Abstract

The statistical model MESAW (Matrix Equations for Source Apportionment on Watershed) was used to estimate the diffuse unit-area source emission coefficients of nitrogen in Estonian rivers. The input data included monitored riverine loads, point sources and land use categories from a total of 50 rivers/catchment areas. Two independent studies were conducted: the estimation of emission coefficients for the whole of Estonia and for a smaller study area near Tallinn. The results from both cases showed that drained peat soils were the highest diffuse source contributor in unit-area loads. The results show that the unit-area loads from drained peat soils were up to 2.3 times higher than from arable land. Moreover, a comparison of emission coefficients for the whole of Estonia and for the Tallinn catchment area indicated that coefficients can vary significantly between sources and single years. Additional detailed studies and monitoring are needed to support these conclusions.

INTRODUCTION

One of the major environmental management goals in the Baltic Sea region, and thus also in Estonia, is the reduction of riverine nutrient loads. However, recent data analyses of Estonian rivers indicate that nitrogen concentrations have increased in some rivers (Iital et al. 2010) despite the decrease in fertilizer use (Vassiljev & Blinova 2012). An increase in nitrogen concentrations has even been detected in watersheds with very low human activity (Iital et al. 2010). Hoffmann et al. (2000) assumed that the increase in nutrient pollution could also be caused by wide-scale melioration. Some authors (Heikkinen 1994; Kløve 2001; Kløve et al. 2010) have reported that drainage of peat soils leads to decomposition of peat and increases the flux of nutrients to watercourses. The drainage of peatlands results in peat oxidation and significantly changes their physical and chemical properties (Litaor et al. 2008; Verhoeven & Setter 2010). This can result in high nitrate-nitrogen concentrations in the pore water of drained peatlands that is caused by the aeration of peat and subsequent mineralization and nitrification of organic N (Tiemeyer et al. 2007). Vassiljev & Blinova (2012) and Vassiljev (2015) have observed an increase in nitrogen concentrations in some Estonian river basins with a high percentage of drained peat soils. Mineralization of nitrogen from peat soils is regarded as another possible source of nutrients in Europe (Eurostat 2011). To illustrate the problem, the correlation between the concentration of total nitrogen and water runoff is shown for the Leivajõgi River and the Hatu River for 2014–2015 (Figure 1). Both of the watersheds have a relatively high percentage of drained peat soils (Leivajõgi 45.1% and Hatu 44.1%) and there is a low level of human activity in the catchment areas. The Leivajõgi is situated about 20 km south-east of a large city (Tallinn, ∼450,000 inhabitants) and the Hatu River is surrounded by natural areas. High concentrations of nitrogen in the Leivajõgi can be seen as a result of diffuse pollution from roads and other human activity arising from its proximity to a large city. However, Figure 1 shows that the concentrations of total nitrogen increase with the increase in water flow in both rivers at the same rate. This clearly indicates that there are other, larger sources than the influence of a large city.

Figure 1

Correlation between total nitrogen concentration and daily runoff for two rivers with a high percentage of drained peat soils.

Figure 1

Correlation between total nitrogen concentration and daily runoff for two rivers with a high percentage of drained peat soils.

The main objective of this paper was to quantify the emission coefficients of various land cover types and drained peat soils. This paper uses the results obtained in Vassiljev et al. (2016) but with a significant increase in the amount of data and the time period analyzed. Moreover, the data analysis in this study is more comprehensive.

CASE STUDIES AND METHODOLOGY

In the current study, a statistical approach was used to estimate the nitrogen emission coefficients from various diffuse source categories. Emission coefficients were estimated for the whole territory of Estonia and, for comparison purposes, also for the drinking water catchment area of Tallinn city where there is a high density of water quality monitoring sites (Figure 2).

Figure 2

Map of Estonia and location of Tallinn catchment area.

Figure 2

Map of Estonia and location of Tallinn catchment area.

The statistical model MESAW (Matrix Equations for Source Apportionment on Watershed) was used for source apportionment and retention estimates of nitrogen (Grimvall & Stålnacke 1996). This model has been shown to be expedient for source apportionment, especially for areas with a high density of water quality monitoring sites (Lidèn et al. 1998; Vassiljev et al. 2008; Stålnacke et al. 2015). The model approach uses non-linear regression for simultaneous estimation of source strength (i.e. export/emission coefficients to surface waters) for the different land use or soil categories and retention coefficients for pollutants in a river basin or lake. The basic principles and major steps in the procedure include: (1) estimation of mean annual riverine N loads for years 2005, 2006, 2007, 2008, 2011, 2012 and 2013 at each water quality monitoring site, (2) subdivision of the entire drainage basin into sub-basins according to the upstream area of the water quality monitoring site, (3) derivation of statistics on land use, lake area, point source emissions and other relevant data for each sub-basin, (4) application of a general non-linear regression expression with the loads at each sub-basin as the dependent/response variable and sub-basin characteristics as covariates/explanatory variables. In the MESAW model, the load at the outlet of an arbitrary sub-basin can be estimated using the following general expression (Equation (1)):  
formula
(1)
where Li= load at the outlet of sub-basin i; Lj= load at the outlet of nearest upstream sub-basin j; Rj,i = retention on the way from the outlet of sub-basin j to the outlet of sub-basin i; n = number of sub-basins located to the nearest upstream; Si = total losses from soil to water in sub-basin i; Pi = point source discharges to waters in sub-basin i; Di = atmospheric deposition on surface waters in sub-basin i; R= retention in sub-basin i; ɛi = statistical error term.

The load at each sub-basin can be divided into contributions from the sources located in sub-basins further upstream (the first term in Equation (1)) and contributions from the sources located within the sub-basin under consideration (the Si, Pi and Di terms). It should be taken into account that for some sub-basins n can be equal to zero (e.g. the uppermost sub-basin or separate basin without any upstream sub-basin). In this case, Equation (1) will be used without the first term. The parameterization of the model is flexible and can be study-area specific. The model is fitted by minimizing the sum of squares for the difference in the observed and estimated loads. In this study, Pi and Di were assumed to be known and Si was assumed to be a simple function of land use according to Si= (β1a1i+β2a2i+β3a3i). Coefficients a1i, a2i and a3i denote the area of arable land (excluding arable land on drained peat soil, ∼5% of total arable land), other land (named natural areas) that included forests, pastures, natural grasslands, bogs (excluding areas on drained peat soils) and drained peat soils (including all land types located on drained peat soils) in the sub-basin i, and β1, β2 and β3 are unknown export coefficients (i.e. emission coefficients, unit-area loads) for the three land use categories calculated by MESAW.

Nutrients are normally retained temporarily or permanently in watercourses. In the model, retention is expressed as a summary expression for all hydrological and biogeochemical processes that may retain nutrients. In MESAW it can be parameterized by any empirical function. In this study, the retention is subdivided into retention in lakes and riverine retention (i.e. instream retention). It was assumed that retention in lakes is a function of the lake area divided by the drainage area, and riverine retention a function of the drainage area.

Both types of retention can be expressed through the following equation:  
formula
(2)
where λ1 and λ2 denote a non-negative parameter and Ri denotes the retention in the ith basin. The first part of the function reflects the instream retention whereas the second part reflects the retention in lakes and reservoirs.
Retention from an arbitrary sub-basin m to the river mouth (Rm, mouth) can be derived from:  
formula
(3)
where Rm,mouth = retention from the outlet of the sub-watershed m on the way to the mouth of the whole river; k = number of sub-basins downstream of sub-basin m; Rj = values of retention within the different sub-basins downstream of sub-basin m.

In case 1 (whole of Estonia), the hydro-chemical parameters were collected from 50 sites in the annual state monitoring program by the Estonian Environmental Agency (EEA). The data collected over a seven-year period from 2005 to 2013 were used in this study. In addition, the EEA measured water runoff at 46 locations that, for some measurement points, did not coincide with the hydro-chemical monitoring sites. In case 2, the hydro-chemical parameters were collected from 12 sites, 11 of them sampled by AS Tallinna Vesi with nitrogen concentration measurements varying from 12 to 52 times per year and one sampled by the EEA (six to 12 times per year). The chemical analysis of water samples was performed in accredited laboratories. The percentage of peat soils in the watersheds was estimated on the basis of a digital soil map obtained from the Estonian Land Board. The digital CORINE land cover map was used to derive land use statistics for each sub-basin. In the model, the emission coefficients were calculated for arable land excluding peat soils, areas with drained peat soils and other land use categories excluding peat soils categorized as natural areas (including forests, pastures, natural grasslands and bogs). Characteristics of the sub-basins for case 1 and case 2 are presented in Tables 1 and 2.

Table 1

Land cover in basins/sub-basins in case 1

  Land cover (km2)
 
River basin Monitoring point Arable (km2Drained peat (km2Other land (km2Lakes (km2
Võhandu Outflow from lake Vagula 223 64 274 12 
Võhandu Himmiste 348 105 376 14 
Võhandu Downstream from Räpina 451 119 553 16 
Väike Emajõgi Tõlliste 338 141 447 9.5 
Väike Emajõgi Pikasilla bridge 385 181 599 9.8 
Õhne Downstream from Suislepa 167 68 317 5.2 
Õhne Upstream from Tõrva, Roobe 57 34 161 4.7 
Tarvastu Upstream from Põdraoja 46 10 35 0.0 
Tänassilma Oiu 173 60 214 1.1 
Emajõgi Rannu-Jõesuu 1,082 455 1,465 275 
Emajõgi Tartu 2,590 1,365 3,460 308 
Emajõgi Kavastu 2,918 1,455 3,743 315 
Pedja Jõgeva Sordiaretusjaam 182 163 318 0.3 
Pedja Tõrve 223 176 371 1.1 
Põltsamaa Rutikvere 285 159 414 3.8 
Porijõgi Reola 116 39 81 1.0 
Ahja Kiidjärve 134 41 158 0.4 
Ahja Lääniste 380 92 456 1.9 
Kääpa Outflow from Kose reservoir 70 50 142 2.8 
Avijõgi Mulgi 88 37 264 0.0 
Rannapungerja Iisaku-Avinurme road, Roostoja 39 33 157 0.0 
Tagajõgi Tudulinna 19 10 212 0.3 
Alajõgi Griini (Alajõe) 16 29 109 0.3 
Pühajõgi River mouth 60 38 121 0.4 
Purtse River mouth (Tallinn-Narva mnt) 163 96 552 0.6 
Kunda Lavi springs 59 58 200 0.2 
Kunda River mouth 141 78 313 0.4 
Seljajõgi River mouth 217 35 164 0.0 
Loobu Vihasoo 110 36 168 0.0 
Valgejõgi Loksa pedestrian bridge 104 60 286 1.9 
Pudisoo Pudisoo 24 9.1 103 0.0 
Jägala Linnamäe 393 323 845 5.4 
Leivajõgi Pajupea 16 35 28 0.0 
Vääna River mouth 120 43 150 1.6 
Keila Keila 245 93 300 0.0 
Keila River mouth 273 102 307 0.0 
Vihterpalu Vihterpalu 71 117 293 0.0 
Kasari Kasari bridge 829 393 1,422 2.1 
Velise Valgu 28 25 89 0.0 
Pärnu Tahkuse 644 334 1,090 0.7 
Pärnu Oore 1,543 728 2,861 6.5 
Vodja Vodja 24 3.5 24 0.0 
Navesti Aesoo 298 190 553 0.0 
Saarjõgi Kaansoo 34 26 127 0.0 
Halliste Riisa 545 74 1,249 5.8 
Reiu Downstream from Lähkma 91 56 385 0.0 
Sauga Nurme weir 147 74 323 0.9 
Pirita Lükati 223 180 399 5.8 
Mustajõgi Mustajõe 27 59 286 2.1 
Rannapungerja Bridge on Mustvee road 71 49 472 0.3 
  Land cover (km2)
 
River basin Monitoring point Arable (km2Drained peat (km2Other land (km2Lakes (km2
Võhandu Outflow from lake Vagula 223 64 274 12 
Võhandu Himmiste 348 105 376 14 
Võhandu Downstream from Räpina 451 119 553 16 
Väike Emajõgi Tõlliste 338 141 447 9.5 
Väike Emajõgi Pikasilla bridge 385 181 599 9.8 
Õhne Downstream from Suislepa 167 68 317 5.2 
Õhne Upstream from Tõrva, Roobe 57 34 161 4.7 
Tarvastu Upstream from Põdraoja 46 10 35 0.0 
Tänassilma Oiu 173 60 214 1.1 
Emajõgi Rannu-Jõesuu 1,082 455 1,465 275 
Emajõgi Tartu 2,590 1,365 3,460 308 
Emajõgi Kavastu 2,918 1,455 3,743 315 
Pedja Jõgeva Sordiaretusjaam 182 163 318 0.3 
Pedja Tõrve 223 176 371 1.1 
Põltsamaa Rutikvere 285 159 414 3.8 
Porijõgi Reola 116 39 81 1.0 
Ahja Kiidjärve 134 41 158 0.4 
Ahja Lääniste 380 92 456 1.9 
Kääpa Outflow from Kose reservoir 70 50 142 2.8 
Avijõgi Mulgi 88 37 264 0.0 
Rannapungerja Iisaku-Avinurme road, Roostoja 39 33 157 0.0 
Tagajõgi Tudulinna 19 10 212 0.3 
Alajõgi Griini (Alajõe) 16 29 109 0.3 
Pühajõgi River mouth 60 38 121 0.4 
Purtse River mouth (Tallinn-Narva mnt) 163 96 552 0.6 
Kunda Lavi springs 59 58 200 0.2 
Kunda River mouth 141 78 313 0.4 
Seljajõgi River mouth 217 35 164 0.0 
Loobu Vihasoo 110 36 168 0.0 
Valgejõgi Loksa pedestrian bridge 104 60 286 1.9 
Pudisoo Pudisoo 24 9.1 103 0.0 
Jägala Linnamäe 393 323 845 5.4 
Leivajõgi Pajupea 16 35 28 0.0 
Vääna River mouth 120 43 150 1.6 
Keila Keila 245 93 300 0.0 
Keila River mouth 273 102 307 0.0 
Vihterpalu Vihterpalu 71 117 293 0.0 
Kasari Kasari bridge 829 393 1,422 2.1 
Velise Valgu 28 25 89 0.0 
Pärnu Tahkuse 644 334 1,090 0.7 
Pärnu Oore 1,543 728 2,861 6.5 
Vodja Vodja 24 3.5 24 0.0 
Navesti Aesoo 298 190 553 0.0 
Saarjõgi Kaansoo 34 26 127 0.0 
Halliste Riisa 545 74 1,249 5.8 
Reiu Downstream from Lähkma 91 56 385 0.0 
Sauga Nurme weir 147 74 323 0.9 
Pirita Lükati 223 180 399 5.8 
Mustajõgi Mustajõe 27 59 286 2.1 
Rannapungerja Bridge on Mustvee road 71 49 472 0.3 
Table 2

Land cover in basins/sub-basins in case 2

  Land cover (km2)
 
River basin Monitoring point Arable (km2Drained peat (km2Other land (km2Lakes (km2
Jägala Sael 144 102 213 0.1 
Pirita Ardu 2.6 12 37 0.0 
Pirita Paunküla 5.5 23 59 4.9 
Pirita Ravila 23 34 84 0.2 
Pirita Kose-Uuemõisa 38 40 111 0.0 
Pirita Saula 95 73 201 0.2 
Pirita Vaida 126 125 346 1.3 
Jägala Kaunissaare 188 180 465 4.9 
Raudoja Veehoidla 0.1 1.0 21 0.6 
Kuivajõgi Kose-Uuemõisa 52 31 80 0.5 
Kurna Oja 7.9 8.4 19 0.0 
  Land cover (km2)
 
River basin Monitoring point Arable (km2Drained peat (km2Other land (km2Lakes (km2
Jägala Sael 144 102 213 0.1 
Pirita Ardu 2.6 12 37 0.0 
Pirita Paunküla 5.5 23 59 4.9 
Pirita Ravila 23 34 84 0.2 
Pirita Kose-Uuemõisa 38 40 111 0.0 
Pirita Saula 95 73 201 0.2 
Pirita Vaida 126 125 346 1.3 
Jägala Kaunissaare 188 180 465 4.9 
Raudoja Veehoidla 0.1 1.0 21 0.6 
Kuivajõgi Kose-Uuemõisa 52 31 80 0.5 
Kurna Oja 7.9 8.4 19 0.0 

RESULTS AND DISCUSSION

One challenge of estimating emission coefficients (expressed as unit-area losses) for a whole country is the differences in flow rates (e.g. specific runoff) between the studied sites. Another problem is that the difference in the water runoff between the sites varies between different years, as shown in Figure 3. The Keila River is located in N Estonia and the Võhandu River in SE Estonia. The figure shows that the runoff is almost the same for some years, while it differs significantly for other years (e.g. years 2004 and 2012). In the model, the load at the outlet sub-basin L was calculated as the sum of daily loads. The daily loads, on the other hand, were calculated by multiplying concentrations with discharge. If the discharges are very different in different rivers across Estonia, the calculated loads will differ significantly, not, however, because of land use or soil type but because of the variation in water flow. It was found necessary to adjust riverine loads in the MESAW model for each year under investigation in order to make the results comparable. Firstly, the average annual runoff was calculated using the flow rates of all the rivers in Estonia. Secondly, the flow rate coefficients for each river were calculated by dividing the average runoff by the river's runoff. Finally, the corrected loads were calculated by multiplying loads by the flow rate coefficients. In these procedures, it was assumed that the correlation between the emissions and the runoff is linear.

Figure 3

Annual water runoff in the Võhandu River (SE Estonia) and the Keila River (N Estonia) in 1997–2013.

Figure 3

Annual water runoff in the Võhandu River (SE Estonia) and the Keila River (N Estonia) in 1997–2013.

As an example, Figure 4 shows the dependency between emission and runoff for the Võhandu River (SE Estonia) and the Keila River (N Estonia). The distance between the river mouths is 280 km. It can be seen that the correlation between the two variables is more or less linear. Thus, this rough method was regarded as a suitable means of obtaining a good estimate of the riverine loads for MESAW.

Figure 4

Dependence between annual total nitrogen emission and water runoff (Keila and Võhandu, 1997–2013).

Figure 4

Dependence between annual total nitrogen emission and water runoff (Keila and Võhandu, 1997–2013).

Another problem is the low sampling frequency. In this study, load was calculated using interpolated concentration values between actual measurements. This can lead to significant errors if, for example, all the samples are taken during low water flow. If concentrations increase with flow, the load estimated using the concentrations measured during low water flow will then be lower than the actual load. Figure 5 shows the measured daily discharge and total nitrogen concentration in the Leivajõgi River for 2013. The state monitoring sampling rate changed after 2010 from 12 to six times per year. The result can be seen in Figure 5: in 2013 all samples were collected during low flow discharge. This means that the actual loads in 2013 must be greater than those estimated using interpolated concentration values. The dependency between the concentrations and discharge for 2011–2013 is shown in Figure 6. As a consequence, the load calculations based on interpolated concentrations were underestimated more than twice in 2013. Thus, they had to be corrected before being used as input in MESAW. When no data were available for concentrations in the case of large discharges (e.g. 2013), a regression from other years (e.g. 2011 and 2012) was used to estimate flow-weighted nutrient fluxes.

Figure 5

Measured discharge and concentrations of total nitrogen in the Leivajõgi River in 2013.

Figure 5

Measured discharge and concentrations of total nitrogen in the Leivajõgi River in 2013.

Figure 6

Correlation between the measured discharge and concentrations of total nitrogen in the Leivajõgi River.

Figure 6

Correlation between the measured discharge and concentrations of total nitrogen in the Leivajõgi River.

MESAW was used for calculations in two case studies. First, the diffuse source emission coefficients of total nitrogen were estimated for seven single years (2005, 2006, 2007, 2008, 2011, 2012 and 2013) for the whole of Estonia (Table 3). Years were selected with different average runoff to cover the whole range of runoff from minimal to maximal values. In 2009 and 2010 runoff at most of the sampling points was in the range covered by other selected years. Also the number of sampling points in those years was smaller than in 2011–2013. Therefore data from 2009 and 2010 were not considered in this study.

Table 3

Results of estimated total nitrogen emission coefficients of the three land cover classes for the whole of Estonia

Year Arable, kg/ha Standard error Natural areas, kg/ha Standard error Drained peat areas, kg/ha Standard error Average water runoff, mm 
2005 10.7 2.0 1.5 0.5 24.4 3.8 285 
2006 8.4 1.5 0.8 0.4 10.7 3.1 160 
2007 16.8 2.3 1.2 0.3 19.1 4.7 245 
2008 19.9 4.2 2.8 1.3 33.6 8.0 409 
2011 17.9 2.9 2.2 0.7 32.7 6.1 338 
2012 21.8 3.2 2.7 1.2 25.8 6.2 392 
2013 11.5 2.6 1.2 0.5 23.2 5.4 253 
Year Arable, kg/ha Standard error Natural areas, kg/ha Standard error Drained peat areas, kg/ha Standard error Average water runoff, mm 
2005 10.7 2.0 1.5 0.5 24.4 3.8 285 
2006 8.4 1.5 0.8 0.4 10.7 3.1 160 
2007 16.8 2.3 1.2 0.3 19.1 4.7 245 
2008 19.9 4.2 2.8 1.3 33.6 8.0 409 
2011 17.9 2.9 2.2 0.7 32.7 6.1 338 
2012 21.8 3.2 2.7 1.2 25.8 6.2 392 
2013 11.5 2.6 1.2 0.5 23.2 5.4 253 

All the estimated coefficients for all years were statistically significant (p < 0.05). The highest unit-area loads for total nitrogen loads in all years were from drained peat soils. The results are somewhat contradictory to official assessments (Estonian Ministry of Environment 2015), which have shown that the highest unit-area loads come from arable land.

In the next step of the analysis, the MESAW model was applied to a smaller area (Tallinn's drinking water catchment area) with better temporal resolution in its sampling frequency. The modelling results of source apportionment in the Tallinn catchment area are presented in Table 4.

Table 4

Results of estimated total nitrogen emission coefficients of the three land cover classes for the Tallinn catchment area

Year Arable, kg/ha Standard error Natural areas, kg/ha Standard error Drained peat areas, kg/ha Standard error Average water runoff, mm 
2005 22.1 6.8 2.8 1.7 24.8 9.3 263 
2006 15.1 6.5 1.0 0.8 21.9 8.7 137 
2007 25.4 9.0 1.6 1.1 32.6 10.1 249 
2008 43.6 13.9 3.4 2.8 42.2 17.6 382 
2011 20.4 6.0 2.8 1.6 34.3 8.3 312 
2012 32.8 7.8 4.5 2.4 38.5 10.3 423 
2013 12.1 5.1 1.9 1.6 28.2 10.4 225 
Year Arable, kg/ha Standard error Natural areas, kg/ha Standard error Drained peat areas, kg/ha Standard error Average water runoff, mm 
2005 22.1 6.8 2.8 1.7 24.8 9.3 263 
2006 15.1 6.5 1.0 0.8 21.9 8.7 137 
2007 25.4 9.0 1.6 1.1 32.6 10.1 249 
2008 43.6 13.9 3.4 2.8 42.2 17.6 382 
2011 20.4 6.0 2.8 1.6 34.3 8.3 312 
2012 32.8 7.8 4.5 2.4 38.5 10.3 423 
2013 12.1 5.1 1.9 1.6 28.2 10.4 225 

All the coefficients for all years were statistically significant (p < 0.05). Similar to the analysis above for the whole of Estonia, the results for the Tallinn catchment area showed that the unit-area losses for drained peat soils are significantly higher than for arable land (except 2008). It is notable that the emission coefficients for all land types are much higher in the Tallinn catchment area than for the whole of Estonia (Tables 3 and 4). This is most likely to be due, in part, to the difference in the sampling frequency but it is even more likely to be due to the spatial scale issue. Table 3 contains average values for the whole of Estonia (relatively large area with significant differences in emission coefficients) while Table 4 contains average values for a smaller catchment area in Northern Estonia.

Previous studies have shown that the emission coefficients from arable land can vary by up to a factor of 25 in different countries (Wickham & Wade 2002) and, according to some studies, even by up to a factor of 8 in Estonia (Deelstra et al. 2014). Vassiljev et al. (2008) studied the emission coefficients for the whole of Estonia during 1995–2005 and found that the coefficient of nitrogen from arable land was 12.3 kg/ha compared to 15.2 kg/ha obtained in this study. Stålnacke et al. (2015) estimated emissions from cultivated areas in the Baltic Sea drainage basin to be 14.3 kg/ha. Emissions from drained peat areas were not estimated. Studies conducted in SE Estonia (Povilaitis et al. 2012) resulted in somewhat lower export coefficients compared to the average values defined for the whole of Estonia. This corresponds well with the findings in this study that show that the export coefficients in the study area (N Estonia) are higher than the average values. This indicates that the sources of nitrogen loads should be defined at the catchment area level rather than at the country level to ensure accurate estimations. The spatial issue is further illustrated in Figure 4 where the emission in SE Estonia is much lower than that in N Estonia. For example, the measured water runoff in 2010 was practically the same (Figure 2) in both watersheds. Land use and strength of point sources (kg/km2) in 2010 are relatively similar (Table 5). However, nitrogen emission was 2.4 times higher in the Keila River watershed (N Estonia) than in the Võhandu River watershed (SE Estonia). This could be caused by different relief, different depth of groundwater table, etc. Evidently, the average emission coefficients for the whole of Estonia will be lower than for N Estonia when SE Estonian rivers are included in the estimation of emission coefficients. Moreover, it is likely that there is more intensive agriculture and some additional sources of pollution (e.g. higher air deposition due to more intensive road transport) in the study area close to Tallinn. From the results, it is evident that an extensive study is required to quantify more precise source emissions of nitrogen in Estonia.

Table 5

Land use and point sources in the Võhandu River and the Keila River in 2010

River-Location Arable, % Drained peat areas, % Other areas (not arable and not peat), % Point sources, kg/km2 
Võhandu-Räpina 40.1 10.6 49.3 16.0 
Keila-Keila 38.4 14.6 47.0 10.0 
River-Location Arable, % Drained peat areas, % Other areas (not arable and not peat), % Point sources, kg/km2 
Võhandu-Räpina 40.1 10.6 49.3 16.0 
Keila-Keila 38.4 14.6 47.0 10.0 

Emission coefficients from drained peat areas have not been studied in Estonia. A field study in Northern Germany (Scholz & Trepel 2004) revealed that the nitrate concentrations in drained peatlands in an unnatural state can be 5–60 times higher than in natural peatlands. The high emission coefficients of drained peat soils obtained in this study can be explained by the high percentage of peatlands in an unnatural state in Estonia, with only 5.5% of peatlands in a near-natural state (Paal & Leibak 2011).

From the modelling results of case study 2 (Tallinn catchment area) it can be seen that the loads from arable land in 2011, 2012 and 2013 are substantially lower at the same water runoff than in 2005, 2006, 2007 and 2008 (Figure 7). This may be related to the last financial crisis when the use of fertilizer decreased in Estonia during the same period (Statistics Estonia 2015). Annual changes in temperature also have an effect on nutrient concentrations. In Tiemeyer et al. (2007) it was found that in some years NO3–N concentrations were low (or zero) at the beginning of the discharge season, while low flow rates at the end of the discharge season (April) coincided with comparatively high NO3–N concentrations. This was explained by different climate conditions during the study period as the net release of nitrogen from peat soils increases with increasing temperature (Koerselman et al. 1993). Additional studies are needed to determine the relationship between temperature and changes in emission coefficients.

Figure 7

Emission of nitrogen from agricultural areas vs annual runoff in the Tallinn catchment area.

Figure 7

Emission of nitrogen from agricultural areas vs annual runoff in the Tallinn catchment area.

The export coefficients in MESAW are estimated by correlating observed and calculated loads. For example, Figure 8 presents the comparison between the modelled and measured loads in the Tallinn catchment area for 2012. The correlation between the observed and modelled loads is linear, indicating that the calculated results are reliable.

Figure 8

Calculated vs observed nitrogen loads for the Tallinn catchment area (2012).

Figure 8

Calculated vs observed nitrogen loads for the Tallinn catchment area (2012).

Calculations of the retention showed that coefficients λ1 and λ2 were different for different years and varied between 0.0008 and 0.0025 for λ1 and between 2.5 and 6.7 for λ2. This means, for instance, that retention varies from 7 to 20% in different years in rivers with a sub-basin area of 100 km2 (Equation (2)). For example, if the lake area divided by the drainage area equals 0.168 (corresponds to the Võrtsjärv sub-basin), the retention in different years varies from 30 to 53%. The variation in retention may be explained by different weather conditions.

According to official assessments (Estonian Ministry of Environment 2015), the highest unit-area loads come from arable land. At the same time, Estonia is one of the most peatland-rich countries in the world with almost 22% of the country covered in peatlands in various conditions (Orru 1992), of which 5.5% of the peatlands are in a near-natural state (Paal & Leibak 2011). The rest of the peatlands are affected by drainage for forestry, agriculture and peat extraction (Karofeld et al. 2016). Therefore, it is necessary to study the nutrient inflow to the waterbodies from peatlands more thoroughly.

A comparison of emission coefficients for Estonia and the Tallinn catchment area indicated that the coefficients can vary significantly, even in a relatively small country. Therefore, it is suggested that the sources of nitrogen loads should be defined at the catchment area level rather than at the country level. The current study showed that average unit-area losses from drained peat soils can be estimated with a sufficient degree of accuracy and precision. However, the processes are far more complex at smaller spatial scales. For example, the mineralization of nitrogen depends on the peat type and characteristics, soil moisture, depth, drainage types and construction time, so the emission coefficients in different spatial locations could differ. It is suggested that drained peatlands should be a particular concern and that additional and more detailed monitoring and field studies are needed.

Effective restoration strategies are required to reduce nutrient inflows to the waterbodies from drained peatlands. These do not depend solely on the restoration technique adopted but on how well integrated the catchment management schemes are and how well the interacting mechanisms are understood. Much more work is required to examine the hydrological and hydro-chemical processes surrounding artificial drainage and peatland restoration (Holden et al. 2004).

CONCLUSIONS

The MESAW model enabled estimation of statistically significant (p < 0.05) diffuse emission coefficients in two case studies for arable land, drained peat soils and other land. The study showed that unit-area losses from drained peat soils can be up to 2.3 times higher than from arable land. This is somewhat different from official assessments, which have shown that the highest unit-area loads result from arable lands. Drained peat soils must be recognized as an additional source of nitrogen. This study showed that it is a significant source in Estonia where the percentage of unnatural peatlands is high (∼20% of total area of Estonia). Therefore it would be necessary to reconsider the nutrient reduction strategies in Estonia.

Unit-area losses from drained peat soils were estimated to vary between 11 and 34 kg/ha in the whole of Estonia and 22–42 kg/ha in the study area around Tallinn city. This indicates that the emission coefficients in different spatial locations could differ because of a number of different factors. Therefore it is suggested that the sources of nitrogen loads should be defined at the catchment area level rather than at the country level because the emission coefficients can vary significantly even in a small country like Estonia.

ACKNOWLEDGEMENTS

This study was supported by the Institutional Research Funding IUT19-17 at Tallinn University of Technology, by the EEA Financial Mechanism 2009-2014 program ‘Integrated marine and inland water management’ (project title ‘Development of data modelling system and the decision support tool for the integrated marine and inland water management’), and by the Central Baltic Programme project Waterchain. The authors would like to thank AS Tallinna Vesi for providing valuable data.

REFERENCES

REFERENCES
Deelstra
,
J.
,
Iital
,
A.
,
Povilaitis
,
A.
,
Kyllmar
,
K.
,
Greipsland
,
I.
,
Blicher-Mathiesen
,
G.
,
Jansons
,
V.
,
Koskiaho
,
J.
&
Lagzdins
,
A.
2014
Hydrological pathways and nitrogen runoff in agricultural dominated catchments in Nordic and Baltic countries
.
Agric. Ecosyst. Environ.
195
,
211
219
.
EC Eurostat
2011
Data Requirements, Availability and Gaps in Agri-Environment Indicators (AEIs) in Europe
,
2011 edn
.
Publications Office of the European Union
,
Luxembourg
.
Estonian Ministry of Environment
2015
Lääne-Eesti veemajanduskava 2016–2020
.
Tallinn
,
Estonia
.
Grimvall
,
A.
&
Stålnacke
,
P.
1996
Statistical methods for source apportionment of riverine loads of pollutants
.
Environmetrics
7
,
201
213
.
Hoffmann
,
M.
,
Johnsson
,
H.
,
Gustavson
,
A.
&
Grimvall
,
A.
2000
Leaching of nitrogen in Swedish agriculture – a historical perspective
.
Agric. Ecosyst. Environ.
80
,
277
290
.
Holden
,
J.
,
Chapman
,
P. J.
&
Labadz
,
J. C.
2004
Artificial drainage of peatlands: hydrological and hydrochemical process and wetland restoration
.
Prog. Phys. Geogr.
28
,
95
123
.
Iital
,
A.
,
Pachel
,
K.
,
Loigu
,
E.
,
Pihlak
,
M.
&
Leisk
,
U.
2010
Recent trends in nutrient concentrations in Estonian rivers as a response to large-scale changes in land-use intensity and lifestyles
.
J. Environ. Monit.
12
,
178
188
.
Karofeld
,
E.
,
Müür
,
M.
&
Vellak
,
K.
2016
Factors affecting re-vegetation dynamics of experimentally restored extracted peatland in Estonia
.
Environ. Sci. Pollut. Res.
23
,
13706
13717
.
Kløve
,
B.
,
Sveistrup
,
T. E.
&
Hauge
,
A.
2010
Leaching of nutrients and emission of greenhouse gases from peatland cultivation at Bodin, Northern Norway
.
Geoderma
154
,
219
232
.
Koerselman
,
W.
,
Van Kerkhoven
,
M.
&
Verhoeven
,
J. T. A.
1993
Release of inorganic N, P and K in peat soils; effect of temperature, water chemistry and water level
.
Biogeochemistry
20
,
63
81
.
Lidèn
,
R.
,
Vasilyev
,
A.
,
Loigu
,
E.
,
Stålnacke
,
P.
&
Wittgren
,
H. B.
1998
Nitrogen source apportionment using a physically and statistically based model
.
Ecol. Modell.
114
,
235
250
.
Litaor
,
M. I.
,
Eshel
,
G.
,
Sade
,
R.
,
Rimmer
,
A.
&
Shenker
,
M.
2008
Hydrogeological characterization of an altered wetland
.
J. Hydrol.
349
,
333
349
.
Orru
,
M.
1992
Eesti Turbavarud (Estonian Peat Resources)
.
Eesti Geoloogiakeskus
,
Tallinn
,
Estonia
.
Paal
,
J.
&
Leibak
,
E.
2011
Estonian Mires: Inventory of Habitats
.
Publication of the Project, Estonian Mires Inventory Completion for Maintaining Biodiversity
.
Regio, Tartu
,
Estonia
.
Povilaitis
,
A.
,
Stålnacke
,
P.
&
Vassiljev
,
A.
2012
Nutrient retention and export to surface waters in Lithuanian and Estonian river basins
.
Hydrol. Res.
43
,
359
373
.
Scholz
,
M.
&
Trepel
,
M.
2004
Water quality characteristics of vegetated groundwater-fed ditches in a riparian peatland
.
Sci. Total Environ.
332
,
109
122
.
Stålnacke
,
P.
,
Pengerud
,
A.
,
Vassiljev
,
A.
,
Smedberg
,
E.
,
Mörth
,
C. M.
,
Hägg
,
H. E.
,
Humborg
,
C.
&
Andersen
,
H. E.
2015
Nitrogen surface water retention in the Baltic Sea drainage basin
.
Hydrol. Earth Syst. Sci.
19
,
981
996
.
Statistics Estonia
2015
Statistical Yearbook of Estonia
.
Tallinn
,
Estonia
. .
Vassiljev
,
A.
2015
Modelling of the influence of drained areas on the environment
. In:
International Conference on Civil, Structural and Environmental Engineering Computing
(
Kruis
,
J.
,
Tsompanakis
,
Y.
&
Topping
,
B. H. V.
, eds).
The Fifteenth Prague
,
Czech Republic
,
paper 280
.
Vassiljev
,
A.
&
Blinova
,
I.
2012
The influence of drained peat soils on diffuse nitrogen pollution of surface water
.
Hydrol. Res.
43
(
4
),
352
358
.
Vassiljev
,
A.
,
Blinova
,
I.
&
Ennet
,
P.
2008
Source apportionment of nutrients in Estonian rivers
.
Desalination
226
,
222
230
.
Vassiljev
,
A.
,
Margus
,
G.
,
Annus
,
I.
&
Stålnacke
,
P.
2016
Investigation of possible nutrient sources in Estonian rivers
.
Proc. Eng.
162
,
188
195
.
Verhoeven
,
J. T. A.
&
Setter
,
T. L.
2010
Agricultural use of wetlands: opportunities and limitations
.
Ann. Botany
105
,
155
163
.
Wickham
,
J. D.
&
Wade
,
G. D.
2002
Watershed level risk assessment of nitrogen and phosphorus export
.
Comput. Electron. Agric.
37
,
15
24
.