [11] B (dbh, n) = ( a + a n) e 0 + c+ r [(b+ b n) dhb ] 0 ε where a, a 0, b, b 0, and c are parameters, and ε is the error term. These parameters were estimated by using the Gauss Newton method in the nlin procedure (SAS Institute Inc. 1999). Comparison with litterfall from litter-trap studies The models developed were compared with: (1) long-term littertrap data of the Finnish Forest Research Institute (Metla) managed by Mr. T. Hokkanen (Table 2) and (2) the litterfall reported in previous studies in Sweden, Norway, and Finland (Table 3). Needle and branch litterfall was collected from six stands throughout Finland (Table 2). The size of the litter traps was 0.5 m 2. Kouki and Hokkanen (1992) describe the collection of litterfall in more detail. The proportion of the litter potential made up of the total needle biomass and the total branch biomass for the litter-collection stands was estimated by dividing the estimated biomass for each litter-collection stand by the average annual litterfall. The average annual needle and branch litterfall for these stands was assessed starting in 1960 (Table 2). The needle biomass of these stands was estimated by applying biomass equations based on dbh, height, and crown length, and branch biomass was estimated by applying biomass equations based on dbh and height to the latest tree-level data (Marklund 1988). The reported litterfall figures were compared with estimates based on the present study. Reported biomasses were multiplied by biomass turnover rate produced in this study. When biomass data were not available, foliage biomass and branch biomass were subtracted from the stand volume (by biomass expansion factor of Lehtonen et al. 2004a). These functions are based on biomass and volume equations that were applied on tree-wise data from permanent sample plots of the Finnish national forest inventory. Functions can be applied to forests less than 250 m 3 ha 1. For stands with tree volumes greater than that, volumes of 250 m 3 ha 1 were used. Litter-trap estimates and model approach estimates of annual litterfalls by compartment were analysed by a pairwise Student s t test and by its nonparametric counterpart, a pairwise Wilcoxon test.

7 Muukkonen and Lehtonen 2523 Fig. 4. Probability and 95% confidence limits of the ordinal regression of the cohort approach. Survival classes on the y-axes represent the percent survival of needles in each of the needle cohorts. Survival classes were estimated visually and classified into six classes: (1) 0% 5%, (2) 6% 25%, (3) 26% 50%, (4) 51% 75%, (5) 76% 95% and (6) 96% 100%. The size of the bubble indicates the probability that the needle cohort of curtain age belongs to the specific survival class. Fig. 5. Predicted survival percentages with 95% confidence limits of needle cohorts in the cohort approach, according to age of the needles. The model of the present study is compared with the models of Flower-Ellis and Mao-Sheng (1987) and Niinemets and Lukjanova (2003). Results and discussion Needles It was found that with a probability of 71%, the youngest observed needle cohort (1 year old) belongs to class 5 (Table 4 and Fig. 4), which indicates survival of 96% 100%. At a probability of 90%, the 12-year-old needle cohort had shed 95% 100% of all its needles. When the survival class model of the ordinal regression is expanded by random simulations, we construct a percentsurvival model (Fig. 5). At the age of 5.5 years, 50% of the needles of that needle cohort have been shed. In addition, when the needles are 12 years old, all needles of the needle cohort are shed. Needles are shed most rapidly when they are 6 8 years old (Fig. 5). The results show that Norway spruce does not shed needles one needle cohort at a time. Rather, it sheds needles simultaneously from several of the oldest needle cohorts, which confirms the conclusions of Salemaa et al. (1993). The dynamics of needle shed found in this study is similar to that reported by Niinemets and Lukjanova (2003) for an Estonia stand (Fig. 5). The dynamics for northern Sweden reported by Flower-Ellis and Mao- Sheng (1987) differs markedly, because of the longer needle longevity on northern sites. The biomass turnover rates of the random simulations vary from 0.07 to 0.13 when the weighting and yellowing effects of needles are taken into account. Both the arithmetic mean and the median show that the turnover rate of needle biomass for Norway spruces in southern Finland is Branches Branch biomass was modelled according to branch diameter, and estimates were calibrated for each sample tree by a

8 2524 Can. J. For. Res. Vol. 34, 2004 Table 5. Needle and branch biomass turnover rates of Norway spruce (standard deviation in parentheses). Modelled Southern Finland Northern Sweden Estimated on the basis of measurements Needles 0.10 (0.0126) 0.14 a 0.05 b (0.0036) 0.08 a,b 0.11 c Branches d a Without weighting and yellowing effects. b Calculated from needle-shed dynamics reported by Flower-Ellis and Mao-Sheng (1987). c Estimated turnover rate was determined from the amounts of litter and biomass (Mork 1942; Viro 1955; Bonnevie- Svedsen and Gjems 1957; Nilsson and Wiklund 1992; Finnish measurements). d Estimated turnover rate was determined from the amounts of litter and biomass (Viro 1955; Nilsson and Wiklund 1992; Finnish measurements). Table 6. Parameter estimates with approximated standard errors (SE) for the model of branch biomass distribution (eq. 9). Parameter Estimate Approximated SE a a b b c c Note: No. observations = 1858; SS error = ; SS total = Fig. 6. Models for potential turnover rate of branch litter as a function of diameter, when stocking varies from 500 to 2500 trees ha 1. Table 7. Parameter estimates with approximated standard errors (SE) for potential branch litter model (eq. 11). Parameter Estimate Approximated SE a a b b c Note: No. observations = 782; SS error = ; SS total = mixed model. Thereafter, estimates of branch biomass were used in modelling the vertical biomass distribution of the branches. Most of the variation in biomass distribution of the vertical branches was explained by relative height on the tree and by the crown ratio (Table 6). The combination of change in the crown base and the vertical distribution of biomass provided an estimate for potential branch litter. This relative potential branch litterfall was shown to be dependent on tree dimensions and stand density. Therefore, a nonlinear model was constructed in which relative potential branch litterfall is modelled as a function of dbh and stand density (stems per hectare) (Table 7; Fig. 6). A similar effect of density to height at the crown base has also been noted in studies of timber quality, where the height of the lowest living branches has been studied with different stockings (Madgwick et al. 1986; Johansson 1992). The average potential branch litterfall was 1.25% of the total branch biomass of Norway spruce. It was found that the relative potential branch litterfall is less than half of the potential litterfall for pines, which for Scots pines was 2.7% (Lehtonen et al. 2004b). Therefore, when carbon fluxes of spruce forests in southern Finland are estimated, the model (eq. 11) with dbh and stocking density or a constant turnover rate of for branches should be applied (Table 7). These estimates are based on average change in height of the crown base during 10 years and because of high interannual variation in litterfall are therefore not comparable to litter measurements from a single year. If silvicultural practises change and density regimes are affected, the average turnover rate could be biased for new conditions, while presented model (eq. 11) takes these changes with stocking density into account. Our model estimates only the branch death in the lower crown but does not address small branch and twig death from the interior areas of the crown. This source of branch litter plays only a minor role in the total branch litterfall and is often reported with litter-trap studies. Also our modelling approach is up-scaling tree-wise measurements of canopy dynamics to stand level; the up-scaling is based on representative sampling of branch biomass and crown base height development for southern Finland. Therefore, we consider our results to be applicable within the region of southern Finland, while the models and idea could be applied after reparameterization for any other forest region. The litterfall of large branches is very seldom measured. Several ecosystem- and soil-carbon models (Wang et al. 2001; Yarie and Billings 2002; Komarov et al. 2003; Masera et al.

9 Muukkonen and Lehtonen 2525 Fig. 7. Comparison of modelled and measured (a) needle litterfall and (b) branch litterfall. 2003; Paul et al. 2003) are estimating branch litter as a constant ratio, meanwhile some studies are fully ignoring branch litterfall because of lack of data. According a literature review, values for branch litterfall for Norway spruce in Scandinavia varied between 100 and 600 kg ha 1 (Fig. 7). Being such a significant litter source, it should not be ignored, and some effort should be made to catch the correct level and the dynamics of the branch litterfall. Thereafter, the combination of models is justified when there is a lack of suitable litter-trap studies. The combination of methods, like branch mortality dating (Maguire 1994), biomass models of living and dead branches (Marklund 1988), and litter-trap measurements (Kouki and Hokkanen 1992) could be used for branch turnover rate assessments. Comparison with measured litterfall Our results show that the rates of biomass turnover calculated in this study provide estimates of litterfall that are similar to measured amounts of litterfall (Fig. 7). Both the Student s t test (t = 1.512, p = 0.146) and the Wilcoxon test (Z = 1.408, p = 0.159) for paired comparisons show that there is no statistically significant difference between measurements and estimations of needle litterfall (Fig. 7a). Nor does branch litterfall differ statistically significantly between measurements and estimations (t = 0.315, p = 0.758; Z = 1.022, p = 0.307) (Fig. 7b). These results indicate that, when there is a need to estimate the amount of branch and needle litterfall, the rates of biomass turnover found in this study are relevant and useful estimators. In Swedish data on litterfall measurement, Norway spruce needles have a significantly higher rate of biomass turnover, and therefore our estimations give lower values than measured litterfall does (t = 3.058, p = 0.022; Z = 1.99, p = 0.046). In Norway (t = 1.327, p = 0.233; Z = 0.734, p = 0.463) and Finland (t = 0.001, p = 0.999; Z = 0.059, p = 0.953), there are no statistically significant differences between predictions and measurements. When litterfall of branches is compared for Swedish data, our model estimates significantly overestimate branch litterfall (t = 6.137, p = 0.001; Z = 2.201, p = 0.028). In Finland, there is no significant difference between estimated and measured branch litterfall (t = , p = 0.620; Z = 0.178, p = 0.859). Litterfall collections and measurements represent situations on selected forest sites. In many cases, stand level branch litterfall measurements consist of only smaller branches and twigs. The litter measurements are normally made by litter collectors, which do not collect larger branches. We assumed that comparisons using Finnish litterfall measurements and the reported values of Nilsson and Wiklund (1992) would be more reliable since, in Finnish data, the biomass equations are used to tree-wise data, while Nilsson and Wiklund (1992) give pre-estimated biomass values. For other sources (Mork 1942; Viro 1955; Bonnevie-Svedsen and Gjems 1957) the biomass values were calculated from stand volume, which in comparisons might give distorted results. When modelled and measured litterfall were compared for the whole of Scandinavia, we succeeded in predicting litterfall satisfactorily. Both, the litter measurements and Finnish data used for modelling needle biomass turnover rate, consist of only few study sites. Therefore, broader discussion about the causes of differences is not desirable. Applicability of results Species-specific estimation of litter production by compartment is essential for understanding the carbon cycle and flows of forests. In studies concerning the carbon balance of forests, rates of biomass turnover are usually estimated from litterfall measurements, or inverse number of the maximum number of needle cohorts are used. In some cases, rates of biomass turnover for other tree species are applied to Norway spruce. For example, if the rates of biomass turnover for Scots pine are applied to Norway spruce, we overestimate the input of needle and branch litter to the soil system. The results of this study can be used in modelling the carbon balance of boreal Norway spruce forests. Biomass turnover

10 2526 Can. J. For. Res. Vol. 34, 2004 rates derived in this study are applicable as average values for large areas. Our modelled rates of biomass turnover for needles and branches are suitable for conditions corresponding with those in southern Finland. Turnover rates for northern Sweden calculated from needle-shed dynamics reported by Flower-Ellis and Mao-Sheng (1987) are suitable for conditions corresponding with those in northern Finland. Acknowledgements The authors thank the Academy of Finland for financing Integrated method to estimate carbon budgets of forests (project No ), which is part of the Research Programme on Sustainable Use of Natural Resources (SUNARE). We are also grateful to the National Forest Inventory for providing data for the national tree research (VAPU) and to Mr. Tatu Hokkanen for providing data on litterfall measurement. The authors also thank Dr. Raisa Mäkipää and Mr. Mikko Peltoniemi for their comments on the manuscript, Dr. Joann von Weissenberg for checking the English language of the article, and Kati Liukkonen for drawing Figs. 1, 2, and 3. References Balster, N.J., and Marshall, J.D Decreased needle longevity of fertilized Douglas-fir and grand fir in the Northern Rockies. Tree Physiol. 20: Bender, R., and Benner, A Calculating ordinal regression models in SAS and S-Plus. Biom. J. 42: Bille-Hansen, J., and Hansen, K Relation between defoliation and litterfall in some Danish Picea abies and Fagus sylvatica stands. Scand. J. For. Res. 16: Bonnevie-Svedsen, C., and Gjems, O Amount and chemical composition of litter from larch, beech, Norway spruce and Scots pine stands and its effect on the soil. Medd. Nor. Inst. Skogforsk. 14: Cannel, M.G.R World forest biomass and primary production data. Academic Press Inc., London. p Fleming, R.A., and Piene, H. 1992a. Spruce budworm defoliation and growth loss in young balsam fir: period models of needle survivorship for spaced trees. For. Sci. 38: Fleming, R.A., and Piene, H. 1992b. Spruce budworm defoliation and growth loss in young balsam fir: cohort models of needlefall schedules for spaced trees. For. Sci. 38: Flower-Ellis, J.G.K., and Mao-Sheng, Y Vertical and cohort life-tables for needles of Norway spruce from northern Sweden. Sver. Lantbruksuniv. Inst. Skogsekol. Skoglig Marklara Rapp. 57: Holstener-Jørgensen, H., Veracion, V.P., and Yao, C.E Litterfall studies in an irrigation trial in Norway spruce. Forstl. Forsogsvaes. Dan. 36: Hynynen, J., Ojansuu, R., Hökkä, H., Siipilehto, J., Salminen, H., and Haapala, P Models for predicting stand development in MELA System. Finn. For. Res. Inst. Res. Pap Jalkanen, R Fluctuation in the number of needle sets and needle shed in Pinus sylvestris. Scand. J. For. Res. 13: Johansson, K Effects of initial spacing on the stem and branch properties and graded quality of Picea abies (L.) Karst. Scand. J. For. Res. 7: Kärkkäinen, M Puutieteen perusteet. Metsälehti Kustannus, Hämeenlinna, Finland. [In Finnish.] Komarov, A., Chertov, O., Zudin, S., Nadporozhskaya, M., Mikhailov, A., Bykhovets, S., Zudina, E., and Zoubkova, E EFIMOD 2 amodel of growth and cycling of elements in boreal forest ecosystems. Ecol. Model. 170: Korhonen, K.T., and Maltamo, M Männyn maanpäällisten osien kuivamassat Etelä-Suomessa. Finn. For. Res. Inst. Res. Pap. 371: [In Finnish.] Kouki, J., and Hokkanen, T Long-term needle litterfall of a Scots pine Pinus sylvestris stand: relation to temperature factors. Oecologia, 89: Lehtonen, A., Mäkipää, R., Heikkinen, J., Sievänen, R., and Liski, J. 2004a. New approach to formulate biomass expansion factors (BEF) by stand age for Scots pine, Norway spruce and birch. For. Ecol. Manage. 188: Lehtonen, A., Sievänen, R., Mäkelä, A., Mäkipää, R., Korhonen, K.T., and Hokkanen, T. 2004b. Potential litterfall of Scots pine branches in southern Finland. Ecol. Model. 180: Madgwick, H.A.I., Tamm, C.O., and Fu, M.Y Crown development in young Picea abies stands. Scand. J. For. Res. 1: Maguire, D.A Branch mortality and potential litterfall from Douglas-fir trees in stands of varying density. For. Ecol. Manage. 70: Marklund, L.G Biomassafunktioner för tall, gran och björk i Sverige. Sverig. Lantbruksuniver. Inst. Skogstaxering Rapp. 45: [In Swedish.] Masera, O.R., Garza-Caligaris, J.F., Kanninen, M., Karjalainen, T., Liski, J., Nabuurs, G.J., Pussinen, A., de Jong, B.H.J., and Mohren, G.M.J Modelling carbon sequestration in afforestation, agroforestry and forest management projects: the CO2FIX V.2 approach. Ecol. Model. 164: Mork, E Om strøfallet i våre skoger. Medd. Nor. Inst. Skogforsk. 29: [In Norwegian.] Niinemets, Ü., and Lukjanova, A Needle longevity, shoot growth and branching frequency in relation to site fertility and within-canopy light conditions in Pinus sylvestris. Ann. For. Sci. 60: Nilsson, L.-O., and Wiklund, K Influence of nutrient and water stress on Norway spruce production in south Sweden the role of air pollutants. Plant Soil, 147: Paul, K.I., Polglase, P.J., and Richards, G.P Predicted change in soil carbon following afforestation or reforestation, and analysis of controlling factors by linking a C accounting model (CAMFor) to models of forest growth (3PG), litter decomposition (GENDEC) and soil C turnover (RothC). For. Ecol. Manage. 177: Pedersen, L.B., and Bille-Hansen, J A comparison of litterfall and element fluxes in even aged Norway spruce, sitka spruce and beech stands in Denmark. For. Ecol. Manage. 114: Ranta, E., Rita, H., and Kouki, J Biometria. Helsinki University Press, Helsinki. [In Finnish.] Reichle, D.E. (Editor) Dynamic properties of forest ecosystems. Cambridge University Press, Cambridge. Ross, J., Kellomäki, S., Oker-Blom, P., Ross, V., and Vilikainen, L Architecture of Scots pine crown: phytometrical characteristics of needles and shoots. Silva Fenn. 20: Salemaa, M., Jukola-Sulonen, E.-L., Nieminen, T., and Nöjd, P Latvatunnukset ja puun kasvu elinvoimaisuuden ilmentäjinä. Finn. For. Res. Inst. Res. Pap. 446: [In Finnish.] SAS Institute Inc SAS system [computer program]. SAS Institute Inc., Cary, N.C. Schoettle, A.W., and Fahey, T.J Foliage and fine root longevity of pines. Ecol. Bull. 43: Viro, P.J Investigations on forest litter. Comm. Inst. For. Fenn. 45: 1 65.

11 Muukkonen and Lehtonen 2527 Wang, S., Grant, R.F., Verseghy, D.L., and Black, T.A Modelling plant carbon and nitrogen dynamics of a boreal aspen forest in CLASS the Canadian Land Surface. Ecol. Model. 142: Yarie, J., and Billings, S Carbon balance of the taiga forest within Alaska: present and future. Can. J. For. Res. 32: