Advertisement

Phosphorus and nitrogen fluxes carried by 21 Finnish agricultural rivers in 1985–2006

  • Petri Ekholm
  • Katri Rankinen
  • Hannu Rita
  • Antti Räike
  • Heidi Sjöblom
  • Arjen Raateland
  • Ljudmila Vesikko
  • José Enrique Cano Bernal
  • Antti Taskinen
Article

Abstract

The Finnish Agri-Environmental Programme aims to reduce nutrient load to waters. Using national monitoring data, we estimated the agricultural load (incl. natural background) of total phosphorus (TP) and total nitrogen (TN) transported by 21 Finnish rivers to the northern Baltic Sea and analysed the flow-adjusted trends in the loads and concentrations from 1985 to 2006. We also related the loads to spatial and temporal patterns in catchment and agricultural characteristics. Agricultural load of TN increased, especially in the rivers discharging into the Bothnian Bay, while the load of TP decreased in most of the rivers, except those discharging into the Archipelago Sea. The trends may partly be related to a decrease in grassed area (TP, TN) and increased mineralisation (TN), but the available data on catchment and agricultural characteristics did not fully explain the observed pattern. Our study showed that data arising from relatively infrequent monitoring may prove useful for analysing long-term trend. The mutual correlation among the explaining variables hampered the analysis of the load generating factors.

Keywords

Agriculture Catchment Nitrogen Phosphorus River Trend Agri-environmental measures 

Introduction

The impact of agriculture on surface waters is a topic that has received both scientific and public attention in Finland. The first studies quantifying agricultural nutrient load were published in the 1960s and 1970s (Kajosaari 1965; Kauppi 1979), and as far back as 30 years ago, agriculture was identified as the largest source of total phosphorus (TP) and total nitrogen (TN) in surface waters (Kauppi 1984). More recent chemical and biological monitoring data indicates that most lakes, rivers and coastal waters under agricultural influence fail to achieve a good ecological status (Varanka and Luoto 2011; Ekholm et al. 2007; Ekholm and Mitikka 2006; Rask et al. 2010; Sutela and Vehanen 2009; Kauppi et al. 1990). Yet, the significance of agricultural activities in nutrient loading and eutrophication has been repeatedly questioned in public discussion (Valve et al. 2013). Disputes over the issue may partly arise from the diffuse nature of the agricultural nutrient load, which hampers its quantification under a fluctuating climate and in catchments where other loading sources are also present.

The implementation of the EU’s Water Framework and Marine Strategy directives calls for a reduction in agricultural nutrient losses to surface water in Finland. The key instrument is formed by the Finnish Agri-Environmental Programme, launched in 1995 to conform to the EU’s agri-environmental policy (EEC 1992). Through the programme, more than 90 % of the farmers have been subsidised for carrying out environmentally beneficial agricultural practices. Due to subsidies and other economic and political drivers (Laukkanen and Nauges 2014), fertiliser application rates of P in particular have decreased, autumn ploughing has made way for less intensive tillage operations, buffer strips and riparian zones have been implemented and wetlands constructed (Aakkula et al. 2011; Turtola 2007). At the same time, changes in agricultural production have taken place that may have counteracted the measures listed above. For instance, the total area of fallow land has decreased and crop and animal production have become even more spatially separated, hindering the recycling of manure (Niemi and Ahlstedt 2012). On top of agri-environmental measures and changes in agricultural production, climate affects the load: milder winters may have increased the mobilisation of N and particulate P (Ekholm et al. 2007; Lepistö et al. 2008; Puustinen et al. 2007). Finally, the time lag between measures and responses is likely to delay the desired outcome in the receiving bodies of water (Bouraoui and Grizzetti 2011).

The efficiency of agri-environmental measures has generally been studied on a small scale, e.g. on a field plot. However, such studies do not provide information on the combined effects on a catchment scale, which are an outcome of various environmental and agronomical conditions. A catchment-scale response can be evaluated by examining monitoring data from agricultural rivers and streams (Granlund et al. 2005; Ekholm et al. 2007; Cherry et al. 2008; Vuorenmaa et al. 2002; Stålnacke et al. 2014). Problems here include the difficulty of differentiating the effect of agri-environmental measures from other human pressures, potentially masked by climatic fluctuation (Kronvang et al. 2005; Vagstad et al. 2004). In addition, infrequent sampling increases the risk of missing peak load events (Johnes 2007).

By using data from national monitoring programmes, we made an effort (1) to separate the share of agricultural TP and TN load from the total nutrient flux transported by 21 Finnish rivers to the northern Baltic Sea, (2) to analyse whether the agricultural load shows a trend over the period 1985–2006 and (3) to examine the relationship between the load and spatial and temporal patterns in catchment characteristics. The nutrient fluxes were calculated using an empirical method relying on a river-specific relationship between concentrations and flow (Q). This method allowed us to account for infrequent sampling and variation in Q. The rivers discharge into four Baltic Sea sub-basins, of which the Gulf of Finland, the Archipelago Sea and the Bothnian Sea are considered primarily N-limited, but the northernmost sub-basin, the Bothnian Bay, is limited by P (Tamminen and Andersen 2007). However, a reduction of both nutrients in all the sub-basins is required, since P promotes cyanobacterial blooms in N-limited areas, and N can be transported from the P-limited Bothnian Bay to southern N-limited sub-basins.

Materials and methods

Catchments

The 21 river catchments (357–4923 km2) belong to a national long-term monitoring project on riverine material inputs into the Baltic Sea. Ten of the rivers are located in southern and 11 in western Finland (Table 1, Fig. 1). The rivers in southern Finland discharge into the Gulf of Finland, into the Archipelago Sea and into the Bothnian Sea, and most of the rivers in western Finland discharge into the Bothnian Bay. The mean precipitation in the area ranges from 600 to 700 mm and the mean temperature is about 5 °C. Generally, soil texture tends to be fine with clayey soils dominating in southern Finland, whereas coarser soils are typical of western Finland, where acid sulphate soils are also present (Lilja et al. 2009). Fields in western Finland also have a flatter topography than fields in southern Finland.
Table 1

Characteristics of the rivers. Field%, Lake%, Built-up%, Forest% and Peat% refer to the proportion of these land use types in the catchment area and OrgField% to the proportion of fields on organic soils in the catchment; hydrology refers to the years 1985–2006

Baltic Sea sub-basin/ river

Region

Catchment

Flow, m3 s−1

Base flow index (−)

Area, km2

Field%

Lake%

Built-up%

Forest%

Peat%

OrgField%

Mean

Min.

Max.

Gulf of Finland

 Virojoki

South

357

13.5

3.8

1.4

78.5

13.5

1.11

4.3

0.1

38.1

0.53

 Koskenkylänjoki

South

895

30.3

4.4

2.1

60.2

2.5

0.55

8.1

0.6

77.9

0.73

 Porvoonjoki

South

1273

31.2

1.3

4.1

59.1

4.2

1.24

12.7

0.7

130.9

0.53

 Mustijoki

South

783

30.3

1.5

3.6

61.1

10.2

2.49

6.9

0.1

122.5

0.45

 Vantaanjoki

South

1686

23.8

2.3

9.2

57.2

8.2

1.24

15.9

1.0

175.2

0.49

 Karjaanjoki

South

2046

17.7

12.2

4.6

62.5

6.2

0.90

18.9

0.0

70.6

0.80

Archipelago Sea

 Kiskonjoki

South

629

17.1

8.1

3.0

65.2

8.4

1.22

10.3

0.0

54.6

0.80

 Paimionjoki

South

1088

42.8

1.6

2.5

49.0

3.5

0.57

9.5

0.0

121.4

0.40

 Aurajoki

South

874

36.8

0.3

4.8

52.7

8.6

1.04

8.5

0.0

163.6

0.29

Bothnian Sea

 Eurajoki

South

1336

23.5

12.9

2.3

57.7

11.6

1.54

9.5

0.0

61.1

0.71

 Lapväärtinjoki

West

1098

13.5

0.2

0.8

82.9

25.1

1.09

14.1

1.1

146.3

0.51

 Närpiönjoki

West

992

21.6

0.4

1.3

73.2

18.7

1.22

8.9

0.2

142.3

0.44

Bothnian Bay

 Kyrönjoki

West

4923

24.6

1.2

1.7

67.2

23.4

2.38

42.9

1.1

407.4

0.46

 Lapuanjoki

West

4122

21.1

2.9

1.4

70.1

23.0

2.76

32.5

0.2

330.6

0.55

 Ähtävänjoki

West

2054

13.7

9.8

1.3

70.7

24.2

1.85

15.0

2.4

65.2

0.87

 Perhonjoki

West

2524

10.1

3.4

0.8

83.0

33.4

2.42

22.5

1.1

283.2

0.58

 Lestijoki

West

1373

10.5

6.2

0.8

80.7

30.9

0.69

12.0

1.4

204.3

0.55

 Kalajoki

West

3658

15.5

1.9

1.2

78.7

21.6

2.14

40.6

2.4

591.7

0.47

 Pyhäjoki

West

3712

9.0

5.2

1.0

80.1

26.8

1.87

31.8

2.6

514.1

0.52

 Siikajoki

West

4318

8.0

0.5

0.5

85.5

44.0

1.87

40.1

0.7

468.8

0.55

 Kiiminginjoki

West

3814

1.3

3.0

0.7

92.1

42.6

0.69

41.8

2.3

609.0

0.61

Fig. 1

River catchments chosen for this study

The data was divided into four periods of around five years. Period I (1985–1989) and period II (1990–1994) represent the situation before the Finnish Agri-Environmental Programme, but in period II a higher proportion (up to 20 %) of fields were fallowed due to compulsory measures to restrict overproduction. Period III (1995–1999) and period IV (2000–2006) represent the first and second phases of the Finnish Agri-Environmental Programme, respectively. For the description of individual measures required by the programme, see Aakkula et al. (2011) and Laukkanen and Nauges (2014).

The catchments were characterised using geoinformation databases and annual statistics. If data was available only on a municipal or regional level, it was allocated to the catchments by calculating area-weighted means. The proportion of fields in the catchment (Field%) was based on total agricultural land (in the year 2000) in the SLICES (Separated Land Use/Land Cover Information System) database (resolution 25 × 25 m), in which agricultural fields were digitised from orthophotos (Sucksdorff and Teiniranta 2001). The proportion of lakes in the catchment (Lake%) was based on shoreline data at a scale of 1:20,000. Data on the field use types, i.e. the proportions of cereal and oil seed crops (Crop%), grass (Grass%) and fallow (Fallow%) in total agricultural land, and the corresponding yields, was available from agricultural yearbooks by rural districts or centres. More spatially accurate data by municipality was available from 1995. The proportion of peat land in the catchment (Peat%) was obtained from the Finnish Soil Database (Lilja et al. 2009), which is based on the interpretation made by Agrifood Research Finland of the geological map of Quaternary Deposits of the Geological Survey of Finland. It describes the soil at 1 m according to the Finnish classification based on the content of organic matter and the texture of mineral material. The proportion of fields on organic soils in the catchment (OrgField%) was obtained by combining the above information on soil classes with that from CORINE Land Cover 2006.

Nitrogen balance has been considered a key agri-environmental indicator of the potential N losses (Yli-Viikari et al. 2007; Salo and Turtola 2006). The balances of N were calculated by Agrifood Research Finland for the period 1990–2005. The use of N and P in manure was derived from the number of domestic animals by municipality, taken from agricultural yearbooks (excluding fur farms), and the amount of P and N excreted by animals based on values provided by the Ministry of the Environment. Losses of N are also a function of mineralisation, making the losses sensitive to temperature outside the growing season. The daily soil temperature, measured at a depth of 20 cm every fifth day, was available for eight sites from the database of the Finnish Meteorological Institute. The catchments in southern Finland fell outside this area, and thus soil temperature (annual temperature sum, TSumAnn, and temperature sum in December–January, TSumWin) was used as an explanatory variable only for the catchments in western Finland. Soil test P (STP) values of cultivated soil in Finland is based on an extraction with acid ammonium acetate (Vuorinen and Mäkitie 1955), the results correlating with the concentration of dissolved P in runoff (Uusitalo and Jansson 2002). STP values by rural centres were taken from Viljavuuspalvelu Oy (Mäntylahti 2003).

“Agricultural load” (including natural background) was estimated by subtracting the reported point-source load and the estimated loads caused by forestry practices and the sparse population from the total riverine nutrient flux. Municipal and industrial load was taken from the VAHTI database (data starting from 1994) and from the statistics of the National Board of Waters. Load from scattered settlements (taken from the VEPS database of the Finnish Environment Institute, www.syke.fi) was based on the number of inhabitants whose homes were not connected to sewer systems, treatment type and distance from a water body. Finally, the load from forested land (the VEPS database) was based on specific loads for each forestry practice (Kenttämies and Mattson 2006).

Field% ranged from 1.3 to 42.8 % (mean 21 %, Table 1) and was higher in southern than in western Finland, with the Paimionjoki and the Aurajoki having the highest Field%, and the northernmost catchment (the Kiiminginjoki) the lowest (Table 1). Lake% ranged from 0.2 to 12.9 % (mean 3.8 %), the Karjaanjoki, the Kiskonjoki, the Eurajoki and the Ähtävänjoki having the highest values. The remaining area was mainly forested, with the percentage of forestry land (Forest%) increasing towards the north; the share of built-up areas (Built-up%) was below 10 % in all the catchments (mean 2.3 %, Table 1). Peat% ranged from 2.5 to 44 %, being highest in western Finland. OrgField% accounted from 0.55 % to 2.76 % of the total catchment area.

Daily Q values (m3 s−1) were based on daily water level recordings with calibrated flow-rating curves (the databases of the Finnish Environment Institute). The mean annual runoff (Runoff, in metres) in each period was calculated by dividing mean Q by the catchment area above the discharge measurement site. Base flow index, the ratio of mean annual base flow to mean annual flow, ranged from 0.29 to 0.80 (Table 1).

The temporal and spatial differences in the catchment characteristics were analysed by a mixed linear model. An unstructured variance-covariance structure was applied for modelling the dependencies among repeated measurements. No difference in Runoff between regions (i.e. southern and western Finland) was found, but Runoff decreased from period I to period IV (Table 2). Crop% was higher in southern than in western Finland, whereas the opposite was true for Grass%. The temporal variation in Crop% mainly reflected the variation in Fallow%, except that Grass% decreased and Crop% increased in western Finland from period III to period IV. STP was lowest in the crop production areas of southern Finland. It increased from period I to period III, but decreased to period IV, especially in western Finland. The amount of P applied in manure decreased by periods (Table 2). The balance of N was higher in western than in southern Finland and decreased from period to period in both regions (Table 3). Manure N showed no difference between regions but also decreased over time. Annual and winter temperature sums were highest in period IV. Some of explaining catchment and agricultural variables correlated with each other (Tables 4 and 5). For example, Field% correlated strongly with Crop% and Grass% and Runoff with P and N in manure. For such cases, the results are obviously uncertain.
Table 2

Mean values for the agricultural load of total phosphorus (TP) and the explaining catchment and agricultural variables in southern and western Finland by periods

 

TP, kg km−2 y−1

Runoff, m

Crop%

Grass%

Fallow%

STP, mg l−1

P in manure, kg ha−1

South

West

South

West

South

West

South

West

South

West

South

West

South

West

Period I

33.5

23.0

0.318

0.321

77

49

16

40

5.6

6.5

11.7

13.7

2.83

2.84

Period II

29.4

22.9

0.302

0.290

63

41

14

40

18.5

16.6

12.0

13.8

2.51

2.64

Period III

31.1

19.1

0.304

0.279

67

47

16

42

10.4

8.2

13.9

16.0

2.03

1.93

Period IV

27.7

18.9

0.271

0.282

68

54

13

33

10.7

7.7

13.6

14.4

1.43

1.91

Region

ns

ns

***

***

*

ns

ns

Period

**

***

***

***

***

***

**

Region  period

ns

***

***

***

***

***

*

The bottom part shows ANOVA results

STP soil test phosphorus, ns not significant

*p < 0.05, **p < 0.01, ***p < 0.001

Table 3

Mean values for the agricultural load of total nitrogen (TN) and N-specific explaining variables in southern and western Finland by periods

 

TN, kg km−2 y−1

N balance, kg ha−1

N in manure, kg ha−1

TSumAnn, °C

TSumWin, °C

South

West

South

West

South

West

West

West

Period I

537

341

74.1

86.5

15.5

15.5

364

−4.00

Period II

553

316

59.1

71.3

14.5

15.8

357

−1.88

Period III

547

311

58.6

68.4

12.4

11.3

351

−9.46

Period IV

553

396

42.1

58.1

10.8

13.9

409

2.90

Region

ns

**

ns

Period

ns

***

**

***

***

Region  period

**

ns

ns

The bottom part shows ANOVA results

TSumAnn annual temperature sum, TSumWin temperature sum in December–January, ns not significant

*p < 0.05, **p < 0.01, ***p < 0.001

Table 4

Correlation among variables used to explain agricultural load of total phosphorus (TP)

Variable

TP

Lake%

Runoff

Field%

Crop%

Grass%

Fallow%

STP

Lake%

−0.59***

       

Runoff

0.15

−0.37***

      

Field%

0.71***

−0.06

−0.19

     

Crop%

0.31**

0.13

−0.01

0.73***

    

Grass%

−0.29**

−0.17

0.09

−0.77***

−0.93***

   

Fallow%

0.00

0.01

−0.07

0.19

−0.06

−0.27*

  

STP

0.08

0.03

−0.23*

−0.20

−0.29**

0.28*

−0.24*

 

P in manure

−0.16

−0.07

0.61***

−0.36***

−0.16

0.21

0.01

−0.08

STP soil test phosphorus, ns not significant

*p < 0.05, **p < 0.01, ***p < 0.001

Table 5

Correlation among variables used to explain agricultural load of total nitrogen (TN)

Variable

TN

Lake%

Runoff

Field%

Crop%

Grass%

Fallow%

Peat%

OrgField%

N balance

N in manure

TSumAnn

Lake%

−0.58***

           

Runoff

0.18

−0.43***

          

Field%

0.80***

−0.18

−0.23

         

Crop%

0.47***

0.20

0.02

0.74***

        

Grass%

−0.47***

−0.24

0.06

−0.78***

−0.92***

       

Fallow%

0.08

0.04

−0.05

0.20

−0.03

−0.26

      

Peat%

−0.54***

−0.24

0.00

−0.83

−0.23***

0.83

−0.19*

     

OrgField%

−0.08

−0.10

−0.37**

0.00

−0.84*

0.23

0.06

0.22

    

N balance

−0.20

−0.10

0.21

−0.29*

−0.37**

−0.52***

−0.34**

0.39**

−0.02

   

N in manure

−0.12

−0.05

−0.60***

−0.46***

−0.23

0.28*

−0.03

−0.24***

−0.45***

0.26*

  

TSumAnn

0.60***

−0.27

−0.11

0.49**

0.83***

−0.82

0.30***

−0.32*

0.22

−0.42**

−0.15

 

TSumWin

0.48**

0.06

−0.18

0.52***

0.64***

−0.65

0.23***

−0.43***

0.21*

−0.03

−0.38*

0.64***

TSumAnn annual temperature sum, TSumWin temperature sum in December–January, ns not significant

*p < 0.05, **p < 0.01, ***p < 0.001

Water quality

The rivers were sampled by the regional environment authorities 12–28 times per year with a special focus on the high flow events in spring and autumn. Data on TP and TN concentrations (μg l−1) was retrieved from the database of the Finnish Environment Institute. Pre-screening of the data revealed some outlying observations, but only those observations were excluded for which the laboratory traced an error either in sampling or analysis. Determination of TP was performed by the molybdenum blue method with ascorbic acid as a reductant and digestion with potassium peroxodisulphate. TN determination was initiated by digestion with peroxodisulphate, followed by reduction of NO3 with a Cd amalgam and determination of NO2 by the azo colour method.

Estimating nutrient fluxes

Due to a sparse sampling frequency, it was imperative that the calculation method of nutrient fluxes efficiently made use of available data. Visual inspection revealed that, despite a lot of scatter, the concentrations of TP and TN had a tendency to increase with Q, typical of diffuse polluted rivers. However, in some rivers, the concentration-flow pattern was “U-shaped” in that the concentrations first decreased then increased with Q, suggesting an initial dilution, say, of a point-source load. To account for such a concentration-flow regime, the daily concentrations of TP (C P) and TN (C N) were predicted from daily Q using Eq. 1:
$$ {C}_{\mathrm{P},\ \mathrm{N}}=a+\frac{b}{Q}+cQ $$
(1)
where a, b and c ≥ 0. The form of Eq. 1 was taken from Wartiovaara (1975), with the exception that we predicted concentrations rather than loads, as done in the original application. The approach resembles that used by Jarvie et al. (2010) in differentiating point and non-point load in rivers in the UK. Constant a represents the fraction of nutrients apparently independent of Q, constant b the fraction inversely dependent on Q and constant c the fraction which size is positively dependent on Q. The constants were estimated for each nutrient, river and period by fitting the equation to observed data (i.e. the concentrations of TP and TN and Q on the sampling days) using the Newton algorithm. The model converged for most of the rivers and periods, although not always statistically significantly, especially in the case of C N. Daily nutrient fluxes were calculated by multiplying C P or C N by Q and annual and periodic fluxes by summing the daily ones. Generally, model residuals were relatively normally distributed and independent, though in a few intensively sampled rivers the residuals exhibited autocorrelation (data not shown). The model explained TP better than TN, which supports the finding that the relation between Q and TN in diffuse load-dominated rivers is relatively weak (Kauppila and Koskiaho 2003; Arheimer et al. 1996). The concentration of TP in agricultural rivers tends to be driven by erosion (Uusitalo et al. 2007), a process probably more related to Q than the mineralisation of organic matter that largely governs the losses of TN (Lepistö et al. 2008). Figure 2 shows examples on the observed and predicted concentrations in relation to Q for a river receiving both point and nonpoint loading (“U-shaped” pattern) and for a river receiving only nonpoint loading. Annual fluxes estimated using the above method differed on average for 1.5 % in the case TN and 5.7 % in the case of TP from those obtained using the periodic method (Rekolainen et al. 1991), in which each water quality observation represented the concentration from the midpoint of the preceding and current observation to the midpoint of the current and the next observation.
Fig. 2

Observed (dots) and predicted concentrations (solid line) of total nitrogen (C N ) as a function of flow (Q) in the Porvoonjoki and the Paimionjoki during period I (1985–1989) and period IV (2000–2006). Predicted concentrations are based on the nonlinear model C N = a + b/Q + cQ. The Porvoonjoki received point-source load from municipalities and diffuse load. In the river, the observed TN concentrations first decreased then increased with Q. The proportion of concentrations are inversely related to Q (i.e. b/Q) can be interpreted to represent point-source load. In period I, the point-source load of the river was 1210 kg day−1, giving a theoretical concentration of 7010 μg l−1 at 2 m3 s−1 (mean annual minimum Q in 1986–2006); the dashed line shows the dilution of such a load. The predicted TN concentration, 1930 + 12,100/2 + 15.6 • 2 = 8010 μg l−1 is close to the theoretical concentration. In period IV, the point-source load was decreased to 800 kg day−1, giving a theoretical concentration of 4620 μg l−1, while the predicted TN was 2910 + 3220/2 + 18.7 • 2 = 4550 μg l−1. The coefficient c increased from 15.6 in period I to 18.7 in period IV, suggesting an increase in diffuse TN load. In the Paimionjoki, the coefficient b was zero, which is in line with the fact that the point-source load was negligible in both periods. The coefficient c also increased in this river from period I to period IV

Calculating the trends

We estimated the trends both in the fluxes and concentrations of TP and TN. As for the fluxes, trends were estimated separately for the agricultural load and for the total riverine flux from which the different loading sources were not removed. Here, the fluxes were re-calculated using constants a, b and c specific for each river and period but the entire 22 years’ data on Q. Second, a linear regression equation was fitted on the fluxes in each period, which revealed a “flow-adjusted” trend in nutrient fluxes over periods. The analysis was performed for all the rivers with at least three periods of data (all rivers for TP and 16 for TN). Missing data was caused by problems in convergence of the estimation of the constants a, b and c (see “Estimating nutrient fluxes”).

Trends in the concentrations were detected by using a non-parametric partial Mann-Kendall test having Q as a covariate and accounting for seasonality (Libiseller and Grimvall 2002). As the concentrations are affected by all loading sources, not only agriculture, the results from this analysis were compared with the trends in total riverine nutrient fluxes.

Error caused by sampling strategy

To analyse how well the monitoring had captured different flow events, the hydrograph for each river was divided into ten classes. The first class included the days when Q was between the maximum Q and 90 % of the maximum (P 90 < Q ≤ P 100), the second class included the days when Q was between 80 and 90 % of the maximum (P 80 < Q ≤ P 90) and so on. The number of samples falling into these classes was calculated and the presence of a systematic trend in sampling towards higher or lower flow classes was estimated by dividing the mean Q of the sampling days by the mean Q over the entire period. To analyse the effect of sampling frequency on the flux estimates, the data was thinned out by removing the observations falling into class P 95 < Q ≤ P 100 and recalculating the flux as explained in “Estimating nutrient fluxes”. Then observations in the class P 90 < Q ≤ P 100 were removed and the flux was recalculated, and so on. Finally, the fluxes based on thinned and full data were compared and the mean difference and its 90 % uncertainty bounds (showing scatter among the rivers and periods) were calculated.

Explaining nutrient loads

To understand the load-generating factors, agricultural TN and TP loads were explained by variables characterising environmental and agricultural features of the catchments using linear multiple regression models. The parameters of the models were estimated for all the rivers and periods simultaneously to increase sample size and the precision of the estimates. Residual normality and other prerequisites of the regression analysis were evaluated by graphical and other descriptive methods. We used mixed models to account for the hierarchical structure of data in that repeated measures of TN and TP loads as well as many catchment characteristics (e.g. runoff, field use) were available. Here, restricted maximum likelihood approach was used. As the structure of variance-covariance matrix was unknown, we ran the models assuming three different variance-covariance structures: compound symmetry, first-order autoregressive and unstructured. In general, these variance-covariance structures produced similar sets of statistically significant variables for TP load, but in the case of TN load the results differed more often (in which case they are discussed below).

We introduced the explaining variables one by one in the model. The logic was to start with those catchment characteristics that affect nutrient load but are “exogenous” for this analysis, then proceed with variables characterising agriculture and finally include variables that potentially modify the agricultural effect on nutrient load. The “exogenous” variables included Lake% and Runoff, which are known to affect TP and TN load (Ekholm and Mitikka 2006). The agricultural features were divided into field use (Crop%, Grass% and Fallow%) and nutrient use (soil P status, P in manure, N balance). The modifying variables included TSumAnn, TSumWin, Peat% and OrgField%. As in the case of trend analysis, the statistical analysis included all the rivers with at least three periods of data (21 rivers for TP and 16 for TN). For such an exploratory use of multiple regression analysis, see Babbie (2007), Kukkonen et al. (2007), Pellikka et al. (2007) and Penttilä et al. (2006). All statistical analyses in this study were performed with SAS for Windows (SAS Institute Inc., Cary, NC).

Results and discussion

Effect of sampling strategy on nutrient loads

Despite the aim of the national river monitoring programme to concentrate sampling on high flow events, most of the samples were taken during low flows. An obvious reason for the tendency to sample at low flows is the peaky hydrograph in the rivers; in some rivers, the flow only reached the class P 90 < Q ≤ P 100 on a single day during the 22-year study period. However, the highest relative sampling frequency, i.e. the number of sampling days per number of days in a flow class, was found at flow class P 70 < Q ≤ P 80 (Fig. 3). More importantly, there appeared to be no major temporal change in the distribution of samples among the different flow classes; no trend existed in sampling day runoff divided by mean runoff of the period in any of the rivers (data not shown). As the actual sampling strategy appears to have remained the same over the periods, the trends in river fluxes were probably caused by other factors than changes in sampling.
Fig. 3

Number of flow observations (circles) and water quality samples (bars) in each flow class. Whiskers show the maximum and bars the mean number of water samples

When the nutrient loads were recalculated with thinned data, the loads were higher than when all the samples were included. For TN, the mean difference exceeded 10 % when the samples representing P 85 > Q were excluded. In turn, for TP a 10 % difference occurred only at about P 70 > Q (Fig. 4). Even though the difference between the minimum and maximum deviation was large, the 90 % confidence bound was relatively narrow as long as there were water quality samples in the P 75 > Q class. Because sampling in all the rivers in all the periods fulfilled this requirement, the general uncertainty of the calculations is relatively low.
Fig. 4

Mean difference (line) and its 90 % uncertainty bounds (shaded area) of TP and TN fluxes as a function of samples used in the flux calculation (P 100 = all the samples used, P 95 = 95 % of the samples used, and so on). Positive values indicate overestimation of the fluxes

Nutrient fluxes

Mean TP flux was lower in rivers of western Finland than in southern Finland, but due to a large river to river variability, the difference was not statistically significant (Table 2). Mean TP flux decreased from period I to period IV, in part due to a concomitant decrease in the mean runoff. In all the periods, the riverine fluxes of TP were the largest in the two rivers having the highest Field% (the Paimionjoki and the Aurajoki) and the lowest in the rivers with the highest Lake% (the Karjaanjoki, the Kiskonjoki, the Eurajoki and the Ähtävänjoki) and in the Kiiminginjoki with the clearly lowest Field% (1.3 %, Fig. 5). As for TP, the fluxes of TN were positively related to Field% and inversely to Lake%. However, TN fluxes also reflected wastewater load. While the fluxes of TN in wastewater loaded rivers decreased from period I to period IV, the opposite was true for rivers in western Finland, where the fluxes of TN were higher in period IV than in the other periods (Fig. 5).
Fig. 5

Source apportionment of the riverine TP and TN fluxes

The riverine nutrient fluxes were divided into original sources assuming no retention. The proportion of municipal wastewater loads in riverine TP fluxes decreased from period I to period IV, exceeding 10 % of the total riverine TP flux only in the Porvoonjoki (periods I–III) and the Karjaanjoki (all periods, Fig. 5). Because N is less efficiently removed from wastewaters, its share was higher (up to 29 % in the Porvoonjoki, period I), yet it also decreased over time. Nutrient input from industry was negligible, except in the Karjaanjoki (10–18 % of the TP fluxes; 7–12 % of the TN fluxes) and the Eurajoki (5–8 % for TP; 3–8 % for TN).

Diffuse load dominated in all the rivers (Fig. 5). Scattered settlements accounted for more of the riverine TP fluxes than municipalities, accounting for 20 % of the riverine TP fluxes in the Vantaanjoki, the Karjaanjoki and the Ähtävänjoki. For TN, the percentages were lower due to the higher share of municipalities. Forestry was responsible for more than 10 % of the riverine TP fluxes in the Karjaanjoki, the Ähtävänjoki and the Kiiminginjoki (for TN the share was lower). The “agricultural load”, i.e. the fraction left when the loads from the above sources were subtracted from the total riverine nutrient flux, was the largest source of both nutrients in all the rivers except the Karjaanjoki. For TP, agricultural load was clearly highest in the Paimionjoki and the Aurajoki, whereas for TN, the differences among the rivers were lower.

Trends in nutrient loads and concentrations

The flow-adjusted agricultural TP load decreased in most rivers (Fig. 6). The decrease was most pronounced in rivers discharging into the Bothnian Bay, up to 30 %, when estimated from the linear change from period I to period IV. Little change was found in the Paimionjoki and the Aurajoki discharging into the Archipelago Sea. By contrast, the flow-adjusted agricultural TN loads generally increased, by up to 50 % in the rivers discharging into the Bothnian Bay (Fig. 7). Only in the Aurajoki and the Lapväärtinjoki was the TN load decreasing. For the Aurajoki, this result may be due to one high TN concentration observed during period I.
Fig. 6

Development in flow-adjusted agricultural TP load (kg km−2 y−1) by rivers over periods. The bars show the estimated load and the line the linear regression fit from period I to period IV

Fig. 7

Development in flow-adjusted agricultural TN load by rivers (kg km−2 y−1) over periods

The flow-adjusted trends in total riverine nutrient fluxes (i.e. the fluxes from which no loading sources were subtracted) were in accordance with those obtained for nutrient concentrations with the non-parametric test; a river having a significant trend in the concentration according to the non-parametric test also identified more than a 10 % change in the flux (Fig. 8). Few rivers with a high point-source load formed exceptions here. They had high nutrient concentrations during low Q, which did not have a significant effect on annual nutrient fluxes. In other countries surrounding the Baltic Sea, decreasing rather than increasing trends in agricultural TN loads are typical, especially in Denmark and Sweden (Stålnacke et al. 2014; Windolf et al. 2012).
Fig. 8

Relationship between trends in riverine fluxes (y-axis) and concentrations (x-axis). The MK value refers to the Mann-Kendall statistics, with dark diamonds indicating a statistically significant change in the concentrations

Explaining agricultural nutrient load

We started explaining agricultural TP load with a model that had only Region (southern Finland and western Finland) as an explaining variable (model 1, Table 6). Introducing other variables one-by-one in the model (also testing for their interactions) resulted in a model having Region, Lake%, Runoff, Field%, interaction between Field% and Region, and Grass% as significant variables (model 6, Table 6). Other field use types (Crop%, Fallow%), P in manure and STP did not emerge as significant in any combinations.
Table 6

Effects of catchment and agricultural characteristics on agricultural load of total phosphorus (kg km−2 y−1)

Model

Intercept

Lake%

Runoff

Field%

Grass%

AIC

BIC

South

West

South

West

1

24***

17***

     

557

568

2

45***

30***

−3.3***

    

528

538

3

18***

−6.9

−2.9***

104***

   

495

506

4

−30***

−23**

−1.3***

101***

1.4***

  

475

486

5

−35***

−18**

−1.3***

105***

1.6**

0.77***

 

469

479

6

−36***

−24**

−1.3***

96***

1.6*

1.0***

0.19*

468

479

Only models with statistically significant effects are presented. Field% and Lake% refer to the proportion of fields and lakes, respectively, in the catchment area; Runoff is given in metres; Grass% refers to the proportion grassed land in the agricultural land

AIC Akaike Information Criteria, BIC Bayesian Information Criteria

*p < 0.05, **p < 0.01, ***p < 0.001. Results using an unstructured variance-covariance structure

The negative slope for Lake% (−1.3; model 6) is within realistic bounds, as is the positive slope for Runoff (96). The slope of Field% was 1.6 in southern Finland and 1.0 in western Finland. The fields in southern Finland have a finer soil texture and more sloping fields, which probably makes them more sensitive to erosion and increases TP losses. On the other hand, the presence of acid sulphate soils may decrease the TP losses in western Finland (Vuorenmaa et al. 2002). Using model 5 (Table 6), a catchment with 100 % fields in southern Finland would have a TP load of 150 kg km−2 y−1 (Runoff = 0.3 m). As this value includes the natural background, it corresponds relatively well to the specific agricultural loss (110 kg km−2 y−1) found for three intensively monitored small agricultural catchments in southern Finland between 1981 and 1997 (Vuorenmaa et al. 2002). The size of natural background for TP losses is about 5 kg km−2 y−1 in Finland (Mattsson et al. 2003; Kortelainen et al. 2006). However, this estimate is based on semi-natural forested catchments, which may underestimate natural background in more fertile agricultural soils. A similar “reality check” cannot be made for agriculture in western Finland, because this region does not have monitored agricultural catchments (other than the rivers examined here).

That agricultural TP load increased with Grass% may appear surprising, because grass should protect the soil from P losses by erosion (Uusitalo 2004). Perhaps grass increases the losses of dissolved P; these have been found to be high in grassland receiving surface application of manure, for example (Uusi-Kämppä and Heinonen-Tanski 2008).

Grass% decreased from period III to period IV, which is in line with the decreasing agricultural TP losses (see “Trends in nutrient loads and concentrations”). Apart from the changes in Grass%, the regression analysis did not reveal the causes behind the TP decrease. In western Finland, some of the decrease might be related to the fact that STP decreased from period III to period IV in the region, in which case the correlation between STP, Runoff and the field use types (Table 4) had prevented the emergence of STP in the regression models. The correlation between Grass% and other explaining variables also hampers solid conclusions on the effect of Grass% on TP load.

As in the case of TP, the agricultural TN load carried by the rivers was lower in western than in southern Finland, and the load increased with Field% and Runoff and decreased with Lake% (Table 7). Other variables significantly related to TN load were N balance and TSumAnn. In addition, with autoregressive and compound symmetry variance-covariance structures, Grass% became inversely related to TN load. Surprisingly, N balance was inversely related to TN load. Because Grass% and N balance were strongly correlated with each other (and Grass% with several other variables, as noted above), it is difficult to draw conclusions on the relationships. TSumAnn was positively correlated to TN load (model 7; Table 7), which suggests that the increasing soil temperature has promoted mineralization and N losses. Here, too, the slope for Field% in southern Finland was approximately at the same level as found elsewhere: the specific TN load for field land would be about 1700 kg km−2 y−1 (model 5; Table 7), while 1500 kg km−2 y−1 was found for the small agricultural catchments (Vuorenmaa et al. 2002).
Table 7

Effects of catchment and agricultural characteristics on agricultural load of total nitrogen (kg km−2 y−1)

Model

Intercept

Lake%

Runoff

Field%

N balance

TSumAnn

AIC

BIC

South

West

South

West

1

556***

290***

      

733

741

2

816***

411***

−43***

     

708

716

3

888***

300***

−23*

−54*

    

696

704

4

580***

−84

−6.5

−55**

891***

   

677

685

5

79

−187

−9.7

−32*

1001***

12.7***

  

663

671

6

119

−132

−3.7

−32*

1350**

12.6**

−2.1**

 

653

661

7

−377

−8.0

762*

2.9

−0.8

1.49***

348

350

Only models with statistically significant effects are presented. Field% and Lake% refer to the proportion of fields and lakes, respectively, in the catchment area; Runoff is given in metres. TSumAnn is the annual temperature sum and is tested only for western Finland

AIC Akaike Information Criteria, BIC Bayesian Information Criteria

*p < 0.05, **p < 0.01, ***p < 0.001. Results using an unstructured variance-covariance structure in mixed models

Conclusions

We estimated agricultural nutrient load using river monitoring data that contained a lot of noise, in part due to infrequent sampling but also to weather-driven hydrological processes. Total riverine nutrient fluxes were estimated using an empirical method that predicted the concentrations from a relation between concentration and Q, which allowed accounting for changes in the fluxes caused by variation in hydrology. Agricultural nutrient load (incl. natural background) was separated from the total riverine nutrient flux by subtracting other loading sources from the riverine flux. Despite the coarse approach used, our load estimates for agricultural TP and TN load corresponded relatively well with those found earlier for more intensively sampled small agricultural catchments in Finland. Agricultural TP load decreased, whereas TN load generally increased between 1985 and 2006, despite the agri-environmental measures implemented since the start of Finnish Agri-Environmental Programme in 1995. Available data on catchment and agricultural characteristics did not fully explain the observed pattern, but a decrease in grassed area may have partly affected the TP and TN loads and increased mineralisation the TN load. A Finnish governmental aim was to decrease the agricultural load of TP and TN by 30 % from the level of the early 2000s up to the year 2015. Although we were able to show the direction of agricultural load and link it to some explaining factors, it remains challenging to precisely quantify the change in agricultural load to verify whether such political goals have been met.

Notes

Acknowledgments

We would like to thank the Ministry of the Environment, the Finnish Field Drainage Association, Maa- ja Vesitekniikan Tuki ry and the EU Erasmus programme for financially supporting this study. We are grateful to Jukka Aroviita for commenting on the manuscript and Hannu Sirviö, Niina Kotamäki and Heli Rita for the help with Proc Mixed.

Compliance with ethical standards

We see no ethical issues or potential conflicts of interest in this research.

References

  1. Aakkula, J., Kuussaari, M., Rankinen, K., Ekholm, P., Heliölä, J., Hyvönen, T., et al. (2011). Follow-up study on the impacts of agri-environmental measures in Finland. Workshop on evaluating agri-environmental policies. Braunschweig, Germany, 20–22 June 2011. In Evaluation of Agri-Environmental Policies Selected Methodological Issues and Case Studies (pp. 111–127). OECD Publishing.Google Scholar
  2. Arheimer, B., Andersson, L., & Lepistö, A. (1996). Variation of nitrogen concentration in forest streams—influences of flow, seasonality and catchment characteristics. Journal of Hydrology, 179, 281–304.CrossRefGoogle Scholar
  3. Babbie, E. R. (2007). The practice of social research (11th edition): Thomson Wadsworth.Google Scholar
  4. Bouraoui, F., & Grizzetti, B. (2011). Long term change of nutrient concentrations of rivers discharging in European seas. Science of the Total Environment, 409, 4899–4916.CrossRefGoogle Scholar
  5. Cherry, K. A., Shepherd, M., Withers, P. J. A., & Mooney, S. J. (2008). Assessing the effectiveness of actions to mitigate nutrient loss from agriculture: a review of methods. Science of the Total Environment, 406(1–2), 1–23.CrossRefGoogle Scholar
  6. EEC. (1992). Council Regulation (EEC) No 2078/92 of 30 June 1992 on agricultural production methods compatible with the requirements of the protection of the environment and the maintenance of the countryside. Official Journal of the European Communities, L215, 85–90.Google Scholar
  7. Ekholm, P., & Mitikka, S. (2006). Agricultural lakes in Finland—current water quality and trends. Environmental Monitoring and Assessment, 116, 111–135.CrossRefGoogle Scholar
  8. Ekholm, P., Granlund, K., Kauppila, P., Mitikka, S., Niemi, J., Rankinen, K., et al. (2007). Influence of EU policy on agricultural nutrient losses and the state of receiving surface waters in Finland. Agricultural and Food Science, 16, 282–300.CrossRefGoogle Scholar
  9. Granlund, K., Räike, A., Ekholm, P., Rankinen, K., & Rekolainen, S. (2005). Assessment of water protection targets for agricultural nutrient loading in Finland. Journal of Hydrology, 304(1–4), 251–260.CrossRefGoogle Scholar
  10. Jarvie, H. P., Withers, P. J. A., Bowes, M. J., Palmer-Felgate, E. J., Harper, D. M., Wasiak, K., et al. (2010). Streamwater phosphorus and nitrogen across a gradient in rural–agricultural land use intensity. Agriculture, Ecosystems and Environment, 135, 238–252.CrossRefGoogle Scholar
  11. Johnes, P. J. (2007). Uncertainties in annual riverine phophorus load estimation: impact of load estimation methodology, sampling frequency, baseflow index and catchment population density. Journal of Hydrology, 332, 241–258.CrossRefGoogle Scholar
  12. Kajosaari, E. (1965). On the washing of phosphorus into watercourses (in Finnish with an English summary). Vesitalous, 1965(3), 17–25.Google Scholar
  13. Kauppi, L. (1979). Phosphorus and nitrogen input from rural population, agriculture and forest fertilization to watercourses. Publications of the Water Research Institute, 34, 35–46.Google Scholar
  14. Kauppi, L. (1984). Contribution of agricultural loading to the deterioration of surface waters in Finland. Publications of the Water Research Institute, 57, 24–30.Google Scholar
  15. Kauppi, L., Knuuttila, S., Sandman, K., Eskonen, K., Luokkanen, S., & Liehu, A. (1990). Role of landuse in the occurrence of blue-green algal blooms. [Verh. Internat. Verein. Limnol.]. Verhandlungen der internationationale Vereinigung für theoretische und angewandte Limnologie, 24, 671-676.Google Scholar
  16. Kauppila, P., & Koskiaho, J. (2003). Evaluation of annual loads of nutrients and suspended solids in Baltic rivers. Nordic Hydrology, 34(3), 203–220.Google Scholar
  17. Kenttämies, K., & Mattson, T. (2006). Metsätalouden vesistökuormitus, MESUVEprojektin loppuraportti. Suomen ympäristökeskus.Google Scholar
  18. Kortelainen, P., Mattsson, T., Finér, L., Ahtiainen, M., Saukkonen, S., & Sallantaus, T. (2006). Controls on the export of C, N, P and Fe from undisturbed boreal catchments, Finland. Aquatic Sciences, 68, 453–468.CrossRefGoogle Scholar
  19. Kronvang, B., Bechmann, M., Lundekvam, H., Behrendt, H., Rubaek, G. H., Schoumans, O. F., et al. (2005). Phosphorus losses from agricultural areas in river basins: effects and uncertainties of targeted mitigation measures. Journal of Environmental Quality, 34(6), 2129–2144.CrossRefGoogle Scholar
  20. Kukkonen, M., Rita, H., Hohnwald, S., & Nygren, A. (2007). Treefall gaps of certified, conventionally managed and natural forests as regeneration sites for neotropical timber trees in northern Honduras. Forest Ecology and Management, 255, 2163–2176.CrossRefGoogle Scholar
  21. Laukkanen, M., & Nauges, C. (2014). Evaluating greening farm policies: a structural model for assessing agri-environmental subsidies. Land Economics, 90(3), 458–481.CrossRefGoogle Scholar
  22. Lepistö, A., Kortelainen, P., & Mattson, T. (2008). Increased organic C and N leaching in a northern boreal river basin in Finland. Global Biogeochemical Cycles, 22, 1–10.CrossRefGoogle Scholar
  23. Libiseller, C., & Grimvall, A. (2002). Performance of partial Mann-Kendall tests for trend detection in the presence of covariates. Environmetrics, 13(1), 71–84.CrossRefGoogle Scholar
  24. Lilja, H., Uusitalo, R., Yli-Halla, M., Nevalainen, R., Väänänen, T., & Tamminen, P. (2009). Finnish soil database user’s guide. MTT Tiede 6 (in Finnish.).Google Scholar
  25. Mäntylahti, V. (2003). Suomen peltojen viljavuus 1986-1990, 1991-1995, 1996-2000. Viljavuuspalvely Oy.Google Scholar
  26. Mattsson, T., Finer, L., Kortelainen, P., & Sallantaus, T. (2003). Brook water quality and background leaching from unmanaged forested catchments in Finland. Water, Air, and Soil Pollution, 147, 275–297.CrossRefGoogle Scholar
  27. Niemi, J., & Ahlstedt, J. (Eds.). (2012). Finnish agriculture and rural industries 2012 (MTT Taloustutkimus julkaisuja 112a): MTT Economic Research, Agrifood Research Finland.Google Scholar
  28. Pellikka, J., Rita, H., Lindén, H., & Svensberg, M. (2007). Motives for voluntary wildlife monitoring in Finland. Wildlife Biology, 13, 1–10.CrossRefGoogle Scholar
  29. Penttilä, R., Lindgren, M., Miettinen, O., Rita, H., & Hanski, I. (2006). Consequences of forest fragmentation for polyporous fungi at two spatial scales. Oikos, 114(2), 225–240.CrossRefGoogle Scholar
  30. Puustinen, M., Tattari, S., Koskiaho, J., & Linjama, J. (2007). Influence of seasonal and annual hydrological variations on erosion and phosphorus transport from arable areas in Finland. Soil & Tillage Research, 93(1), 44–55.CrossRefGoogle Scholar
  31. Rask, M., Olin, M., & Ruuhijärvi, J. (2010). Fish-based assessment of ecological status of Finnish lakes loaded by diffuse nutrient pollution from agriculture. Fisheries Management and Ecology, 17(2), 126–133.CrossRefGoogle Scholar
  32. Rekolainen, S., Posch, M., Kämäri, J., & Ekholm, P. (1991). Evaluation of the accuracy and precision of annual phosphorus load estimates from 2 agricultural basins in Finland. Journal of Hydrology, 128(1–4), 237–255.CrossRefGoogle Scholar
  33. Salo, T., & Turtola, E. (2006). Nitrogen balance as an indicator of nitrogen leaching in Finland. Agriculture, Ecosystems & Environment, 113(1–4), 98–107. doi: 10.1016/j.agee.2005.09.002.CrossRefGoogle Scholar
  34. Stålnacke, P., Aakerøy, P. A., Blicher-Mathiasen, G., Iital, A., Jansons, V., Koskiaho, J., et al. (2014). Temporal trends in nitrogen concentrations and losses from agricultural catchments in the Nordic and Baltic countries. Agriculture, Ecosystems & Environment, 198, 94–103. doi: 10.1016/j.agee.2014.03.028.CrossRefGoogle Scholar
  35. Sucksdorff, Y., & Teiniranta, R. (2001). Land cover and land use mapping in Finland. In G. Groom & T. Reed (Eds.), Strategic landscape monitoring for the Nordic countries (pp. 523) (pp. 89–93). Copenhagen: TemaNord. Nordic Council of Ministers.Google Scholar
  36. Sutela, T., & Vehanen, T. (2009). Responses of fluvial fish assemblages to agriculture within the boreal zone. Fisheries Management and Ecology, 17, 141–145.CrossRefGoogle Scholar
  37. Tamminen, T., & Andersen, T. (2007). Seasonal phytoplankton nutrient limitation patterns as revealed by bioassays over Baltic Sea gradients of salinity and eutrophication. Marine Ecology-Progress Series, 340, 121–138.CrossRefGoogle Scholar
  38. Turtola, E. (2007). The Agri-Environmental Program in Finland: Effects on nutrient loading from agriculture into surface waters in 2000-2006—preface. Agricultural and Food Science, 16(4), 279–281.CrossRefGoogle Scholar
  39. Uusi-Kämppä, J., & Heinonen-Tanski, H. (2008). Evaluating slurry broadcasting and injection to ley for phosphorus losses and fecal microorganisms in surface runoff. Journal of Environmental Quality, 37, 2339–2350.CrossRefGoogle Scholar
  40. Uusitalo, R. (2004). Potential bioavailability of particulate phosphorus in runoff from arable clayey soils (Agrifood Research Reports 53). Jokioinen: MTT.Google Scholar
  41. Uusitalo, R., & Jansson, H. (2002). Dissolved reactive phosphorus in runoff assessed by soil extraction with an acetate buffer. Agricultural and Food Science in Finland, 11(4), 343–353.Google Scholar
  42. Uusitalo, R., Turtola, E., & Lemola, R. (2007). Phosphorus losses from a subdrained clayey soil as affected by cultivation practices. Agricultural and Food Science, 16, 352–365.CrossRefGoogle Scholar
  43. Vagstad, N., Stålnacke, P., Andersen, H. E., Deelstra, J., Jansons, V., Kyllmar, K., et al. (2004). Regional variations in diffuse nitrogen losses from agriculture in the Nordic and Baltic regions. Hydrology and Earth System Sciences, 8(4), 651–662.CrossRefGoogle Scholar
  44. Valve, H., Åkerman, M., & Kaljonen, M. (2013). ‘You only start filling in the boxes’: natural resource management and the politics of plan-ability. Environment & Planning A, 45, 2084–2099.CrossRefGoogle Scholar
  45. Varanka, S., & Luoto, M. (2011). Environmental determinants of water quality in boreal rivers based on partitioning methods. River Research and Applications, 28, 1034–1046.CrossRefGoogle Scholar
  46. Vuorenmaa, J., Rekolainen, S., Lepisto, A., Kenttamies, K., & Kauppila, P. (2002). Losses of nitrogen and phosphorus from agricultural and forest areas in Finland during the 1980s and 1990s. Environmental Monitoring and Assessment, 76(2), 213–248.CrossRefGoogle Scholar
  47. Vuorinen, J., & Mäkitie, O. (1955). The method of soil testing in use in Finland. Agrogeological Publications, 63, 1-44.Google Scholar
  48. Wartiovaara, J. (1975). Jokien ainevirtaamista Suomen rannikolla [English abstract: Amounts of Substances Discharged by Rivers Off the Coast of Finland] (Vesientutkimuslaitoksen julkaisuja - Publications of the Water Research Institute 13): Vesihallitus-National Board of Waters.Google Scholar
  49. Windolf, J., Blicher-Mathiesen, G., Carstensen, J., & Kronvang, B. (2012). Changes in nitrogen loads to estuaries following implementation of governmental action plans in Denmark: a paired catchment and estuary approach for analysing regional responses. Environmental Science & Policy, 24, 24–33.CrossRefGoogle Scholar
  50. Yli-Viikari, A., Hietala-Koivu, R., Huusela-Veistola, E., Hyvönen, T., Perälä, P., & Turtola, E. (2007). Evaluating agri-environmental indicators (AEIs)—use and limitations of international indicators at national level. Ecological Indicators, 7, 150–163.CrossRefGoogle Scholar

Copyright information

© Springer International Publishing Switzerland 2015

Authors and Affiliations

  • Petri Ekholm
    • 1
  • Katri Rankinen
    • 1
  • Hannu Rita
    • 2
  • Antti Räike
    • 1
  • Heidi Sjöblom
    • 1
  • Arjen Raateland
    • 1
  • Ljudmila Vesikko
    • 1
  • José Enrique Cano Bernal
    • 1
  • Antti Taskinen
    • 1
  1. 1.Finnish Environment Institute SYKEHelsinkiFinland
  2. 2.University of HelsinkiHelsinkiFinland

Personalised recommendations