CASE STUDY REPORT on Gauja/Koiva River to test and demonstrate the elaborated significance criteria

Transcription

1 CASE STUDY REPORT on Gauja/Koiva River to test and demonstrate the elaborated significance criteria Analysis of diffuse pollution sources and the methods of their assessment in the Estonian part of the Gauja/Koiva river basin district 2013 Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839)

2 Author Peeter Marksoo CASE STUDY REPORT ANALYSIS OF DIFFUSE POLLUTION SOURCES AND THE METHODS OF THEIR ASSESSMENT IN THE ESTONIAN PART OF THE GAUJA/KOIVA RIVER BASIN DISTRICT The report prepared in the frame of the Estonia Latvia programme project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) Tallinn 2013 Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 2

3 Table of contents Introduction EXPERIENCE OVER THE PAST DECADE IN DETERMINING DIFFUSE POLLUTION TUT METHOD LATVIAN MODEL THE WENNERBLOM MODEL CALCULATIONS OF NUTRIENT LOADS TO THE MUSTJÕGI RIVER AND IN ESTONIA AS A WHOLE; COMPARISON OF METHODS CONCLUSIONS AND RECOMMENDATIONS References Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 3

4 Introduction The objective of this report is to analyse under the project Towards joint management of the transboundary Gauja/Koiva river basin district the data available on diffuse pollution in the Estonian part of the Koiva river basin district (the Mustjõgi sub-basin), to assess diffuse load to surface water bodies and to make recommendations for the improvement of data collection and monitoring, where necessary. For the purposes of the part of the project dealing with point source loads, the term point pollution is mainly used. As regards diffuse sources, this paper mainly uses the term diffuse load because it is difficult to draw a line between background and anthropogenic nutrient loads and to determine when natural, background load becomes human-induced, i.e. pollution. This report aims in particular at comparing the diffuse nutrient loads calculated by using the method of estimating diffuse loads developed by the Department of Environmental Engineering of Tallinn University of Technology, the adapted Wennerblom model used in Latvian river basin management plans and the classic Wennerblom model suited for the conditions in Sweden, and at making recommendations for further cooperation. Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 4

5 1. EXPERIENCE OVER THE PAST DECADE IN DETERMINING DIFFUSE POLLUTION Over the past decade, the assessment of diffuse nitrogen and phosphorus loads has been addressed in several studies, mainly conducted by experts from the Department of Environmental Engineering of Tallinn University of Technology and AS Maves [1, 2, 3, 4, 5, 6], in connection with the preparation of river basin management plans. Different methods and models have been also used to calculate diffuse loads in the course of preparing river basin management plans for sub-basins in the period These were discussed in a greater detail in a study conducted by AS Maves in 2006 [2]; a summary of the findings is presented in the following three paragraphs and in Table 1. Different experts have calculated nutrient balances for sub-basins under different projects, using different models and methods. The level of accuracy of the calculations differs greatly, mainly due to different scopes of the projects (determined by the availability of external funding). Therefore, although catchment balances were successfully calculated, they have been presented in different forms, which are difficult to compare. For example, in the Harju sub-basin an analysis has been made for each river separately but the catchment balance for the whole river basin is missing. Moreover, while most models use, inter alia, surface area runoffs as source information, the MIKE Basin tool that was applied for example in the case of the Matsalu sub-basin uses the number of livestock and the quantity of mineral fertilisers as source data. The descriptions of processes and definitions used in different models are so different that it is impossible to compare the proportions of point and diffuse loads from different types of land use. Another reason why the models are difficult to compare is a significant climatic difference between the years on which the calculations are based. There is nearly a twofold difference between the loads in years of high and low water levels. The assessments of loads in different sub-basins concern, however, different years. A comparison of water management plans based on different sub-basins and other major catchment balances in Estonia is presented (Table 1) nonetheless. Besides water management plans, we have used the reports prepared under the MANTRA-East project, which include the total balance for the Lake Peipsi catchment area (the PolFlow model) and the balance for its Estonian part (the MESAW model). The balance for the catchment area of the Gulf of Finland was prepared under the project SEGUA. Nutrient balances for the Baltic Sea catchment area Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 5

6 are regularly prepared within the activities of HELCOM. The book Veekaitse põllumajanduses (Water protection in agriculture), published in 1985, includes a national balance, albeit outdated. More recent studies have demonstrated that if only one mathematical (empirical or physical) model is applied, no reliable results can be obtained, as a rule, for either of the two nutrients. Therefore, the authors [ 2 ] believe that no catchment balance prepared in Estonia is reliable. Although at least two models were applied in the case of the Lake Peipsi sub-basin, the results have not been integrated. Because the models are considered to be unreliable, expert assessments are used nowadays quite successfully. In Estonia, expert assessments have been used to determine the nutrient balance in the sub-basin of the Western islands; in other subbasins different models were applied. Table 1. Completed projects to determine nutrient balances in hydrological basins in Estonia [2] Project 1. In the course of preparing river basin management plans The Lake Peipsi, Viru and Lake Võrtsjärv subbasins The Harju sub-basin The sub-basin of Western islands The Pärnu sub-basin The Matsalu sub-basin The Koiva sub-basin 2. Other projects Implementation of Article 5 of the water framework directive in Estonia. Aggregated report on river basins (Ministry of the Environment, 2005) HELCOM PLC (HELCOM, 2004) Estimation of wastewater generation by source categories (Oras et al., 2006) Veekaitse põllumajanduses (Water protection in agriculture) (Maastik, 1984) SEGUE Method/model MESAW, PolFlow MESAW Expert assessment The Wennerblom model The MIKE Basin model N/A HELCOM HELCOM Expert assessment Expert assessment PolFlow Upon the completion of management plans for river basin districts in 2010, the management plans for sub-basins expired and are no longer available on the website of the Ministry of the Environment ( Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 6

7 The study conducted by AS Maves introduced a new method of calculating diffuse loads - first, the total potential load, i.e. the amount of nutrients from background and anthropogenic point and diffuse sources, part of which may leach into ground and surface waters, is calculated. Then the amount of removed nutrients, e.g. nutrients used in agricultural products, removed in the course of sewage treatment and volatilised nutrients, is subtracted from the total amount. The result is gross load - the theoretical load of nutrients before retention that would reach surface water at the river measuring section. Retention is a phenomenon that occurs when part of nutrients is retained, due to sedimentation, uptake by plants and volatilisation, as water flows from up- to downstream. When retention is subtracted from gross load, the result is net load, i.e. the (measurable) real load at the river measuring section (or in a body of ground water) that is reflected in the monitoring data. Figures 1 and 2 illustrate the nutrient balances in sub-basins calculated by using the above method. P, kg/a Potentsiaalne koormus, kg Kõrvaldamine, kg Hajuheide, kg Punktheide, kg Brutokoormus, kg Sisekoormus, kg Netokoormus, kg Peipsi koos Võrtsjärvega Viru Harju Matsalu Läänesaarte Pärnu Koiva Figure 1. Potential nitrogen load, removal, diffuse & point discharge, gross load, internal load and net load in sub-basins in Estonia (2004) [2] The authors point out several weaknesses of such approach, which are described below and with which we can only agree. Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 7

8 The assessment of nutrient loads is complicated due to the lack of measurable parameters and data. The accuracy of assessment is influenced by the accuracy of the assessment of background load (including precipitation) and retention. Since pure background load is impossible to measure, it is deemed to be the net load in catchment areas with low population density and a small share of arable land. However, the proportion of natural wetlands and woodlands is so significant in Estonia that even a one-kilogram error in the assessment of the load per hectare may conceal a large part of the impact of human activity. That part of the assessment should be differentiated from the rest by distinguishing as precisely as possible between the loads from different land parcels and areas with different usage intensity. P, kg/a Potentsiaalne koormus, kg Kõrvaldamine, kg Hajuheide, kg Punktheide, kg Brutokoormus, kg Sisekoormus, kg Netokoormus, kg Peipsi koos Võrtsjärvega Viru Harju Matsalu Läänesaarte Pärnu Koiva Figure 2. Potential phosphorus load, removal, diffuse & point discharge, gross load, internal load and net load in sub-basins in Estonia (2004) [2] When assessing potential load, the estimates of the amounts of diffuse domestic discharge as well as nitrogen and phosphorus released from the soil are subject to a high degree of uncertainty. Also, it is unknown how much liquid manure leaks directly into surface water and how much manure is spread directly on snow or frozen soil. As regards removal, it is difficult to determine how much nitrogen volatilises. A significant problem in the case of Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 8

9 diffuse discharge is that the difference between potential load and removal is very small. A 10% error in the assessment of potential load or removal may result in a negative value of discharge. Therefore, the part of the table concerning diffuse discharge is very sensitive. For example, a harvest from a natural grassland (in statistics, more than 5 year old cultural grasslands are included in natural grasslands) will result in a negative calculated value of diffuse discharge. A minor error in source data may result in a major error in the assessment of nutrient loads. The assessment of the soil and groundwater buffer and retention is even more complicated. In fact, more or less reliable data are only available on agriculture, precipitation, point discharge and net load, whereas the net load of the Narva River is subject to a high degree of uncertainty because the load from Russia to the Narva River prevents us from determining the load from the Estonian side. 2. TUT METHOD The above mentioned shortcomings, in particular as regards the phosphorus and nitrogen runoff coefficients for different land cover types, have been attempted to be removed in the works of the researchers of the Department of Environmental Engineering of Tallinn University of Technology (Enn Loigu, Arvo Iital, Karin Pachel, Ülle Leisk) [3, 6]. Experts have mainly based their assessment on the generalisation of the available surface water monitoring data which were compared with the data from neighbouring countries. Unit values of diffuse load are indicated separately for agricultural land cover types as well as for woodlands and wetlands. The method also takes into account the impact of clear-cutting and peat production on nutrient runoff. The areas of land cover types are those from the land cover database CORINE This method is hereinafter referred to as the TUT method. This method was used in calculations in the reports of 2008 and 2012 required by the Nitrates Directive. An overview of nutrient runoff coefficients of different land cover types used in the TUT method is presented in Table 2. These are long-term average runoff coefficients independent of the annual precipitation. Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 9

10 Table 2. Nutrient runoff coefficients of different land cover types used in the TUT method CORINE Ntot Ptot Land cover type No kg/ha/y kg/ha/y 111 Continuous urban fabric Discontinuous urban fabric Industrial and/or commercial units Road and rail networks and associated land Airports Green urban areas Non-irrigated arable land Fruit trees and berry plantations Complex cultivation patterns (>75% of arable land) Land principally occupied by agriculture, with significant areas of natural vegetation (<75%) Pastures Natural grassland Broad-leaved forest Coniferous forest Mixed forest Moors and heathland Transitional woodland/scrub on mineral land Transitional woodland/scrub on mire Open fens and transitional bogs Open lawn and pool communities Peat extraction areas , 512 Water courses, water bodies 4.5 Nutrient loads from agricultural land cover types were calculated by using averaged runoff coefficients and, as indicated above, the static model does not take into account any changes in the runoff conditions. The principles of calculating unit loads for different land cover types are explained in detail in the report Hajureostuse koormuse andmete täpsustamine, 2007 (Specification of diffuse pollution load data, 2007) [3]. Coefficients of nutrient runoff from agricultural landscapes were specified by using the data of hydrochemical monitoring of surface waters (from the mid-1990s to the present) by three automatic monitoring stations (Räpu/Arkma, Rägina/Kirna, Tõnga/Tõnga) that operate or operated in mainly agricultural catchment areas and from other monitoring points in catchment areas with a larger proportion Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 10

11 of arable land. The use of fertilisers on arable lands of different catchment areas varied from 50 to 170 kg N/ha and from 10 to 35 kg P/ha. It should be taken into account that agricultural land cover types (CORINE 211, 242 and 243) consist of other land cover types, including natural land cover types, and therefore, their total area is considerably larger than the area of land that is actually cultivated; this means that nutrient loads from such landscapes do not only represent agricultural diffuse and point source loads. The calculations of anthropogenic agricultural nutrient loads do not include the loads from pastures, in the case of which nutrient runoff is equal to that of dry forests, which represent natural, background load. Runoff unit values for different land cover types are also based on the results of monitoring carried out in catchment areas with diverse land cover, including, therefore, partially nutrient loads from natural land parcels, scattered settlements and, naturally, from precipitation, which is not differentiated in the case of any land cover type. The runoff coefficients of forests and swamps are based on national monitoring data and on comparison with the data published in neighbouring countries. Nutrient loads from clear-cut areas were calculated assuming that there is a nearly twofold increase in nutrient runoff during the first three years after cutting. Nitrogen loads deposited by precipitation directly on the surface of water bodies are based on the air monitoring data; in 2009, for example, it was between 3 and 6 kg N/ha in Estonia. As the phosphorus content in precipitation is, unfortunately, not measured, the value suggested by the TUT model is 31 mg P/m 2, based on previous studies. Therefore, the total quantity of phosphorus deposited by precipitation on the surface of water bodies would be 68.5 tonnes, which seems improbably large and, therefore, is not used in this report; the nutrient load deposited by precipitation on the surface of water bodies is deemed to be zero in the TUT model (according to the Wennerblom model, the relevant quantity is 17.7 tonnes). The nutrient load from unchanneled rainwater was calculated for high density areas, industrial and commercial units, roads and airports, assuming that rainwater either diffuses or is directed from water resistant surfaces to adjacent terrain or directly into inland water bodies. Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 11

12 Although the TUT method was designed mainly to determine diffuse load, it also takes into account direct load from manure storage facilities (which is unfortunately difficult to calculate) so that animal husbandry as an important source of nutrients would not be left out. If farms complied with the relevant legislation and good agricultural practice, no point source pollution would be created by farms and manure storage facilities. The fact is that some farms fail to comply with these requirements and recommendations to a greater or smaller extent. Different studies and experts have suggested different proportions of potential manure pollution from farms that end up as point source pollution. The TUT method suggests, as a compromise between different studies, that 10% of nitrogen and 1% of phosphorus created in a farm from manure reach water bodies as point source pollution. The nutrient loads calculated by the TUT method are attached to the report as an Excel spreadsheet (TTU_Mustjogi_Eesti). Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 12

13 3. LATVIAN MODEL In Latvian water management plans, diffuse loads were calculated by using the Wennerblom model developed by Tord Wennerblom, Sten-Åke Carlsson and Hans Kvarnäs and adapted in 2001, in cooperation between Swedish and Latvian experts, to take into account the requirements of Latvian water management plans and certain conditions that differ from those in Sweden. The model is hereinafter called the Latvian model. Latvian experts pointed out at a meeting of Estonian and Latvian experts, held on 14 and 15 February 2013 in Pärnu that the model was designed to help make decisions and could not be used to calculate scientific nutrient loads. The majority of nutrient coefficients are based on the indicators of the so-called classic Wennerblom model and only the diffuse load from agricultural lands was calculated based on Latvian monitoring data. In the future, other indicators are planned to be replaced with runoff coefficients based on local monitoring data. Latvian colleagues provided the author with calculations for the Melnupe (called the Peetri River in Estonia) and Vaidava rivers made by using the Latvian model; the calculations are attached to this report as files Melnupe1, Melnupe2 and Vaidava. Unfortunately, the Latvian expert who made the calculations is no longer available and therefore we could not ask him any questions. The Latvian model uses average runoff data because it was mainly designed to facilitate the development of measures. Although based on the Wennerblom model, the Latvian model, unlike the Wennerblom model, does not react to changes in precipitation or runoff inputs. Therefore, we can argue that the TUT and Latvian models are based on longterm average runoff coefficients that are independent of the annual rainfall and of the three methods, only the classic Wennerblom model helps to calculate nutrient runoff depending on rainfall. A detailed comparison of the three models is provided in Chapter 5, which includes calculations for the Mustjõgi river and Estonia as a whole made by using each model. Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 13

14 4. THE WENNERBLOM MODEL In the course of preparing this report, it was agreed with the client that besides the TUT and Latvian models the classic Wennerblom model, which is based mainly on data of Central Sweden, would also be used to calculate diffuse loads to the Mustjõgi river (hereinafter the WB model ). Although the authors of the model stress that Swedish runoff coefficients should be replaced by coefficients based on local data, the calculations made by using the WB model are interesting to be compared with those made by the TUT and Latvian models. Estonia is also in the process of adapting the WB model to local conditions and the calculations and comparisons contained in this report may be beneficial to that process. The calculations made by using the WB model are attached to the report as Excel spreadsheets (wennerblom_eesti and wennerblom_mustjogi). 5. CALCULATIONS OF NUTRIENT LOADS TO THE MUSTJÕGI RIVER AND IN ESTONIA AS A WHOLE; COMPARISON OF METHODS The aim of this report was to calculate diffuse loads in the Mustjõgi sub-basin (or the Estonian part of Koiva basin) by using the TUT and Latvian methods; in the course of work, calculations made by using the Wennerblom (WB) model were added. As the Latvian side made data available on land use on the Latvian side of the Mustjõgi river (the river, not the sub-basin) and other data required for calculating pollution loads, it was more practicable to calculate the nutrient loads to the catchment area of Mustjõgi river, not to Mustjõgi sub-basin, because both river discharge and water quality have been measured on the Mustjõgi river estuary since 2010 and, therefore, the calculated nutrient loads can be compared with the measured loads. Land use along the Mustjõgi river according to CORINE 2006 is presented on Figure 3. Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 14

15 Figure 3. Land use along the Mustjõgi river according to CORINE The calculations of diffuse loads in the sub-basin as a whole would not be comparable with the monitoring data. As the primary objective of this paper was to compare the methods of calculating diffuse loads, the author decided in favour of Mustjõgi river catchment area. However, since human impact on the Mustjõgi river is smaller than Estonian average and the natural conditions (the climate and the hilly landscape) in the region are atypical of Estonia, the report includes comparative calculations for Estonia as a whole. First, we will look at the three methods from the perspective of calculating the point source load, antrophogenic diffuse load, background diffuse load and their sum or the total load (Table 3, Figures 4-9). The percentages in the table indicate the proportion of a given source in total nutrient load. It should be mentioned that the loads for the Latvian part of the Mustjõgi river are those calculated in 2005 by using the Latvian and WB models, as presented by the Latvian colleagues. Although the share of treatment plants should be slightly higher, it should not change the complete picture much. Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 15

16 Table 3. Comparison of point source (PS) loads, anthropogenic (HI) diffuse loads, background (b/g) diffuse loads and their sums calculated by using the TUT, Latvian and Wennerblom models (based on the data of 2011) TUT method Latvian model Wennerblom model t/y % t/y % t/y % Estonia PS load Ntot 1, , , PS load Ptot Mustjõgi PS load Ntot PS load Ptot Estonia HI diff. load Ntot 18, , , HI diff. load Ptot Mustjõgi HI diff. load Ntot HI diff. load Ptot Estonia b/g load Ntot 11, , , b/g load Ptot Mustjõgi b/g load Ntot b/g load Ptot Estonia TOTAL Ntot 31, , , TOTAL Ptot Mustjõgi TOTAL Ntot 1, TOTAL Ptot % 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% TUT, Est TUT, Mustj Lat, Est Lat, Mustj WB, Est WB, Mustj point load atrop. diffuse load background load Figure 4. Comparison of total nitrogen point source loads, anthropogenic diffuse loads, background diffuse loads and their sums calculated for the Mustjõgi river and the total for Estonia by using the TUT, Latvian and Wennerblom models (based on the data of 2011) Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 16

17 tons Ntot/y TUT Latvian WB point load atrop. diffuse load background load Figure 5. Comparison of total nitrogen point source loads, anthropogenic diffuse loads and background diffuse loads calculated for all of Estonia by using the TUT, Latvian and Wennerblom models (based on the data of 2011) tons Ntot/y TUT Latvian WB point load atrop. diffuse load background load Figure 6. Comparison of total nitrogen point source loads, anthropogenic diffuse loads and background diffuse loads calculated for the Mustjõgi river by using the TUT, Latvian and Wennerblom models (based on the data of 2011) When comparing different models, it is often difficult to say which one is more accurate. As regards the WB model, it should be taken into account that it was not designed for Estonian conditions and, therefore, the differences are easier to explain. The pollution loads from agricultural land calculated by using the Latvian model are based on the conditions in Latvia, which match those in the southern part of Estonia. These may not suit for Estonia as a whole, taking into account the limestone bedrock prevalent in North-Estonia and on the Estonian islands. The calculations are based on the data of 2011, the year during which annual precipitation was close to the long-term average. The TUT method gives the biggest total nitrogen load, followed by those calculated by using the Latvian and WB models. According to the calculations based on monitoring data, approximately 23,000 tonnes of nitrogen were carried by rivers to the sea from the territory of Estonia in If we add the quantities discharged directly to the sea, Lake Peipsi, the Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 17

18 Narva and Koiva rivers (which reach outside Estonia), the estimated total nitrogen load is about 32,000 tonnes. This is approximately equal to the load calculated by the TUT method. The Latvian and WB models indicate significantly lower levels of nitrogen load (Figure 5). The measured nitrogen runoff in the Mustjõgi river was 664 tonnes in All three methods indicated higher levels of nitrogen runoff (Figure 6) whereas the TUT method gave the highest retention level. When comparing the relative shares of total nitrogen point source loads, anthropogenic diffuse loads and background diffuse load calculated for all of Estonia by using the TUT, Latvian and Wennerblom models (Figure 4), we can see that there are significant differences between the models. According to the WB model, background load only accounts for one fifth of the total load (and even less in case of the Mustjõgi river). According to the Latvian model, the share of background load is about half of the total load for Estonia as a whole and 2/3 for the Mustjõgi river. According to the TUT method, the share of background load accounts for slightly more than 1/3 of the total load. The differences between the shares of background load and between absolute figures also arise from the differences in the opinions about what should be considered as background load. The TUT method considers all load from forests, wetlands and pastures and the load deposited by precipitation directly on the surface of water bodies to be background load. Background load from agricultural lands, calculated on the basis of the runoff coefficient of pastures, is added to this amount. According to the TUT method, anthropogenic load is any nutrient load from agricultural lands (minus background load), diffused load from unchanneled rainwater, clear-cutting and peat production and point source load from manure storage facilities and other point sources. Table 4 illustrates the division of background and anthropogenic loads according to the WB model. The main difference between the results obtained by the WB model and the TUT method arises from the calculation of background load from agricultural land. While according to the TUT model, 21% of nitrogen load and 68% phosphorus load originates from background load from agricultural land, according to the WB model, the relevant percentages are 12% and 10%, respectively. This means that the difference between nitrogen load from agricultural land is about twofold and the figures for phosphorus load differ as much as seven times. Also, the WB model considers a significant part of nitrogen deposited by precipitation on surfaces of water bodies to be human-induced, i.e. caused by air pollution. The Latvian model, although based on the WB model, differentiates between background and Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 18

19 antrophogenic loads similarly to the TUT model (presuming that the author has interpreted the Latvian data correctly). Table 4. Division of the total nutrient load within the territory of Estonia into background and anthropogenic loads according to the Wennerblom model (based on the data of 2011) Nitrogen Phosphorus t/y % kg/y % NATURAL NUTRIENT RUNOFF Woodlands 2, , Arable land 1, , Wetlands , Other land Deposited by precipitation on the surface of lakes , TOTAL natural nutrient runoff 4, , share of total runoff (%) ANTHROPOGENIC NUTRIENT RUNOFF Lakes - deposited from atmosphere ,851 1 Woodlands due to air pollution 1, Arable land due to air pollution 1, Forestry - clear-cutting ,575 1 Forestry - irrigation Forestry - fertilisation Agriculture 9, , Milking facilities Manure storage facilities 2, , Individual sewerage systems , Wastewater treatment plants (except overflow) ,600 5 Overflow and rainwater ,640 1 Industry discharge directly to receiving water courses Fish farming TOTAL anthropogenic nutrient runoff 18, , share of total runoff (%) TOTAL 23, , When comparing the three models, it should be noted that phosphorus loads differ significantly. While nitrogen load was the lowest in the case of the WB model, phosphorus load is significantly higher than that obtained by the TUT and Latvian models. As regards Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 19

20 phosphorus load, the results obtained by the TUT and Latvian models were similar although the TUT method gave a slightly higher level (Figures 8 and 9). 100% 80% 60% 40% 20% 0% TUT, Est TUT, Mustj Lat, Est Lat, Mustj WB, Est WB, Mustj point load atrop. diffuse load background load Figure 7. Comparison of total phosphorus point source loads, anthropogenic diffuse loads and background diffuse loads calculated for the Mustjõgi river and all of Estonia by using the TUT, Latvian and Wennerblom models (based on the data of 2011) tons Ptot/y TUT Latvian WB point load atrop. diffuse load background load Figure 8. Comparison of total phosphorus point source loads, anthropogenic diffuse loads and background diffuse loads calculated for all of Estonia by using the TUT, Latvian and Wennerblom models (based on the data of 2011) tons Ptot/y TTÜ Läti WB point load atrop. diffuse load background load Figure 9. Comparison of total phosphorus point source loads, anthropogenic diffuse loads and background diffuse loads calculated for the Mustjõgi river by using the TUT, Latvian and Wennerblom models (based on the data of 2011) Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 20

21 According to the calculations based on monitoring data, approximately 620 tonnes of phosphorus were carried by rivers to the sea from the territory of Estonia in If we add the quantities discharged directly to the sea, Lake Peipsi, the Narva and Koiva rivers (which reach outside Estonia), the estimated total phosphorus load is about 650 tonnes. Differences in phosphorus levels are significantly smaller than differences in nitrogen levels. The measured phosphorus runoff in the Mustjõgi river was 29 tonnes in According to the TUT and Latvian models, retention is relatively modest compared with measured phosphorus runoff; according to the WB model, retention is about 1/3. As regards the Mustjõgi river, the Latvian method results in significantly smaller retention compared with measured runoff (29 t/y) and the TUT method gives a slightly lower level. According to the WB model, retention in the Mustjõgi river is quite significant (approx. 1/4). The reason why the level of phosphorus load calculated by the WB model is more than twice higher than those calculated by the TUT and Latvian models is that the nutrient load from arable land is calculated differently (Table 4). However, the load from woodland and consequently the whole background load are smaller than those resulting from the TUT and Latvian models. When phosphorus load is divided into background and anthropogenic loads, the background load resulting from the WB model is significantly smaller than the background load resulting from the TUT and Latvian models (Figure 7). The TUT and Latvian models are quite similar, the only difference being in the share of nitrogen load which is slightly bigger in the case of the TUT model; as regards phosphorus load, the situation is the opposite way around. One important difference between the TUT model and other models is that while according to the TUT model nutrient load from scattered dwelling areas is included in agricultural load, the Latvian and WB models have it as a separate item (Table 5). For nitrogen, such load accounts for a couple of per cent but for phosphorus it may account for more than 10% of total load. Therefore, when comparing nutrient loads from agriculture the load from scattered dwelling areas should be added to agricultural load in the case of the Latvian and WB models. Nutrient loads from agriculture comprise the largest part of the anthropogenic diffuse load and, therefore, are of key importance when planning measures to reduce nutrient loads. When we compare the anthropogenic agricultural loads calculated for Estonia as a whole by Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 21

22 using the three models (adding loads from scattered dwelling areas in the case of the Latvian and WB models), the results are quite different (Table 6; Figures 10 and 11). Table 5. Comparison of nutrient loads from different components of diffuse load, manure storage facilities and fish farms calculated by using the TUT, Latvian and Wennerblom models (based on the data of 2011) TUT method Latvian model WB model t/y % t/y % t/y % Estonia woodland Ntot 5, , , Estonia woodland Ptot Mustjõgi woodland Ntot Mustjõgi woodland Ptot Estonia clear-cutting Ntot Estonia clear-cutting Ptot Mustjõgi clear-cutting Ntot Mustjõgi clear-cutting Ptot Estonia wetlands Ntot 1, Estonia wetlands Ptot Mustjõgi wetlands Ntot Mustjõgi wetlands Ptot Estonia peat extraction areas Ntot Estonia peat extraction areas Ptot Mustjõgi peat extraction areas Ntot Mustjõgi peat extraction areas Ptot Estonia arable land Ntot 17, , , Estonia arable land Ptot Mustjõgi arable land Ntot Mustjõgi arable land Ptot Estonia grassland Ntot Estonia grassland Ptot Mustjõgi grassland Ntot Mustjõgi grassland Ptot Estonia other land Ntot Estonia other land Ptot Mustjõgi other land Ntot Mustjõgi other land Ptot rainwater from Estonia settlements Ntot Estonia rainwater from settlements Ptot rainwater from Mustjõgi settlements Ntot rainwater from Mustjõgi settlements Ptot deposited on the surface of water Estonia courses Ntot Estonia deposited on the surface of water courses Ptot deposited on the Mustjõgi surface of water Ntot Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 22

23 courses deposited on the surface of water Mustjõgi courses Ptot scattered dwelling Estonia areas Ntot Estonia scattered dwelling areas Ptot scattered dwelling Mustjõgi areas Ntot scattered dwelling Mustjõgi areas Ptot manure storage Estonia facilities Ntot , , Estonia manure storage facilities Ptot manure storage Mustjõgi facilities Ntot manure storage Mustjõgi facilities Ptot Estonia fish farming Ntot Estonia fish farming Ptot Table 6. Comparison of anthropogenic nutrient loads from agriculture and low density areas, diffuse load and background loads calculated by using the TUT, Latvian and Wennerblom models (based on the data of 2011) TUT method Latvian model WB model t/y % t/y % t/y % Arable land + scattered dwelling areas Ntot 17, , , Arable land + scattered dwelling areas Ptot Background load Ntot 11, , , Background load Ptot Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 23

24 tons Ntot / y TUT method Latvian model WB model Arable land + scattered dw ellings Background loadlooduskoormus Figure 10. Comparison of antrophogenic diffuse loads from agriculture and low density areas and background total nitrogen loads for all of Estonia calculated by using the TUT, Latvian and Wennerblom models (based on the data of 2011) tons Ptot / y TUT method Latvian model WB model Arable land + scattered dw ellings Background loadlooduskoormus Figure 11. Comparison of antrophogenic diffuse loads from agriculture and low density areas and background total phosphorus loads for all of Estonia calculated by using the TUT, Latvian and Wennerblom models (based on the data of 2011) As mentioned above, the TUT and Latvian models give relatively comparable results concerning both phosphorus and nitrogen background loads. According to the WB model, background load is about two times smaller. Although the total nitrogen background load calculated by the Latvian model is only 1.1 times bigger than the relevant indicator calculated by the TUT, there are significant differences in the assessments of its two main components: nitrogen from woodlands and wetlands. While according to the Latvian model 1.8 times more nitrogen comes from woodland, according to the TUT model nitrogen load from wetlands is Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 24

25 as much as 22 times greater than according to the Latvian model (Table 5). Such significant difference may also mean that the author has not correctly understood the Latvian model. There are also significant differences between the models regarding anthropogenic diffuse loads from agriculture and scattered dwellings areas. The TUT model gives nearly two times higher levels of nitrogen load than the Latvian model, while the results of the WB model remain in between the first two. As regards phosphorus load, according to the WB model, the anthropogenic agricultural and scattered dwellings diffuse load is three times higher than the load calculated by the TUT model and about 5.5 times higher than that calculated by the Latvian model. Agriculture and scattered dwelling areas are the areas where measures to reduce nutrient loads should be implemented. Therefore, we should focus on specifying the runoff coefficients for these areas. The high phosphorus loads resulting from the WB model may be put aside as unsuitable for Estonian and Latvian conditions. The rest of the anthropogenic diffuse load sources (clear cutting, peat extraction areas, unchanneled rainwater from settlements, air pollution) comprise a small part of the total load and, therefore, we will not deal in detail with the minor differences in Table 5. As regards the importance of diffuse load in shaping the ecological condition and water quality, we must distinguish between the quantities of nutrients carried to the sea and to lakes by watercourses, i.e. between nutrient runoff and the quality of water in watercourses. About 4/5 of the nutrient runoff takes place during the 3-4 months of spring floods and autumn rains. During the remaining 8 or 9 months rivers feed mainly on ground water, which contains nitrogen but very little phosphorus. Therefore, diffuse load has an impact on the nitrogen content in rivers all year round which is confirmed by a rather strong correlation between the average nitrate content of rivers covered by the hydro-chemical monitoring programme and the proportion of arable land in the catchment area covered by the monitoring programme (Figure 12). The calculations do not include the following rivers: rivers originating from lakes because of their high retention levels due to denitrification; the Piusa and Narva rivers, which are located on the state border, and for which the land use in the catchment area is not known; Oostriku and Vodja rivers and upper course of Kunda river where the topographical catchment areas very likely do not coincide with the hydrological catchment area; and the Leivajõgi river that is characterised by high content of humic substances. Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 25

26 80 70 y = 1,5902x + 6,6456 R 2 = 0, arable land % mg NO3/l Figure 12. Correlation between the proportion of arable land in the monitored catchment areas and the average nitrate content in rivers in the period between 2008 and kg N/ha 8 6 y = 1,1217x - 0,4137 R 2 = 0, mg N/l Figure 13. Correlation between the potential stocking density in the monitored catchment areas and the average total nitrate content in rivers in the period between 2008 and Since the share of arable land in total agricultural land is quite even in Estonia, similar strong correlation (R 2 = 0,75) is seen between the share of agricultural land and nitrate content in rivers. The nitrate content in rivers that run through areas with the share of arable land above the Estonian average is twice higher than the nitrate content in rivers that run through areas with the share of arable land below the Estonian average. The correlation between the potential nitrogen load from livestock in the catchment area and the nitrogen content in rivers (Figure 13) is weaker than in the case of arable land. Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 26

27 There is no correlation between the average phosphorus content in rivers and the phosphorus load from arable land or stocking density. The average annual phosphorus content mainly depends on the quantity of wastewater from settlements and industry and the prevailing natural conditions. In Pandivere, for example, which has a big share of arable land, the phosphorus content is below the Estonian average due to natural conditions (limestone has high phosphorus-binding capacity and rivers feed mainly on groundwater). 6. CONCLUSIONS AND RECOMMENDATIONS 1. The calculation of diffuse loads has been quite extensive in Estonia over the past decade yet a good solution/model that would satisfy all requirements has not been found. The reason lies in the complexity of assessing diffuse loads and scarcity of measured parameters. One of the obstacles is the lack of a relevant database - the necessary data are either missing or it takes a lot of time to enter the data in the models. It should be considered to add to national monitoring programmes some monitoring stations that measure the nutrient load from agricultural land and natural landscapes - these data would make the future assessment of nutrient runoff rates easier. 2. This report compares the amounts of diffuse nutrient loads for the Mustjõgi river catchment area and calculated for Estonia as a whole by using the diffuse load assessment method developed by the Department of Environmental Engineering of Tallinn University of Technology (the TUT model), the adapted Wennerblom model used in water management plans in Latvia (the Latvian model) and the classic Wennerblom model suited for the conditions in Sweden, and analyses the differences between the results obtained by applying these models and the reasons for such differences. 3. The TUT method gives the biggest total nitrogen load, followed by those calculated by using the Latvian and WB models. According to the calculations based on monitoring data, approximately 23,000 tonnes of nitrogen were carried by rivers to the sea from the territory of Estonia in If we add the quantities discharged directly to the sea, Lake Peipsi, the Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 27

28 Narva and Koiva rivers (which reach outside Estonia), the estimated total nitrogen load is about 32,000 tonnes. This is approximately equal to the load calculated by the TUT method. The Latvian and WB models indicate significantly lower levels of nitrogen load (Figure 5). The measured nitrogen runoff in the Mustjõgi river was 664 tonnes in All three methods indicated higher levels of nitrogen runoff, whereas the TUT method gave the highest retention level. 4. When comparing the three models, it should be noted that phosphorus loads differ significantly. While nitrogen load was the lowest in the case of the WB model, phosphorus load is significantly higher than that obtained by the TUT and Latvian models. As regards phosphorus load, the results obtained by the TUT and Latvian models were similar although the TUT method gave a slightly higher level. According to the calculations based on monitoring data, approximately 620 tonnes of phosphorus were carried by rivers to the sea from the territory of Estonia in If we add the quantities discharged directly to the sea, Lake Peipsi, the Narva and Koiva rivers (which reach outside Estonia), the estimated total phosphorus load is about 650 tonnes. Differences in phosphorus levels are significantly smaller than differences in nitrogen levels. The measured phosphorus runoff in the Mustjõgi River was 29 tonnes in The three models define background load differently. The TUT method considers all load from forests, wetlands and pastures and the load deposited by precipitation directly on the surface of water bodies to be background load. Background load from agricultural lands, calculated on the basis of the runoff coefficient of pastures, is added to this amount. The WB model considers a significant part of nitrogen deposited by precipitation on the surfaces of water bodies to be human-induced, i.e. caused by air pollution. The main difference between the results obtained by the WB model and the TUT method arises from the calculation of background load from agricultural land. While according to the TUT model, 21% of nitrogen load and 68% phosphorus load originates from background load from agricultural land, according to the WB model, the relevant percentages are 12% and 10%, respectively. This means that the difference concerning nitrogen load from agricultural land is about twofold and the figures for phosphorus load differ as much as seven times. The Latvian model, although based on the WB model, differentiates between background and anthropogenic loads similarly to the TUT model. Project Towards joint management of the transboundary Gauja/Koiva river basin district (Nr. EU 38839) 28