Effects of climate change and agricultural adaptation on nutrient loading from Finnish catchments to the Baltic Sea.

Science of the Total Environment 529 (2015) 168–181

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

Effects of climate change and agricultural adaptation on nutrient loading from Finnish catchments to the Baltic Sea Inese Huttunen a,⁎, Heikki Lehtonen b, Markus Huttunen a, Vanamo Piirainen a, Marie Korppoo a, Noora Veijalainen a, Markku Viitasalo a, Bertel Vehviläinen a a b

Finnish Environment Institute SYKE, P.O. Box 140, FIN-00251 Helsinki, Finland Natural Resources Institute Finland LUKE, Latokartanonkaari 9, FIN-00790 Helsinki, Finland

H I G H L I G H T S • • • •

Climate change is expected to increase winter runoff and slightly increase the annual runoff sum. Climate change is expected to affect crop growth, level of fertilization and production intensity. Combined climate and agricultural changes are expected to increase nitrogen and phosphorus loading to the Baltic Sea. Higher yields are needed for reducing nutrient balances, and more land should be allocated to water protection programs.

a r t i c l e

i n f o

Article history: Received 4 November 2014 Received in revised form 12 May 2015 Accepted 14 May 2015 Available online xxxx Editor: Eddy Y. Zeng Keywords: Climate change Nutrient loading Baltic Sea Agricultural scenarios Water quality modeling VEMALA

a b s t r a c t Climate change is expected to increase annual and especially winter runoff, shorten the snow cover period and therefore increase both nutrient leaching from agricultural areas and natural background leaching in the Baltic Sea catchment. We estimated the effects of climate change and possible future scenarios of agricultural changes on the phosphorus and nitrogen loading to the Baltic Sea from Finnish catchments. In the agricultural scenarios we assumed that the prices of agricultural products are among the primary drivers in the adaptation to climate change, as they affect the level of fertilization and the production intensity and volume and, hence, the modeled changes in gross nutrient loading from agricultural land. Optimal adaptation may increase production while supporting appropriate use of fertilization, resulting in low nutrient balance in the fields. However, a less optimal adaptation may result in higher nutrient balance and increased leaching. The changes in nutrient loading to the Baltic Sea were predicted by taking into account the agricultural scenarios in a nutrient loading model for Finnish catchments (VEMALA), which simulates runoff, nutrient processes, leaching and transport on land, in rivers and in lakes. We thus integrated the effects of climate change in the agricultural sector, nutrient loading in fields, natural background loading, hydrology and nutrient transport and retention processes. © 2015 Elsevier B.V. All rights reserved.

1. Introduction According to modeling results, future climatic conditions in northern Europe will be warmer and wetter, and temperature increase will be higher in northern than in southern Europe (IPCC, 2007, 2013). According to Ruosteenoja et al. (2010) the annual average temperature in Finland will increase by 2–6 °C by the end of the 21st century. In winter the increase will be 3–9 °C, in summer 1–5 °C. Furthermore, the annual precipitation has been predicted to increase by 12–22% (10–40% in winter, 0–20% in summer). This means a 30–45 days longer growing season by 2100. The temperature sum during the growing period is forecasted to increase from 1300 to 1900 degree days in southern Finland, from ⁎ Corresponding author. E-mail address: Inese.Huttunen@ymparisto.fi (I. Huttunen).

http://dx.doi.org/10.1016/j.scitotenv.2015.05.055 0048-9697/© 2015 Elsevier B.V. All rights reserved.

1100 to 1600 degree days in central Finland and from 900 to 1200 degree days in the north. Since the summer precipitation may increase only slightly, increasing temperature stress and early summer droughts, which are considered as one of the main causes of low crop yields currently in Finland (Peltonen-Sainio et al., 2009), may become more common. On the other hand, climate change also implies increasing frequency of rainy days and heavy rainfall events. Decreased length of the thermal winter, reduced snow cover and number of frost days are also probable consequences of climate change (Peltonen-Sainio et al., 2009; Höglind et al., 2013). 1.1. Climate change impact on nutrient loading Recent publications suggest that nutrient loading from land to water bodies, and hence to the Baltic Sea, will increase due to climate change

I. Huttunen et al. / Science of the Total Environment 529 (2015) 168–181

(Arheimer et al., 2012; Eriksson Hägg et al., 2014). The estimates vary, and depend on the type of models and methodologies used in the studies. Arheimer et al. (2012) concluded that the simulation of nutrient loads under future climate conditions indicates that the nitrogen inflow may be reduced, but the phosphorus inflow may be slightly increased to the marine basins. However, some climate projections indicate the opposite. Eriksson Hägg et al. (2014) concluded that in the catchments draining into the northern sub-basins of the Baltic Sea (Bothnian Bay, Bothnian Sea and Gulf of Finland), the potential impact of climate change on nutrient loads is higher than the changes caused by a lifestyle shift. In a small scale study of nitrogen leaching, Rankinen et al. (2013) concluded that adjusting the N balance at a parcel level was more important than the vegetation cover. In the future, adaptation at the crop level is potentially an efficient way to manage the nutrient loading risk. 1.2. Climate change impact on crop growth Crop growth is one of the key drivers in the development of agricultural production, land use and nutrient balances. Rötter et al. (2012) summarized the impacts of climate change on the most relevant agrometeorological indicators of crop yields. They concluded that a longer growing season and higher effective temperature sum are likely to increase the crop yield potential, whereas an increasing number of dry days and more frequent adverse weather events are factors that may significantly affect crop yield levels and increase their inter-annual variation. On the other hand, Peltonen-Sainio et al. (2009) concluded that an increase in yield potential requires crop cultivars capable of utilizing a longer growing period. Höglind et al. (2013) simulated grass yields in various locations of northern Europe under the A1B climate scenario. The yield increase due to climate change was estimated at + 11% for grass production in southern Finland, while central Finland (Kuopio region) would reach approximately 20% increase in grass yields with the assumption of optimal overwintering conditions and current CO2 level. Rötter et al. (2013) estimated yields of cereal crops in Finland for the 21st century under the SRES A2 climate scenario. The results indicated decreasing yields of current cultivars in the A2 and increasing yields in the B1 and A1B climate scenarios. Moreover, the yield potential of major crops under climate change decreases under A2, but might be sustained close to the current level even in the A2 climate scenario if new cultivars better tuned to a longer growing season are adopted. 1.3. Impact of agricultural change on crop growth Crop yields are also likely to affect land use and nutrient balances, which in turn are among the key drivers of nutrient loading. Nutrient balance, i.e. the difference between the annual (N and P) fertilization level per ha and the amount of N and P nutrients per ha harvested, has been decreasing significantly in all regions of Finland since 1995. The average nitrogen balance per ha of farmland in Finland has decreased from 94 kg/ha in 1990 to 47 kg/ha in 2007–2013. The average phosphorus balance has decreased from 28.6 kg/ha in 1990 to 2.9 kg ha−1 in 2007–2013 (Salo et al., 2007; Salo and Lemola, 2014). Phosphorus balance has decreased considerably since 2000 because of specific limits imposed on phosphorus fertilization, on the basis of the soil P status of each field parcel, and, most probably, because of significantly increased prices of P fertilizers. Few previous studies (Abler et al., 2002; Hattermann et al., 2007) have combined national and regional level agricultural scenarios with detailed nutrient modeling. In this paper, we estimate the effects of climate change and possible future scenarios of agricultural changes on the phosphorus (P) and nitrogen (N) loading to the Baltic Sea from Finnish catchments. Our objectives are: (1) to present a method to assess the effects of changes in climate and agricultural practices on nutrient loading by coupling two models: the economic agricultural sector model DREMFIA and the nutrient loading model VEMALA and (2) to show and discuss the simulated results of the TP and TN loading trends

169

in the different sub-basins of the Baltic Sea under different climate and agricultural scenarios.

2. Methods and materials 2.1. Study area Finnish catchments draining to the Baltic Sea cover 301,300 km2. The Baltic Sea around the Finnish territory is divided into the following marine sub-basins (Fig. 1): Gulf of Finland (GF), Archipelago Sea (AS), Bothnian Sea (BS), and Bothnian Bay (BB). In our study we also considered the Vuoksi river catchment (VUO, Fig. 1), which contributes to the Gulf of Finland through the Neva River. The annual discharge to the Baltic Sea from Finnish catchments in 2006 was 2050 m3 s−1, which represented 15% of the total riverine flow to the Baltic Sea (HELCOM, 2011). The annual riverine loading of total nitrogen (TN) and total phosphorus (TP) from Finnish catchments in 2006 was 79,000 TN t yr−1 and 3490 TP t yr−1, which represented 12% of the total riverine TN and TP loading to the Baltic Sea.

2.2. Climate change scenario Three climate scenarios were used in this study. All scenarios use the emission scenario A1B, in which greenhouse gas emissions are rather high in the first part of the century and start to decrease from 2050 (IPCC, 2000). This emission scenario produces intermediate greenhouse gas concentrations compared to other SRES scenarios by the end of the 21st century. The importance in the choice of the emission scenario increases by the end of the century, but until 2050 the differences between the greenhouse gas concentrations produced by different emission scenarios are small (IPCC, 2000). One of the climate scenarios is the Mean A1B scenario, which is an average scenario calculated from 19 global climate models for Finland by the Finnish Meteorological Institute (Ruosteenoja et al., 2007). The other two scenarios are from Regional Climate Models (RCMs) obtained from the ENSEMBLES data base (van der Linden and Mitchell, 2009). The scenarios were HIRHAM-ARPEGE-A1B (referred to as the “dry” scenario since the precipitation increase is small) and RCA3-HadCM3-A1B (referred to as the “wet” scenario since the precipitation increase is large). These two scenarios were selected from a larger ensemble of RCM scenarios to represent the uncertainty associated with climate change. The greatest difference in the results was obtained with regard to changes in precipitation and therefore the RCM scenarios were selected from the high and low ends of the range of projected precipitation change. The climate scenarios were calculated as monthly changes in average temperature (°C) and precipitation (%) for the periods 2010–39 and 2040–69 compared to the control period 1971–2000 (Ruosteenoja et al., 2007). The results were provided as a 2.5 degree grid for Finland and surrounding areas for the Mean A1B scenario and as a 0.25 degree grid for the RCM scenarios. The results from the four closest grids were used to calculate values for each sub-catchment of the hydrological model. Precipitation increase by 2040–69 compared to 1971–2000 was 11.5% in the Mean A1B scenario, 4.7% in the dry scenario and 16.2% in the wet scenario. Corresponding temperature increases were 3.2, 2.6 and 2.5 °C, respectively.

2.3. Agricultural scenarios In addition to the climate change scenarios, we chose a baseline and three socio-economic scenarios of agricultural adaptations and change for 2014–2050. More details of the baseline scenario are given in Lehtonen (2013). Adaptation scenarios are described in full in Lehtonen (2015).

170


Fig. 1. Finnish catchments and their contribution to the Baltic Sea sub-basins Bothnian Bay, Bothnian Sea, Archipelago Sea, Gulf of Finland and Vuoksi (left), and the location of the Baltic Sea basins within the region (right).

2.3.1. Mean A1B scenario and no changes in agriculture This scenario is used for TN loading runs to estimate the effect of only climate change on TN loading. Crop distribution and fertilizer use are exactly the same as during the decade 2001–2010. 2.3.2. The Baseline In this “business as usual” scenario unchanged yields and agricultural policy are assumed. Future prices of agricultural outputs are not assumed to change significantly from the prices of 2008–2013; they were projected by the OECD-FAO agricultural outlook 2013 (www-agri-outlook.org). 2.3.3. Successful adaptation, very high prices (SuA_VHP) In this optimistic scenario yields increase by 30% for cereals and grasslands (0.9% annually) and by 60% for oilseeds and winter cereals in 2013–2050. Prices, policies, research and development imply effective adaptation, including new cultivars suitable for a longer growing season. Higher crop yield levels are also driven by increased fungicide use for cereals, and liming which increases the soil pH. Drainage and soil structure investments are also assumed to be encouraged by higher crop prices and subsidies. Drought problems are mitigated because of new cultivars better suited to the altered climate. Furthermore, in this scenario cereal prices increase by 30%, meat prices by 15% and milk product prices by 10%. Prices of nitrogen fertilizers increase only by 10%. Nitrogen fertilization increases by almost 30% due to the nutrient needs of plants, which increase with the increasing yields. The high increase in crop yields in this scenario is mainly driven by 30% higher EU level cereal prices already by 2030, compared to the baseline. High prices are also combined with policies allowing higher fertilization. However, due to the higher yields, nutrient balances of N and P per crop decrease slightly and remain close to the baseline level (2007–2013 levels). Reaching 30% higher crop yields compared to the baseline average crop yields for 1995–2012 is challenging, but possible. Increased fungicide use for cereals may increase yields by 10–15% (Purola, 2013). Liming, which increases soil pH from mean levels of 5.5–6.5 up to 6.0–7.0, could provide another 10–15% increase in yields. It is uncertain to what extent increasing nitrogen fertilization, even though implying a non-increasing nitrogen balance, is compatible with future agri-

environmental policies and e.g. the Baltic Sea Action Plan (BSAP), extending to 2021 (HELCOM, 2013). Nevertheless, this scenario represents a state of the world in which agricultural commodity prices are high, and farmers' efforts for higher yields are permitted by agrienvironmental and other policies 2.3.4. Moderate adaptation (MoA) Yields increase linearly by 10% between 2013 and 2050 (0.25% per year). Fertilization also increases by 10%, to provide the necessary nutrients for plant growth. Prices and policies are the same as in the baseline. However, the moderate market prospects and yield increase imply somewhat higher needs of liming and drainage investments (although to a lower extent than in the successful adaptation scenario) on specialized cereal farms which also utilize new cultivars. The policy environment does not encourage a large group of part-time unspecialized farmers in such investments. 2.3.5. Little adaptation (LiA) Adaptation is rare, even if some individual farms can avoid cereal yield reductions with the help of some adaptations based on fungicide use and liming. Since adaptations at the farm and crop level are rare, yields of both cereals and grasses decrease by 10%, which is the opposite change in yields compared to the moderate adaptation-scenario. Fertilization, however, does not decrease since no change in real prices is expected. Agricultural policy does not encourage productivity improvements in crop production. In order to test the DREMFIA model results, we also specified two “worse-case” scenarios, in which prices were 10% higher compared to the baseline but adaptation was either low (yields − 10%) or nonexistent (yields − 20%): LiA_HP (Little adaptation, high prices; yields − 10%, prices + 10%) and NoA_HP (No adaptation, high prices; yields −20%, prices +10%). Decreasing yields may result if policies discourage maintaining yields e.g. through sufficient fertilization and land quality improvements, or climate turns out to be more challenging and current cultivars, vulnerable to climate change, are used. This reasoning is consistent with the farm modeling outcomes of Lehtonen et al. (2015) and some crop modeling results of current barley cultivars, especially on drought-prone sandy soils (Rötter et al., 2013).


Agricultural commodity prices remain at the baseline level in the little and moderate adaptation scenarios, but increase significantly in the successful adaptation scenario already by 2030. Nelson et al. (2013) compared the outcomes of 9 global trade models and found prices of agricultural commodities varying between −5% and +30%, depending on the model. Thus, the prices in the different scenarios are in a plausible range. However, decreasing real prices, although less likely than stable or increasing real prices, are not considered in this study.

2.4. Economic agricultural sector model DREMFIA Changes in regional level agriculture throughout Finland under these scenarios are evaluated using an economic agricultural sector model DREMFIA. The model simulates production and foreign trade of agricultural commodities, as well as land use (areas under crops and set aside) and production intensity (fertilization, manure use) annually from 1995 to 2020 and produces a steady state static equilibrium for 2030, 2040 and 2050. The model assumes rational economic behavior and competitive markets, replicates realized production and land use in 1995–2012, and produces realistic development paths of agriculture (see Lehtonen, 2001, 2013 for details). Four main areas are included in the model: southern Finland, central Finland, Ostrobothnia (the western part of Finland), and northern Finland (Fig. 2). Demand and foreign trade is determined at the level of the main regions. The products move between the main regions, to cover the demand of each region, at a certain transportation cost. Since the model is very exact in terms of agricultural policy, the main regions are further divided into sub-regions according to regional disaggregation specific to agricultural support payments. Production is thus determined at the level of each of 17 sub-regions. Fig. 2 shows 14 regions: subregions A, B, C1, C2 in Southern Finland; subregions B, C1, C2, C2 north in Middle Finland; subregions C1, C2, C3 in Ostrobothnia; and subregions C2 north, C3, C4 in Northern Finland. Three small regions of Yläneenjoki (with 34,500 ha in Southern Finland), Taipaleenjoki (with 12,000 ha in Middle Finland) and Simojoki (6400 ha in Northern Finland)

Fig. 2. Spatial aggregation of the DREMFIA sector.

171

with 6000–12,000 ha are not shown. More details of the model are available in Lehtonen (2001). Demand in each main region is covered by the production in its own sub-regions, by transportations from other regions, and by imports from abroad. The model is solved to reach the most economic outcome. The underlying hypothesis is that producers engage in profit-maximizing behavior and consumers engage in utility-maximizing behavior in competitive markets. Decreasing marginal utility of consumers and increasing marginal cost lead to production and foreign trade levels where the marginal cost of supply is equal to the market prices. Each region specializes in products that provide the greatest profitability, taking into account the profitability of production in other regions, import prices and consumer demand. The use of different production resources, including farmland, is optimized, taking into account differences in resource quality, technology, and costs of production and transportation. Prices of nitrogen fertilizer and crops affect fertilization levels. If crop prices increase, farmers increase fertilization as long as the value of increased crop yield is higher than the value of increased fertilization. In the model the annual yield of each crop is calculated for the 17 production regions by determining the optimum fertilization from the relation between nitrogen and the crop product. Notably, if manure from livestock is available, less commercial fertilizers are used. The yield and price scenarios are given to the model in the form of (1) future crop yields — the model adjusts fertilization according to the needs of plants to reach the specific yield levels; different fertilization response functions are used for different crops (Mitscherlich function for barley, wheat, oats, mixed cereals and peas, and quadratic function for rye, potatoes, sugar beet, hay, silage, green fodder and oilseeds; cf. Lehtonen, 2001; Ylätalo, 1996); (2) future prices, which affect the use of inputs, crop yields and change in the level of production and foreign trade. Changed yields and prices imply changes in the regional allocation of both livestock and crop production, to reach the most competitive production structure and supply (also changing imports and exports) with respect to the food demand, which is assumed to remain unchanged from the 2010–2012 level. The baseline, an important point of comparison for the specified scenarios presented above, is rooted in the model through a multi-phase validation process. Known statistical data from official agricultural statistics for 1995–2012 is extensively used, as well as some selected empirical data from research databases for the parameter values in different production functions determining e.g. feed use and yield levels of animals, fertilizer use and crop yields. Available data on milk quota prices, land rental values and value and quantity of inputs are used in the model validation. Some few model specific behavioral parameters, which have no correspondence in the literature or known data sources, have been adjusted so that the model replicates very closely the historically observed development of production, land use and prices. Such calibration parameters are (1) substitution elasticities in the demand functions, coupled with the price elasticities of demand, as well as (2) the farm type specific savings rates and (3) the propensity to invest in larger farm size at dairy farms (see Lehtonen (2001, 2013) for details). Endogenous structural and technical change in the dairy sector can be validated, using a unique combination of parameters, to follow the development of the farm structure statistics. Equilibrium properties of the model (increasing marginal costs in terms of production quantity, decreasing marginal utility of consumers) stabilize the overall production very close to the observed production quantities at the whole country level and at the 4 main regions' level. Region-specific budget limits imposed on production support of milk and bovine animals also contribute to the production allocation. The main trends in observed production allocation are replicated, e.g. decreasing dairy production in southern Finland and slightly increasing dairy production in central parts of the country. However there are small differences between the observed and simulated production levels at the national level as follows (simulated/observed, average 2008–2012): Cereal area − 1.1%, grassland area −7.8%, milk yield per

172


cow + 0.2%, milk production + 2.6%, beef production + 4.0%, pork −0.3%, and poultry meat production −5.1%. Grassland area is smaller in DREMFIA than in reality, since horses, lambs and reindeers (users of roughage) are not included. Further examples of comparisons between the model results and reality are available in Lehtonen (2013). We used the following outputs of the DREMFIA model, separately from each of the 4 main regions, as an input to the VEMALA model in the scenario runs: annual fertilizer use (both mineral and manure), crop distribution (land area in hectares under each crop) and predicted yield increase for each crop. 2.5. Nutrient loading model VEMALA VEMALA is an operational, national scale, nutrient (phosphorus and nitrogen) loading model for Finnish catchments. It simulates runoff processes, nutrient processes, leaching and transport on land, in rivers and in lakes (Huttunen et al., submitted for publication). The VEMALA model provides an estimate of external loading, outflow loading, retention of nutrients in all ca. 58,000 lakes in Finland, as well as nutrient loading source apportionment into main sources — agriculture, forests, scattered settlements and point sources. The present version of the VEMALA model consists of a catchment scale, semi-process-based model of TN loading, VEMALA-N, a field scale process-based model for TP loading, VEMALA-ICECREAM and river and lake sub-models (Huttunen et al., submitted for publication). The agricultural loading is simulated via a process-based model for TP and a semi-processbased model for TN. The non-agricultural loading from land areas is simulated via a concentration-runoff-based model for TP and a semi-process-based model for TN. The loading from scattered settlements, point loading and atmospheric deposition are included as input data to the VEMALA model. 2.5.1. VEMALA-N — total nitrogen loading model Total nitrogen loading is simulated by the VEMALA-N model (Huttunen et al., submitted for publication), which is a catchment scale N leaching model with six land uses defined: spring cereals, winter cereals, grassland, root crops, green fallow and forest. In the VEMALA-N model, nitrate (NO− 3 ) and organic nitrogen are described separately. Organic nitrogen is modeled using a concentration-discharge relationship in which subsurface and base flow are characterized by different organic nitrogen concentrations. Nitrate is simulated using a semi-process-based model similar to the INCA approach (Wade et al., 2002; Rankinen et al., 2004). Ammonium (NH+ 4 ) leaching is neglected because, on average, ammonium loading represents only a small fraction (around 6%) of total nitrogen loading in Finnish catchments (Mattsson et al., 2005). By contrast, NH + 4 storage in the soil is modeled and linked to the soil organic nitrogen and nitrate. The processes included to simulate nitrate leaching are mineralization, nitrification, denitrification, immobilization, plant uptake, fertilizer input, dissolution and nitrogen leaching. Most of the nitrogen processes in the soil are simulated as first order kinetic processes that depend on the mass of the nitrogen fractions in the soil, soil temperature and soil moisture: Process ¼ Storage krate ktemp ksoilm

soil moisture in the soil. Soil temperature is not included because both mineral and organic soils have substantial N2O production at temperatures close to zero (Martikainen et al., 2002). The growth of plant biomass is related to air temperature sums over the growing season (Rekolainen and Posch, 1993). The mass balance of + − NO− 3 and NH4 in the soil is simulated for each time step. NO3 leaching is simulated depending on daily NO− 3 concentration in the soil and daily simulated runoff. 2.5.2. VEMALA-ICECREAM — total phosphorus loading model The total phosphorus load from agricultural areas is simulated with the ICECREAM model, which is a field-scale, process-based model calculating erosion and phosphorus losses on a daily time scale. It is based on the earlier CREAMS (Knisel, 1980) and GLEAMS (Knisel, 1993) models and has been adapted to Finnish climatic conditions by Rekolainen and Posch (1993). Further developments and testing of the model have been reported in Tattari et al. (2001), Yli-Halla et al. (2005), Paasonen-Kivekäs et al. (2006), Bärlund et al. (2009) and Jaakkola et al. (2012). Modifications made to ICECREAM for the VEMALA application have been reported by Huttunen et al. (submitted for publication). In the ICECREAM model, water flow in the soil is divided into three pathways: first surface runoff and macropore flow (in clay soils) are subtracted from precipitation/snow melt and the rest of the water infiltrates through the soil profile as matrix flow. Surface runoff is calculated by the SCS curve number method (Soil Conservation Service, 1972), with different parameters for frozen and frost-free soil conditions. Macropore flow is calculated for clay soils according to Jaakkola et al. (2012), with some modifications (Huttunen et al., submitted for publication). Matrix flow is simulated with the simple bucket model based on four parameters: soil hydraulic conductivity, wilting point, field capacity and soil porosity. Evapotranspiration is calculated with the Penman–Monteith method (Monteith and Unsworth, 1995) and corrected against values from the VEMALA model. Soil phosphorus is described with three inorganic and two organic P pools. Initialization of the pools and phosphorus flow between the pools is described in detail by Tattari et al. (2001) and the modifications made to the equations by Huttunen et al. (submitted for publication). Phosphorus can be added to the soil through mineral fertilization, manure application or through plant residues. Phosphorus can be lost from the soil through plant uptake or it can be transported to waterways in dissolved form or attached to eroded soil particles. Dissolved phosphorus (DP) and particulate phosphorus (PP) can be transported via surface runoff and macropore flow, and DP also via matrix flow. Erosion is simulated with the modified USLE model (Foster et al., 1977). For calculation of interrill detachment, rainfall erosivity is calculated according to Lombardi (1979), who linked daily precipitation to rainfall erosivity with a simple power-law function. Posch and Rekolainen (1993) derived monthly values for the parameters of the function from Finnish rainfall data from May to October. In climate change scenarios these monthly values could not be used, as the intensity and quantity of rain are predicted to change in the future climate. The average values for October were used in this study, as most harvesting has been completed by October and the soil is then most vulnerable to erosion.

ð1Þ

where Process is mineralization, nitrification, immobilization, kg · ha − 1 day − 1, Storage is storage of nitrogen fraction in the soil (kg ha−1), krate is rate of the nitrogen transformation process (d−1), ktemp is a temperature effect coefficient and ksoilm is a soil moisture effect coefficient. The mineralization as a function of soil temperature is described with an s-shaped logistic function. According to this function, mineralization is low at low soil temperatures around 0 °C and increases rapidly between 5 and 15 °C, finally reaching a plateau at ca. 20 °C (cf. DessureaultRompre et al., 2010). Denitrification depends on the NO− 3 availability and

2.5.3. River and lake nutrient mass balance models There are numerous lakes in Finnish catchments, and therefore there is a need to have a nutrient balance model included in VEMALA. Nutrient mass balance in Finnish lakes is simulated by estimating external loading, outflow, sedimentation, internal loading (for phosphorus) and denitrification (for nitrogen). External loading, outflow and sedimentation greatly depend on the water balance model for each lake. The water balance model simulates inflow discharge, outflow discharge, lake volume and also residence time. Lake-specific sedimentation rate defines the fraction of the nutrient mass sedimenting during one day. For phosphorus the sedimentation rate is derived with the method


a

100 80

60 40 20 Y1995 Y1998 Y2001 Y2004 Y2007 Y2010 Y2013 Y2016 Y2019 Y2022 Y2025 Y2028 Y2031 Y2034 Y2037 Y2040 Y2043 Y2046 Y2049

0

kg/ha 5000

b

4600

base

4200

MoA

3800

LiA

3400

SuA_VHP

3000 Y1995 Y1998 Y2001 Y2004 Y2007 Y2010 Y2013 Y2016 Y2019 Y2022 Y2025 Y2028 Y2031 Y2034 Y2037 Y2040 Y2043 Y2046 Y2049

2.5.4. Evaluation of VEMALA model performance The observed stream concentrations, loads and concentrations in the lakes for the period 1991–2013 were used to calibrate the model VEMALA-N and models for river and lake nutrient mass balance and transport. If the terrestrial loading from agricultural areas in VEMALAICECREAM was not in line with the observations, it was calibrated by multiplying the load with a coefficient specific to each catchment, or 2nd level catchment in some very large catchments. The coefficient obtained values between 0.5 and 2.05. The total number of observation points used in the calibration for the whole of Finland is about 31,000 in rivers and 35,000 in lakes. The model performance is evaluated on different time scales — (1) daily simulated concentrations and loads are compared with observed concentrations using the Nash–Sutcliffe criteria (NSE) or visual inspection of the graphs and (2) simulated annual nutrient loading is compared with estimated annual nutrient loading using discharge and observed concentrations by an averaging method (HELCOM, 2011). The NSE for daily loading for the 27 biggest rivers in Finland varies in range 0.38–0.90 (mean value 0.75) for TN loading. For TP loading NSE varies in range − 0.88–0.95 (mean value 0.71). Mean NSE above 0.70 represents a good correlation between simulated and observed loadings. The lowest NSE values for daily loading are in Vuoksi catchment where loading pattern is highly affected by large lake Saimaa regulation and retention. A more detailed description of the VEMALA model performance is given in Huttunen et al., submitted for publication.

kg N / ha 120

kg N/ha 50

c

40 30 20

10 0 Y1995 Y1998 Y2001 Y2004 Y2007 Y2010 Y2013 Y2016 Y2019 Y2022 Y2025 Y2028 Y2031 Y2034 Y2037 Y2040 Y2043 Y2046 Y2049

described in Ahlgren et al. (1988) and is fine-tuned by calibration against phosphorus observations. For nitrogen the sedimentation rate is calibrated by observations. In the river model, rivers are divided into discrete stretches. The continuity (mass balance) equation of water and nutrients for each river stretch has been solved by defining a simple calibrated water level/outflow relationship for river stretches (Huttunen et al., submitted for publication). Nutrient transport, sedimentation, erosion and denitrification in the rivers are simulated by the river model.

173

3.1. Agricultural production and land use development in Finland

Fig. 3. a) Nitrogen fertilization for barley (kg N/ha) (top). b) Yield level for barley (kg/ha) (middle). c) Nitrogen balance for barley (kg N/ha) (bottom) in different scenarios, southern Finland, sub-region B. Base = Baseline; MoA = Moderate adaptation (yields +10%); LiA = Little adaptation (yields −10%); SuA_VHP = Successful adaptation, very high prices (yields +30%, crop prices +30%).

Adaptation to climate change influences the amount of fertilization and the resultant yield levels and nutrient balances in the different study areas. Here we used barley, the most cultivated cereal crop, and southern Finland (especially sub-region B, the most important area of barley cultivation), as an example of the fertilization and crop yield changes under the following agricultural scenarios: little adaptation with high prices (LiA), moderate adaptation (MoA) and successful adaptation with very high prices (SuA_VHP) (Fig. 3). Annual level average prices of barley fluctuated in an unusual manner between 94 and 187 €/ton 2007–2009, resulting in wide fluctuations in the barley yield (Fig. 3b), the nitrogen fertilization (Fig. 3a) and nitrogen balance (Fig. 3c). High prices, assumed to have developed already by 2030, result in rapidly increasing N fertilization per ha of barley in the scenario SuA_VHP. However, the 30% increase in the yields actualizes only gradually, by 2050, but high prices already in 2030 trigger higher fertilization. Hence the nitrogen balance of barley first increases up to 2030, then decreases, but remains above the baseline level until 2050 (Fig. 3c). In the scenario of moderate adaptation (MoA), higher yield levels (10%) result in slightly higher nitrogen fertilization and nitrogen balance levels of barley, compared to the baseline, until 2050. In the scenario of little adaptation (LiA) the nitrogen fertilization is very close to the level of the baseline scenario, but the nitrogen balance is higher than in the MoA-scenario. More land area is allocated for cereals in the scenario of successful adaptation with very high prices (SuA_VHP) (Fig. 4a). Together with a yield increase of 30% this means that overall cereal production almost

doubles from the average level of 3.8 million tons in 2000–2013 (Fig. 4b). In the moderate adaptation scenario (MoA), crop yields increase by 10% but the area under cereals remains close to 1.2 million ha (the average in 2002–2013 was slightly below 1.2 million ha). The overall production of cereals decreases in the LiA-scenario with decreasing yields. Low yields also lead to slightly decreasing cereal area. Dairy and beef production is stabilized by national and EU CAP (Common Agricultural Policy) coupled payments which are linked to production, and there are budget limits for these production-linked subsidies. However, dairy milk production increases by 12%, but grassland area decreases because of 30% higher grass yields in the successful adaptation scenario. Pig and poultry meat production are only slightly affected by the yield and price scenarios. The share of cereals out of the total cultivated area increases. The gradually increasing crop yield level brings down the average nitrogen balance from the increased level of 2030 (triggered by high prices) (Fig. 4c) in the SuA_VHP-scenario, until 2050. The production volume of grassland fodder is rather stable in the different scenarios. This implies that grassland areas decrease if grass yields increase, and grassland areas increase if grass yields decrease. Decreasing crop yields in the little adaptation scenario imply lower profits for livestock production. Consequently, 5–10% less beef and milk is produced, and 5–10% less grass forage is needed, if lower yields actualize (−10% compared to the baseline). Probable outcomes of moderate or successful adaptation are larger cereal areas and reduced areas under grasslands. This change is more

3. Results

174


1000 ha 1800

a

1600 1400 1200 1000

Y1995 Y1998 Y2001 Y2004 Y2007 Y2010 Y2013 Y2016 Y2019 Y2022 Y2025 Y2028 Y2031 Y2034 Y2037 Y2040 Y2043 Y2046 Y2049

800

1000 tons 8000

b

7000 6000

base

5000

MoA

4000

LiA SuA_VHP

3000

in Lehtonen, 2015). Hence the results suggest that yields should be increased in the case of increasing agricultural commodity prices — otherwise prices trigger higher fertilization and nutrient balances. Decreasing yields should be avoided, since higher prices, possibly more frequent in the future if the mean and variance of agricultural commodity prices increase, will then lead to increasing nutrient balances. In our very optimistic successful adaptation scenario mainly cereal production expanded, but harvested grass output changed only slightly. Dairy production increased 12%, beef 2%, pork 12%, and poultry production 3%, but overall farm income increased even by almost 100% in the successful adaptation scenario. This is mainly because of decreased production costs per unit produced, not only because of increased production. Farm income increased by 14% in the moderate adaptation scenario, and decreased by 12% in the little adaptation scenario, until 2050. Significantly increasing cereal production in moderate and successful adaptation scenarios is explained by abundant farmland resources and relatively little regulated cereal markets, whereas high feed costs, high opportunity cost of labor and budgetary constraints in national subsidies stabilize livestock production close to the current levels.

Y1995 Y1998 Y2001 Y2004 Y2007 Y2010 Y2013 Y2016 Y2019 Y2022 Y2025 Y2028 Y2031 Y2034 Y2037 Y2040 Y2043 Y2046 Y2049

2000

kg N/ha 70 65

c

60 55 50 45 Y1995 Y1998 Y2001 Y2004 Y2007 Y2010 Y2013 Y2016 Y2019 Y2022 Y2025 Y2028 Y2031 Y2034 Y2037 Y2040 Y2043 Y2046 Y2049

40

Fig. 4. a) Cereals area (1000 ha) (top). b) Cereals production (tons) (middle). c) Nitrogen balance (kg N/ha) (bottom) over all the cultivated area in Finland. Base = Baseline; MoA = Moderate adaptation (yields +10%); LiA = Little adaptation (yields −10%); SuA_VHP = Successful adaptation, very high prices (yields +30%, crop prices +30%); light blue line in Figure a): Cereals area according to official agricultural statistics 1995–2013.

pronounced in areas where the share of grasslands is currently high, i.e. in central and northern Finland. In the context of this study it is important to note that this land use change also affects average nutrient balances. The average nitrogen balance over all cultivated areas (excluding set aside land) is slightly higher in all scenarios than in the baseline (Fig. 4c). This is partly because the increasing nitrogen use efficiency, i.e. the share of nitrogen fertilizer utilized by plants, is not assumed, despite higher yielding cultivars and other improvements in farm management. More frequent adverse weather events are likely to affect soil nitrogen over the whole year, as well as nitrogen uptake during the growing period. Hence it is reasonable not to allow negative or very low nitrogen balance even temporarily in the DREMFIA model, in order to safeguard the availability of plant nutrients. However, it is remarkable that the average nitrogen balance in the little adaptation scenario is always higher than in the baseline scenario 2030–2050. Decreasing production and increasing nitrogen balance, both undesirable outcomes for farmers and society, are clear consequences of little and inadequate adaptation. Our DREMFIA model results from the sensitivity scenarios LiA_HP (Little adaptation, high prices; yields − 10%, prices + 10%) and NoA_HP (no adaptation, high prices; yields −20%, prices +10%) show higher nutrient balances than in the case of LiA scenario. This suggests that if prices increase, nutrient balances also increase, especially if the cereals yields decrease due to little success in adaptation (more details

3.2. Changes in hydrology and seasonal shift of the nutrient loading Precipitation (620–680 mm yr− 1 ) and simulated hydrological variables — runoff (280–350 mm yr − 1 ) and evapotranspiration (245–365 mm yr− 1) vary in different areas in Finland in the reference decade (2001–2010). Detailed results of simulated hydrological variables for reference decade and the decade 2051–2060 for various catchments under the Mean A1B, wet and dry climate change scenarios are shown in the Table S1, electronic supplementary material. The VEMALA model predicts that runoff will increase by 3–11% in the Mean A1B scenario until the year 2060 in comparison to 2001–2010 in all parts of Finland due to the increase in precipitation. However, there is more variation in the predicted runoff changes in the dry and wet scenarios. In the dry scenario, runoff is decreasing by 2–13% due to a decrease or only a slight increase in precipitation and an increase in evapotranspiration. In the wet scenario, runoff is increasing considerably in all parts of Finland by 9–25%. Especially in southern and central Finland a shift from spring flood peak to several winter runoff peaks will take place due to the temperature increase in winter. The lowest increase in runoff is simulated for the Bothnian Bay catchment (BB) for two reasons: the reference decade 2001–2010 had more runoff in BB than in other parts of Finland, and evapotranspiration increased more in BB, probably because of more pronounced increase in temperature. These areal differences in the increase in runoff also influence the areal change in nutrient transport. Accordingly, the differences in runoff changes in the climate change scenarios are determining the shift in nutrient loading. There is a clear seasonal shift in runoff and subsequently in nutrient loading (Fig. 5). Most of the nutrient (TN and TP) loading increase is predicted to occur during winter and autumn months (from November till March). The seasonal shift is more pronounced in southern parts of Finland due to the greater shortening of the snow cover period. The magnitude of the seasonal shift decreases towards northern parts of the country. For example, the increase in winter and autumn TN loading in AS is 37%, whereas in the BB catchment the increase is 29%. The seasonal shift in nutrient loading occurs for two main reasons: increase in runoff in the winter and autumn months outside the growing season for TN and TP, and increased organic nitrogen mineralization and nitrification due to higher soil temperatures leading to an increase in nitrate leaching. During April and May, TN loading will decrease mainly due to the decrease in runoff. There will be less soil moisture storage in soil at the beginning of the growing season due to the lower amount of snowmelt and elevated evaporation. In June and July TN loading will decrease because of a higher nutrient uptake in plants and lower runoff caused by higher evapotranspiration. TP loading has a similar seasonal shift caused by a shift in runoff, and therefore it is not shown in Fig. 5.


175

Fig. 5. Monthly changes in (a) mean discharge and (b) total nitrogen (TN) loading to the Baltic Sea from Finnish catchments in the periods 2001–2010 and 2051–2060 in the Mean A1B, wet RCA3-H-A1B and dry HIRH-A-A1B scenarios. Results from the VEMALA model.

3.3. Nutrient loading terminology and structure of nutrient loading sections Gross loading represents total loading which is formed in the catchment area and flows into river and lake systems. Gross loading consists of diffuse loading (agricultural loading TNAG, TPAG and loading from non-agricultural land areas TNNA, TPNA,), point loading, atmospheric deposition to lakes, rivers and loading from scattered settlements. TN and TN agricultural gross loading is described in Sections 3.5, 3.6 and loading from non-agricultural land areas is explained in Section 3.4. For the purpose of the reduction in nutrient loading it is important to estimate the source apportionment of the nutrient loading entering the inland waters. It is also important to show the climate change effect on the different sources of the diffuse loading. The retention of nutrients in the inland waters is the difference between gross loading to the inland waters and the final riverine nutrient loading to the Baltic Sea. VEMALA also simulates nutrient retention. The specific values are given in Section 3.7. For some Finnish catchments, retention can be very high due to the large surface of lakes covering the catchments. Therefore, it is important to take into account retention processes when calculating nutrient balances. For example, simulated TP retention in the VUO catchment is 77%. The simulated riverine nutrient loading includes nutrient loading with river discharge from river catchments, unmonitored coastal land areas and also direct point loading to the Baltic Sea sub-basins. The riverine loadings are used in the implementation of the Marine Strategy Framework Directive and are therefore presented in Section 3.7.

3.4. Total nitrogen (TNNA) and phosphorus (TPNA) loading from non-agricultural land areas and other non-agricultural sources To estimate the total nutrient gross loading from Finnish catchments into the Baltic Sea, we estimated the amount of loading from other sources than agriculture. These include loading from non-agricultural land areas, point sources (such as municipal and industrial sewage treatment plants), scattered settlements and atmospheric deposition. The non-agricultural loading from land areas is simulated by VEMALA using a concentration-runoff based sub-model for TP and a semiprocess based model for TN. The other loading sources are included as input data to VEMALA. The simulated total nitrogen loading from non-agricultural land areas (TNNA) is 57,000 t yr−1, which is 42% of TN loading. Most of the TN loading from forests is leached as organic nitrogen, whereas only a small fraction of TN is leached as nitrate. Nitrate is mostly leached during the spring flood peak. Nitrate produced during the nitrification process of ammonium in forest soils is efficiently taken up by vegetation.

The organic nitrogen model is based on the concentration-subsurface flow and concentration-groundwater flow relationships. Therefore, organic nitrogen loading increases with increasing runoff in climate change scenarios. The TNNA loading will increase by 2–11% in the Mean A1B scenario until the year 2060 in comparison to 2001–2010 in all parts of Finland (Table S2, electronic supplementary data) due to the increase in runoff. In the dry scenario, TNNA loading is changing (−8 to −2%) due to a decrease in runoff. In the wet scenario, TNNA loading is increasing considerably in all parts of Finland by 9–29%. The simulated total phosphorus loading from the catchments draining to the Baltic Sea from non-agricultural land areas (TPNA) is 1920 t yr−1, accounting for 33% of the total phosphorus loading. TPNA is simulated based on a concentration-runoff relationship and in most scenarios TPNA loading increases and decreases according to the runoff. The only exception is the Mean A1B scenario in the BB catchment, in which the TPNA decreases slightly despite the small (2%) increase in runoff. This is probably caused by the decrease of the high discharge peaks in spring. The simulated TPNA loading for each catchment is shown in Table S2 in the electronic supplementary data. The other non-agricultural sources of nutrients are scattered settlements, point sources and atmospheric deposition. TN loading from point sources and atmospheric deposition to lake and river surfaces is 24,000 t yr−1, which represents 20% of the total TN loading. In climate change runs it is assumed that TN loadings from these sources do not change in the future. By contrast, TN loading from scattered settlements is assumed to decrease by 30% due to the implementation of the Finnish legislation requirements defined in the Government Decree on Treating Domestic Wastewater in Areas Outside Sewer Networks. The TN loading from scattered settlements is currently 2700 t yr−1, which is around 2% of the total nitrogen loading, and will decrease to 1800 t yr−1 (1.4% of TN loading). TP loading from scattered dwellings is predicted to decrease by 49% in the scenarios. Presently, loading from scattered dwellings represents 7% of the TP loading to Finnish catchments. In the future scenarios, this share decreases to 3%. The loading from point sources and deposition to lake and river surfaces is on average 770 t yr−1, representing 13% of the total load. As with nitrogen, the TP loading from these sources is not expected to change much and is set to the same level as in 2010, i.e. 690 t yr−1. 3.5. Total nitrogen loading from agricultural areas (TNAG) The general trend in agricultural loading in different climate change scenarios is that Mean A1B gives an average loading increase, the wet scenario gives the highest predicted TN loading due to the highest increase in runoff and the dry scenario gives the lowest predicted TN

176


loading due to the decrease in runoff. We consider as the main result the predicted TN loading from Mean A1B, but the uncertainty of climate change predictions is taken into account by showing the dry and wet scenario results as a loading variation range.

Baltic Sea sub-basin catchments, except in the BB loading is changing −10–4%.

3.6. Total phosphorus loading from agricultural areas (TPAG) 3.5.1. TNAG loading in the climate change scenarios (Mean A1B, dry and wet), no changes in agriculture Simulated total nitrogen loading from agricultural areas (TNAG) to water bodies in Finnish catchments is 47,500 t yr− 1, which is 35% of total TN loading (Table 1). The agricultural loading varies between different areas in Finland from 61% in the Archipelago Sea (AS) catchment to only 21% in the Vuoksi (VUO) catchment, due to the climatic conditions and suitable soil fertility for agriculture. In the climate change scenario (Mean A1B), TNAG is increasing by 5–11% in the Gulf of Finland (GF), AS and the Bothnian Sea (BS) catchments due to the increasing runoff. In the Bothnian Bay (BB) and VUO catchment, the main crop classes are grasslands and pastures (47% of agricultural areas), which will have slightly higher plant uptake due to the longer growing season. The denitrification from soils will also increase. Therefore the nitrogen balance will slightly decrease and, in combination with almost no change in runoff, TNAG is changing (−11–3%) in the BB and VUO catchments in the Mean A1B scenario. In the dry scenario, TNAG is decreasing in almost all catchments except AS. In the wet scenario, TNAG is increasing in almost all catchments, except in the BB catchment. 3.5.2. TNAG loading into the inland waters in the combined climate and agricultural adaptation scenarios Annual TNAG loading increases in two of the combined climate and agricultural change scenarios. The seasonal shift of TN loading from spring to autumn and winter months occurs in all agricultural scenarios to the same extent. The increase of annual TNAG loading to the Baltic Sea is 18–26% in the little adaptation scenario and −3%–16% in the successful adaptation scenario. The loading slightly changes in the moderate adaptation scenario − 14–1% (Table 1). The TN loading depends on the TN balance in the soil and the changes in crop distribution. The little adaptation scenario is the most pessimistic in terms of nutrient leaching, because TN balance in the soil is increasing due to the lower plant uptake and slightly higher fertilizer use. In the successful adaptation scenario, TN fertilizer use is compensated by higher plant uptake, and therefore TN leaching is relatively lower than in the little adaptation scenario. In the successful adaptation scenario the higher TN loading increase is also caused by a crop change from grasslands to spring cereals. Simulated specific leaching from spring cereal crops (20 kg ha−1 yr−1) is higher than from grasslands (12–13 kg ha−1 yr−1). The spatial distribution of TNAG change in the successful and moderate adaptation scenarios is the same as in the Mean A1B scenario. In MoA and LiA the highest increase is in the AS catchment 0–36%, but the highest decreasing change is in the BB catchment −26–0%, following the runoff changes in the scenarios. In the little adaptation scenario the TNAG increase is more evenly spatially distributed; TNAG is increasing by 8–38% in all

A major part of the TPAG load is transported from fields to waterways with eroded sediments from top soil, where P tends to accumulate in agricultural soils. According to the simulation results of the Mean A1B and wet climate scenarios, erosion will increase in clayey areas, especially in the catchments draining to the Archipelago Sea (data not shown). This is mainly due to the increasing runoff and wetness of soils in autumn, which causes instability in the structure of heavy clay soils, thus enhancing soil erosivity (Aura et al., 2006). On the other hand, erosion of coarse soils with high infiltration decreased in the simulation. As fields in Finland are relatively flat (slope b 3% in 83% of the field area, Puustinen et al., 2010), the major cause of erosion in coarse soils is snow melt in spring. Infiltration in frozen soil is low and thus a lot of surface runoff is produced (Turtola and Kemppainen, 1998). In the climate change simulations, the amount of snow decreased and a larger proportion of the annual precipitation/snow melt entered the soil while it was not frozen, allowing more infiltration. Less surface runoff was thus produced, causing less erosion from the P rich top soil. According to the simulation results, less erosion will take place in the catchments draining to the Bothnian Bay and in the dry scenario also in the Vuoksi catchment and in the catchments draining to the Gulf of Finland. Phosphorus losses from agriculture increased in almost all Mean A1B and wet scenarios (Table 2). In the Bothnian Bay catchment the P loading increased in all climate scenarios despite the decrease in mineral soil losses. Abundance of peat soils was the main reason for this. Higher percolation leached more phosphorus from peat soils, where the P binding capacity is lower than in mineral soils. Although the total runoff decreased in the dry scenario, more percolation was produced as surface runoff was reduced. The decrease in runoff in the dry scenario also resulted in a small decrease in P loading in the Archipelago Sea catchment. Although the wet winters kept erosion at its current level, the soluble P load was decreased with the runoff. Differences in P load between the agricultural scenarios reflect the different P balances attained in the simulations. In the last decade of the simulations, the mean P balances for the whole of Finland were +0.7 to +1.1 kg ha−1 in the baseline, +1.7 to +2.0 kg ha−1 in the little adaptation, −1.1 to −0.5 kg ha−1 in the moderate adaptation and −1.3 to −0.5 kg ha−1 in the successful adaptation scenario, but they differed regionally. The smallest harvest occurred in the dry scenario and the greatest in the wet scenario, so the P balances varied accordingly, the smallest P balance being in the wet scenario. The changes in crop distribution also affected the P loading to a small extent, depending on the proportions of different types of crops and also on the distribution of crops in different types of field plots. For example, if spring cereals were switched to grass on two field plots of the same soil type, erosion

Table 1 Change of TN loading from agricultural areas (TNAG) in the different catchments in 2001–2010 and in the climate change scenarios (Mean A1B, dry and wet) combined with three agricultural adaptation scenarios. Catchments draining into the different Baltic Sea sub-basins

Vuoksi (VUO) Gulf of Finland (GF) Archipelago Sea (AS) Bothnian Sea (BS) Bothnian Bay (BB) Total

Present TNAG (2001–2010)

Share of TNAG from total TN load

Change in TNAG load in the climate, agricultural change scenario (2051–2060)

103 t yr−1

%

A1B, no agricultural changes, %

Successful adaptation, %

M A1B

Wet

Dry

M A1B

Wet

Dry

M A1B

Wet

Dry

M A1B

Wet

Dry

−3 5 11 6 −11 0

2 14 15 14 −5 7

−8 −5 1 −3 −15 −8

−3 14 23 8 −13 2

7 27 36 21 0 16

−8 5 17 2 −14 −3

−8 5 9 0 −24 −5

−3 14 13 10 −17 1

−13 −7 0 −7 −26 −14

17 29 34 27 0 18

22 38 36 36 4 26

8 14 23 19 −10 18

6 10.5 5.3 11.8 13.5 47.5

21 37 61 49 28 35

Moderate adaptation, %

Little adaptation, %


177

Table 2 Change of TP loading from agriculture (TPAG) in the catchments in 2001–2010 and in the climate change scenarios (Mean A1B, dry and wet) combined with three agricultural adaptation scenarios. Catchments draining into the different Baltic Sea sub-basins

Vuoksi (VUO) Gulf of Finland (GF) Archipelago Sea (AS) Bothnian Sea (BS) Bothnian Bay (BB) Total

Present TPAG (2001–2010)

Share of TPAG from total TP load

Change in TPAG load in the climate, agricultural change scenario (2051–2060)

t yr−1

%

Baseline (%)

410 650 390 590 620 2670

41 52 74 60 31 46

Successful adaptation (%)

Moderate adaptation (%)

Little adaptation (%)

M A1B

Dry

Wet

M A1B

Dry

Wet

M A1B

Dry

Wet

M A1B

Dry

Wet

+4 +23 +29 +26 +21 +21

−8 −6 −3 +4 +11 0

+17 +39 +47 +46 +34 +37

0 +20 +26 +20 +18 +17

−11 −9 −2 −1 +8 −3

+12 +34 +42 +39 +30 +32

+2 +22 +27 +22 +20 +19

−9 −7 −1 +1 +10 −1

+14 +37 +44 +42 +33 +35

+5 +22 +31 +27 +23 +22

−6 −7 −2 +6 +13 +2

+9 +38 +50 +49 +37 +38

was prevented more effectively on the steeper plot, but the decrease in TP loading also depended on the P status of the soil. 3.7. Nitrogen and phosphorus riverine loading to the sea 3.7.1. Total nitrogen (TN) riverine loading in the climate and agricultural adaptation scenarios The previous results showed the nutrient gross loading to the inland waters. The VEMALA model also takes into account the nutrient sedimentation and retention in the inland waters by using water and nutrient mass balance models. Around 30% of TN loading is retained in lakes and rivers and represents about 42,500 t yr− 1. The TN retention in Ladoga is assumed to be 30% (Pitkänen and Tallberg, 2007), which is subtracted from the riverine loading reaching the Baltic Sea from the Vuoksi catchment. The future TN retention in the lakes is almost the same in the climate change scenarios as in the present, despite the increasing water temperatures and NO3 loadings in climate change conditions, due to the changes in residence time. There is clear decrease in spring concentrations in all catchments. An increase in autumn concentrations in AS and BS catchments is probably due to more mineralization during warmer autumns. There is a slight increase in winter concentrations in the BB catchment, probably because of higher mineralization. In the GF catchment there are no changes in concentrations in other seasons than spring, probably because of higher denitrification from the high number of lakes. The concentrations at the outlet of the VUO catchment are decreasing throughout the year, because of the denitrification and runoff regulation effect of the large lake Saimaa. The simulated present TN riverine loading into the Baltic Sea from Finnish catchments for the period 2001–2010 is 90,100 t yr−1. The riverine loading also includes direct point loading to the sea. TN loading to the Baltic Sea increases slightly (2%) in the Mean A1B scenario during the period 2051–2060, with a range between −6% in the dry and +13% in the wet climate change scenario due to changes in runoff. In the Mean A1B scenario, TN loading increases to all sub-basins except for the BB (Fig. 6). TN loading decreases to the BB sub-basin, because of a low increase in runoff and due to a decrease in TNAG loading from agricultural areas (cf. Table 1). In the Mean A1B scenario the highest increase (11%) in TN loading takes place to the AS sub-basin and VUO catchment. To the AS sub-basin the TN loading increase is high, because of a high agricultural loading share and 11% increase in agricultural TNAG loading. The AS catchment also has the lowest retention of TN (9%) of all catchments studied due to the low number of lakes. By contrast, the VUO catchment has high retention (65%) and a low share of agricultural loading, and therefore the increase in runoff (11%) determines the increase in TN loading more than other factors. In the Mean A1B scenario, TN riverine loading increases the most in the little adaptation agricultural scenario: a total increase of 26, 14 and 13% takes place to AS, GF and BS sub-basins, respectively, compared to the present loading. These increases are at least double compared to the Mean A1B scenario alone to some of the sub-basins. In the successful

and moderate adaptation scenarios the TN riverine loadings were closer to those in the Mean A1B scenario. To the GF, AS and BS the same pattern was repeated: moderate adaptation led to a lower TN loading compared to the Mean A1B scenario, whereas successful adaptation increased the loading. BB differed from the rest of the sub-basins in that in all scenarios the TN riverine loading remained the same as at present (little adaptation scenario), or decreased slightly. In the wet scenario there was the lowest TN loading increase to the BB sub-basin. Fig. 6 shows total TN riverine loadings to the Baltic Sea sub-basins VUO to GF, GF, AS, BS and BB, which are respectively 6.7, 17.5, 8.1, 16.8 and 41.0 × 103 t yr−1. Compared to the HELCOM data (HELCOM, 2011), which is calculated from observed loadings, the simulated TN loadings are higher. The difference is 13% in GF, 31% in AS, 2% in BS and 16% in BB. The HELCOM data does not include direct point loading to the sea, and therefore the point loading was subtracted from the simulated values before the comparison.

3.7.2. Total phosphorus (TP) riverine loading in the climate and agricultural adaptation scenarios The TP riverine loading into the Baltic Sea depends on the changes in loading, as well as on changes in sedimentation. The simulated net sedimentation of P was 2000 t yr−1 in the rivers and lakes of Finland. The TP sedimentation in Ladoga was assumed to be 70% (Pitkänen and Tallberg, 2007), and that amount was subtracted from the riverine loading reaching the Baltic Sea from the Vuoksi catchment. Phosphorus sedimentation in the Baltic Sea sub-basins changed by + 1–+ 5% in the Mean A1B scenario, + 10%–+ 14% in the wet scenario and − 9% to −6% in the dry scenario due to the changes in the amount of loading. The seasonal shift of loading also shifted the sedimentation from spring to winter. In all climate scenarios the TP concentrations in the rivers increased in winter months compared to present concentrations. This was due to the increased intensity of precipitation and enhanced erosivity of wetter soils. By contrast, the high concentrations during the spring snow melt were substantially decreased. In the present conditions, the highest concentrations were simulated during the spring floods, but in the climate change scenarios the highest concentrations occurred in November, although the highest discharge was still simulated in the spring. The simulated present TP riverine loading into the Baltic Sea from Finnish catchments for the period 2001–2010 was 3590 t year− 1 (Fig. 7). TP loading to the Baltic Sea increased in the Mean A1B and wet scenarios, but decreased in the dry scenario during the period 2051–2060. In contrast to the TN riverine loading, the TP riverine loading was not higher in the successful adaptation scenario than in the moderate adaptation. To all sub-basins the TP loading decreased with the level of adaptation. To BB the simulated TP riverine load was less affected by the climate change scenarios than in the other catchments, due to the small proportion of agriculture in that area.

178


Fig. 6. Total nitrogen (TN) riverine loading from Finland to the Baltic Sea in 2001–2010 and in the climate change scenarios (Mean A1B, wet RCA3-H-A1B and dry HIRH-A-A1B) combined with three agricultural adaptation scenarios. Atmospheric deposition to the sea is not included.

The simulated TP loads were somewhat higher compared to the HELCOM data (HELCOM, 2011). The difference was 8%, 12%, 3% and 3% to the GF, AS, BS and BB sub-basins, respectively. 4. Discussion 4.1. Nutrient loading change comparison with other studies Climate change influences hydrology and biogeochemistry in Scandinavian waters. Generally the increases in precipitation and runoff, and decreases in snow cover and soil frost, have been predicted to increase nutrient runoff into the catchments and eventually into the Baltic Sea (e.g. Arheimer et al., 2012; Meier et al., 2012). Few studies (Abler et al., 2002 and Hattermann et al., 2007) have however taken into account the changes in plant dynamics and human behavior that accompany the climatic changes. Our aim was to analyze how the combined effects of climate change and agricultural adaptation influence nutrient fluxes into the Baltic Sea. We prepared consistent scenarios in which farmers respond to market and climatic conditions by changing land use and fertilization. By combining agro-economic modeling with hydrological and nutrient load modeling, we showed how agricultural adaptation, through climate-induced changes in soil processes, nutrient leaching and nutrient retention in lakes and rivers, affects nitrogen and phosphorus transport into the sea sub-basins surrounding Finland. In our study, we predicted an increase in TN and TP loading to the sea with the Mean A1B climate change scenario. Agricultural change was shown to either worsen or alleviate this increase, depending on the nutrient in question and the level of adaptation. The simulated riverine TP and TN loadings to the Baltic Sea from Finnish catchments were 3590 t yr−1 and 90,100 t yr−1, respectively. The increase in TN loading in the combined agricultural and Mean A1B climate scenario varied from 0 to 8% and in TP loading from 7 to 9%. Dry and wet

scenarios set a wide range of variation of nutrient loading changes mainly due to the changes in runoff. The range for nutrient loading change in all climate and agricultural scenarios was − 9% to 18% for TN and − 7% to + 21% for TP. We predicted the smallest change in nutrient loading to the BB sub-basin, due to the smallest change in runoff, low percentage of agriculture in the catchment, a decrease in the nitrogen balance in the soil and an increase in denitrification in the soil. The major N removal processes in lakes are regulated by a number of factors, of which hydraulic residence time, availability of NO− 3 , temperature, dissolved organic carbon (quality and quantity) and redox potential are among the most important (Ahlgren et al., 1994; Myrstener, 2015). According to our results, there will be a slight decrease in residence times of lakes due to the increase in runoff, as was also reported by Baron et al. (2013), which will partly compensate the increase in denitrification caused by temperature increase. Denitrification is the most sensitive to increased NO− 3 loading to the lakes, and therefore the highest simulated denitrification from lakes is in Mean A1B in combination with the little adaptation scenario, when there is the highest TN loading increase and moderate runoff increase. However, there is a need for more process-based models to assess the interactions between the complex hydrological and biogeochemical processes affecting lake denitrification. Table S3 (Electronic supplementary data) shows the comparison of nutrient loading change to the Baltic Sea in different climate change studies. The results from the different models are not fully comparable because of the different climate scenarios, agricultural scenarios and comparison periods used. In addition, VEMALA includes only the loading from Finnish rivers, whereas the other two models calculate the loading from all rivers flowing to the Baltic Sea. VEMALA and BaltHYPE (Arheimer et al., 2012) are process-based models, whereas Eriksson Hägg et al. (2014) used the population-discharge proxy for future nutrient load simulation.


179

Fig. 7. Total phosphorus (TP) riverine loading from Finland to the Baltic Sea in 2001–2010 and in the climate change scenarios (Mean A1B, wet RCA3-H-A1B and dry HIRH-A-A1B) and three agricultural adaptation scenarios. Atmospheric deposition to the sea is not included.

The results from the two process-based models were rather similar, but closer analysis reveals that the reasoning behind the results is different. Arheimer et al. (2012) predicted the highest increase in runoff (14%–16%) for the BB and BS catchments, whereas in our study the change in runoff in the BB catchment was the smallest (− 5%–9%). Note that the reference periods are different in these two studies — the reference period for this study was 2001–2010 and for Arheimer et al. (2012) it was 1971–2000 (Table S3, electronic supplementary data). Despite the increasing runoff, the TN loading decreased in the BB catchment in their study, which they attributed to an increase in denitrification and retention time in lakes and rivers. In mean climate change scenarios neither model presented a dramatic increase in TN loading from Finnish catchments. Furthermore, Arheimer et al. (2012) predicted increase in TP loading to the Baltic Sea. The increase was especially high (0%–46%) to the BS sub-basin. They suggested that this was caused by increasing mineralization of P. In our simulations the abundance of clay and peat soils together with increased runoff all contributed to the high increase in TP loading to BS. The results of Eriksson Hägg et al. (2014) had the highest mean increase of TN loading to the GF, BS and BB sub-basins. For TP loading their results show less variation in TP loading increase between different sub-basins and scenarios. They predicted the highest increase in TP loading to the GF sub-basin, while the two process-based models predicted the highest increase to the BS sub-basin. The advantage of the process-based models resides in their ability to represent the effect of weather and climate on water and nutrient cycles. However, there are difficulties involved in the application of process-based models to large scale catchments — (1) the need for very detailed input data at a small scale resolution, (2) too few

observation data for validation of the model results and subprocesses modeled, (3) the continuous development of the models, since the less stable assumptions are identified and improved during model applications, (4) the uncertainty in the climate model results used as input, and (5) the challenges and uncertainties in the prediction of agro-economic changes. Nevertheless, there is added value in such interdisciplinary studies to show plausible future development patterns and to study their effects on nutrient loading to the Baltic Sea. However, the processes causing the changes in nutrient loading in the simulations need to be studied carefully to allow comparison between the results from different models. Eriksson Hägg et al. (2014) used the population-discharge proxy for future nutrient load simulation. The advantage of using a simple method is the ease of comparing the model to more complex population models, dynamic population projections and nutrient source allocation/transport models. However, there are several limitations in such a simple approach — (1) extrapolation of the parameters beyond the calibration period and limitations to the assumed connections between consumption and nutrient loads, (2) the in-stream retention changes in climate scenarios cannot be assessed in this type of model, (3) the effect of changing temperature and soil moisture on the processes in the soils cannot be assessed by this approach, and (4) the agricultural development subject to the economic and climatic forces is not explicitly considered. 4.2. How to avoid and mitigate increased nutrient loading The Baltic Sea Action Plan (BSAP; HELCOM, 2013) states that nutrient loads to the Baltic Sea should not be allowed to increase. Our results

180


show that both decreasing and increasing crop yields may increase nutrient leaching in the context of climate change. However, in our study the effects of climate change on phosphorus loading could be alleviated to some extent by efficient agricultural adaptation, although even successful adaptation was not sufficient to keep the TP loading at the current level. In the case of nitrogen loading, the adaptation scenarios did not decrease the loading compared to the present agricultural practices, except in the BB, BS and VUO in the moderate adaptation scenario. Successful adaptation increased loading more than moderate adaptation, because of a substantial increase in the cereal production and a relatively small increase in the livestock production. However, in the successful adaptation scenario there was little increase in nutrient loading compared to the baseline scenario, despite the increasing production. Decreasing yields, in turn, imply higher nutrient balances and increased leaching despite decreased production. Nevertheless, adaptation to climate change and realization of crop yields are important for the economic viability of northern agriculture, despite the substantial farm subsidies paid. The role of agriculture in the Baltic Sea countries requires more attention if decreasing nutrient leaching is the goal, e.g. how much domestically produced food is needed, what kind of agriculture and land use is preferred, and where. Adaptation at different levels needs to be planned and knowledge-based, taking into account various factors that influence nutrient transport from different farming systems and various biogeochemical processes in the soil and in the catchments. According to our results, increase in nutrient loading can be mitigated through higher yields obtained by successful adaptation to climate change. One of the best incentives for farmers to seek higher yields is increased prices on the global markets. Possible farm level adaptation measures include (Peltonen-Sainio et al., 2009) (1) choosing new cultivars, which can utilize the longer growing season in the future and give higher potential yields, (2) investing in soil improvements, e.g. by proper drainage systems and liming of the soils, and (3) crop protection. According to Goulding et al. (2000) and our results, higher fertilizer use will lead to higher leaching of nitrogen in dry years when crop yields are low. Splitting of nitrogen fertilization during the growing period is already practiced in grass production, but split fertilization has also been proposed for cereal crops (Hyytiäinen et al., 2011) in order to avoid nutrient surplus in dry years In order to mitigate the increase in phosphorus loading from fields, our results highlight the importance of two main measures: controlling soil erosion and decreasing P storages in soils (Schoumans et al., 2014). Since P is stored in soils, high P storages have developed in overfertilized fields in the long run (Delgado and Torrent, 1997). Decreasing these storages requires a negative P balance in the fields. Although only a small decrease in P pools due to negative P balance has been found in less than a decade (van der Salm et al., 2009), the model simulation made in this study showed significant effects of the slightly negative P balance on TP loading until 2060. Thus, reaching high yields with minimum necessary fertilizer input is an important measure in reducing phosphorus loading in a long term perspective. To control erosion, in turn, the steepest fields should be changed from annual crops to perennial vegetation cover — preferably to forest or pasture without fertilizer use. This would be possible in the successful adaptation scenario, as higher yields would allow some of the agricultural area to be left for water protection measures. In order to avoid the negative impact of price and yield fluctuations on the environment, a new set of policy instruments should be developed and the existing ones should be adapted to climate change conditions. Existing policy instruments which focus on the amount of fertilizers used per hectare are not encouraging farmers to achieve higher yields or to use new cultivars in changing climate. Instead, the subsidy system should target limiting the nutrient balance in the soil. Such subsidies might allow a larger fertilizer application for more efficient crop growth, and would result in smaller nutrient balances, decreased cultivated area needed for production and less nutrients being released to catchments. However, this

kind of subsidy system would be much harder to monitor than the present system. The aim of a further research effort will be to design a realistic set of mitigation measures and policy instruments to support the use of these measures to reduce agricultural loading in order to meet the target loadings to the Baltic Sea sub-basins set by BSAP. In this study we have created the framework for the models to simulate the complex interactions of the disciplines affecting nutrient loading — climate, hydrology, chemistry, agriculture and agro-economy. This modeling framework allows testing the different combinations of mitigation measures together with policy instruments to reach the target loading under a changing climate and various agricultural development scenarios. 5. Conclusions Nutrient loading to the Baltic Sea sub-basins from Finnish catchments depends on future hydrological changes, and may vary widely depending on the catchment characteristics such as the amount of agricultural area, soil texture and lake percentage as well as farm management practices in the future climate. The nutrient loading change was affected most by the predicted increase in runoff — the wet scenario simulated the highest increase in loading, while the dry scenario simulated a decrease or a low increase in loading. The highest loading increase was predicted for the Archipelago Sea: 0%–30% for TN and −9%–34% for TP. In the little adaptation scenario the increase in nutrient loading was the greatest. This implies that successful agricultural adaptation is essential in order to mitigate the effects of climate change on nutrient loading. New, high yield cultivars, as well as other yieldpromoting management options such as liming, better drainage and crop protection, may help in alleviating the negative effects of climate change on nutrient loading by bringing the nutrient balances down at the farm level. Higher yields may also free up land for mitigation of leaching. Further studies focusing on productivity and land use changes that take into account rational economic behavior under different market and policy scenarios are necessary for determining the most cost-effective measures to diminish agricultural nutrient loading to the Finnish catchments and the Baltic Sea as required by BSAP. Acknowledgments This work was funded by the Academy of Finland by the project MARISPLAN — Marine spatial Planning in a changing climate (IH, VP: Decision number 140871; HL: Decision number 140840; MV: Decision number 140833) and the respective institutes of the authors. We gratefully acknowledge the support of the funding agencies. The authors also thank anonymous reviewers, whose comments and suggestions helped to improve the final version of the manuscript. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.scitotenv.2015.05.055. References Abler, D., Shortle, J., Carmichael, J., Horan, R., 2002. Climate Change, agriculture, and water quality in the Chesapeake Bay Region. Clim. Change 55 (3), 339–359. Ahlgren, I., Frisk, T., Kamp-Nielsen, L., 1988. Empirical and theoretical models of phosphorus loading, retention and concentration vs. lake trophic state. Hydrobiologia 170 (1), 285–303. Ahlgren, I., Sörensen, F., Waara, T., Vrede, K., 1994. Nitrogen budgets in relation to microbial transformations in lakes. Ambio 23, 363–366. Arheimer, B., Dahne, J., Chantal, D., 2012. Climate change impact on riverine nutrient load and land-based remedial measures of the Baltic Sea action plan. AMBIO 41, 600–612. http://dx.doi.org/10.1007/s13280-012-0323-0. Aura, E., Saarela, K., Räty, M., 2006. Savimaiden eroosio. MTT:n selvityksiä 118 (32 pp. in Finnish).

I. Huttunen et al. / Science of the Total Environment 529 (2015) 168–181 Bärlund, I., Tattari, S., Puustinen, M., Koskiaho, J., Yli-Halla, M., Posch, M., 2009. Soil parameter variability affecting simulated fieldscale water balance, erosion and phosphorus losses. Agric. Food Sci. 18, 402–416. Baron, J.S., Hall, E.K., Nolan, B.T., Finlay, J.C., Bernhardt, E.S., Harrison, J.A., Chan, F., Boyer, E.W., 2013. The interactive effects of excess reactive nitrogen and climate change on aquatic ecosystems and water resources of the United States. Biogeochemistry 114, 71–92. http://dx.doi.org/10.1007/s10533-012-9788-y. Delgado, A., Torrent, J., 1997. Phosphate-rich soils in the European Union: estimating total plant-available phosphorus. Eur. J. Agron. 6, 205–214. Dessureault-Rompre, J., Zebarth, B.J., Georgallas, A., Burton, D.L., Grant, C.A., Drury, C.F., 2010. Temperature dependence of soil nitrogen mineralization rate: comparison of mathematical models, reference temperatures and origin of the soils. Geoderma 157, 97–108. Eriksson Hägg, H., Lyon, S.W., Wällstedt, T., Mörth, C.-M., Claremar, B., Humborg, C., 2014. Future nutrient load scenarios for the Baltic Sea due to climate and lifestyle changes. AMBIO 43, 337–351. http://dx.doi.org/10.1007/s13280-013-0416-4. Foster, G.R., Meyer, L.D., Onstad, C.A., 1977. A runoff erosivity factor and variable slope length exponents for soil loss estimates. Trans. ASAE 20, 683–687. Goulding, K.W.T., Poulton, P.R., Webster, C.P., Howe, M.T., 2000. Nitrate leaching from the BroadbalkWheat Experiment, Rothamsted, UK, as influenced by fertilizer and manure inputs and the weather. Soil Use Manag. 16, 244–250. Hattermann, F.F., Gömann, H., Conradt, T., Kaltofen, M., Kreins, P., Wechsung, F., 2007. Impacts of global change on water-related sectors and society in a trans-boundary central European river basin — part 1: project framework and impacts on agriculture. Adv. Geosci. 11, 85–92 (www.adv-geosci.net/11/85/2007/). HELCOM, 2011. The Fifth Baltic Sea Pollution Load Compilation (PLC-5). Baltic Sea Environ. Proc. 128. HELCOM, 2013. Overview of implementation of the HELCOM Baltic Sea Action Plan (BSAP). Document Prepared for the 2013 HELCOM Ministerial Meeting (40 pp. http://helcom.fi/Documents/Ministerial2013/Associated%20documents/Supporting/ BSAP_Overview_with%20cover.pdf (Accessed January 22 2015)). Höglind, M., Thorsen, S.M., Semenov, M.A., 2013. Assessing uncertainties in impact of climate change on grass production in Northern Europe using ensembles of global climate models. Agric. For. Meteorol. 170 (15), 103–113 (March 2013). Huttunen, I., Huttunen, M., Piirainen, V., Korppoo, M., Lepistö, A., Räike, A., Tattari, S., Vehviläinen, B., 2015. A national scale nutrient loading model for Finnish watersheds — VEMALA. Environmental Modeling and Assessment (submitted for publication). Hyytiäinen, K., Niemi, J.K., Koikkalainen, K., Palosuo, T., Salo, T., 2011. Adaptive optimization of crop production and nitrogen leaching abatement under yield uncertainty. Agric. Syst. 104, 634–644. IPCC, 2000. In: Nakicenovic, N., Swart, R. (Eds.), IPCC Special Report, Emission Scenarios. Cambridge University Press, UK (570 pp.). IPCC, 2007. Climate Change 2007. The Physical Science Basis. Cambridge University Press, Cambridge (996 pp.). IPCC, 2013. Climate change 2013: the physical science basis. In: Stocker, T.F., Qin, D., Plattner, G.-K., Tignor, M., Allen, S.K., Boschung, J., Nauels, A., Xia, Y., Bex, V., Midgley, P.M. (Eds.), Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA (1535 pp.). Jaakkola, E., Tattari, S., Ekholm, P., Pietola, L., Posch, M., Bärlund, I., 2012. Simulated effects of gypsum amendment on phosphorus losses from agricultural soils. Agric. Food Sci. 21, 292–306. Knisel, W. (Ed.), 1980. CREAMS, a field-scale model for chemicals, runoff, and erosion from agricultural management systems. Conservation Research Report 26. USDA, Washington, D.C. (643 pp.). Knisel, W., 1993. GLEAMS: groundwater loading effects of agricultural management systems. Version 2.10. Publication No. 5University of Georgia, Department of Biological and Agricultural Engineering, Coastal Plain Experiment Station, Athens, Georgia. Lehtonen, H., 2001. Principles, structure and application of dynamic regional sector model of Finnish agriculture. Academic dissertation. Publications 98Systems Analysis Laboratory, Helsinki University of Technology (Agrifood Research Finland, Economic Research (MTTL). Helsinki. 265 p. http://lib.tkk.fi/Diss/2001/isbn9512256894/). Lehtonen, H., 2013. Sector-level economic modeling as a tool in evaluating greenhouse gas mitigation options. Acta Agric. Scand. Sect. A Anim. Sci. 62 (4), 326–335. http:// dx.doi.org/10.1080/09064702.2013.797011. Lehtonen, H., 2015. Evaluating Adaptation and Production Development of Finnish Agriculture in Climate and Global Change (Manuscript). Lehtonen, H., Liu, X., Purola, T., 2015. Balancing climate change mitigation and adaptation with socio-economic goals at farms in Northern Europe. In: Paloviita, A., Järvelä, M. (Eds.), Chapter 11 in book “Climate Change Adaptation and Food Supply Chain Management”. Routledge. Lombardi, F., 1979. Universal Soil Loss Equation, Runoff Erosivity Factor, Slope Length Exponent, and Slope Steepness Exponent for Individual Storms. (PhD Thesis). Purdue University, W. Lafayette, Indiana. Martikainen, P.J., Regina, K., Syväsalo, E., Laurila, T., Lohila, A., Aurela, M., Silvola, J., Kettunen, R., Saarnio, S., Koponen, H., Jaakkola, T., Pärnä, A., Silvennoinen, H., Lehtonen, H., Peltola, J., Sinkkonen, M., Esala, M., 2002. Agricultural soils as a sink and source of greenhouse gases: a research consortium (AGROGAS). Finnish Global Change Research Programme FIGARE 951-29-2407-2pp. 55–67. Mattsson, T., Kortelainen, P., Räike, A., 2005. Export of DOM from boreal catchments: impacts of land use cover and climate. Biogeochemistry 76 (2), 373–394. Meier, H.E.M., Hordoir, R., Andersson, H.C., Dieterich, C., Eilola, K., Gustafsson, B.G., Höglund, A., Schimanke, S., 2012. Modeling the combined impact of changing climate and changing nutrient loads on the Baltic Sea environment in an ensemble of transient simulations for 1961–2099. Clim. Dyn. 39, 2421–2441. http://dx.doi.org/10. 1007/s00382-012-1339-7.

181

Monteith, J.L., Unsworth, M., 1995. Principles of Environmental Physics. 2nd ed. Arnold, London (291 pp.). Myrstener, M., 2015. Disentangling Denitrification and Its Environmental Drivers in Northern Boreal Lakes. (Master thesis). Umeå University, Sweden, p. 21. Nelson, G.A., Valin, H., Sand, R.D., Havlík, P., Ahammadd, H., Derynge, D., Elliott, J., Fujimori, S., Hasegawah, T., Heyhoed, E., Kylei, P., von Lampe, M., Lotze-Campen, H., Mason d'Croza, D., van Meijl, H., van der Mensbruggh, D., Müller, C., Popp, A., Robertson, S., Schmid, E., Schmitz, C., Tabeau, A., Willenbockel, D., 2013. Climate change effects on agriculture: economic responses to biophysical shocks. PNAS 111 (9) (March 4, 2014 www.pnas.org/cgi/doi/10.1073/pnas.1222465110). Paasonen-Kivekäs, M., Koivusalo, H., Bärlund, I., Tattari, S., Alakukku, L., 2006. Modelling runoff and erosion in agricultural soil: application of ICECREAM model to a field site in southern Finland. In: Tamm, Toomas, Pietola, Liisa (Eds.), NJF Seminar 373 : Transport and Retention of Pollutants From Different Production Systems, Tartu, Estonia, 11–14 June 2006. NJF Report 2, pp. 82–87. Peltonen-Sainio, P., Jauhiainen, L., Hakala, K., Ojanen, H., 2009. Climate change and prolongation of growing season: changes in regional potential for field crop production in Finland. Agric. Food Sci. 18, 171–190. Pitkänen, H., Tallberg, P. (Eds.), 2007. FE15/2007 searching efficient protection strategies for the eutrophied Gulf of Finland: the integrated use of experimental and modelling tools (SEGUE). The Finnish Environment 15/2007. Environmental Protection (90 pp. URN:ISBN:978-952-2671-0). Posch, M., Rekolainen, S., 1993. Erosivity factor in the universal soil loss equation estimated from Finnish rainfall data. Agric. Sci. Finl. 2, 271–279. Purola, T., 2013. Taudinkestävien ja tautialttiiden ohralajikkeiden taloudellinen vertailu. (Master thesis), University of Helsinki, Department of economics and, Management (https://helda.helsinki.fi/handle/10138/40225). Puustinen, M., Turtola, E., Kukkonen, M., Koskiaho, J., Linjama, J., Niinioja, R., Tattari, S., 2010. VIHMA — a tool for allocation of measures to control erosion and nutrient loading from Finnish agricultural catchments. Agric. Ecosyst. Environ. 138, 306–317. Rankinen, K., Kaste, Ø., Butterfield, D., 2004. Adaptation of the Integrated Nitrogen Model for Catchments (INCA) to seasonally snow-covered catchments. Hydrol. Earth Syst. Sci. 8 (4), 695–705. Rankinen, K., Peltonen-Sainio, P., Granlund, K., Ojanen, H., Laapas, M., Hakala, K., Sippel, K., Helenius, J., Forsius, M., 2013. Climate change adaptation in arable land use, and impact on nitrogen load at catchment scale in northern agriculture. Agric. Food Sci. 22, 342–355. Rekolainen, S., Posch, M., 1993. Adapting the CREAMS model for Finnish conditions. Nord. Hydrol. 24 (5), 309–322. Rötter, R.P., Höhn, J.G., Fronzek, S., 2012. Projections of climate change impacts on crop production: a global and a nordic perspective. Acta Agric. Scand. Sect. A Anim. Sci. 62 (4), 166–180. http://dx.doi.org/10.1080/09064702.2013.793735. Rötter, R.P., Höhn, J., Trnka, M., Fronzek, S., Carter, T.R., Kahiluoto, H., 2013. Modelling shifts in agroclimate and crop cultivar response under climate change. Ecol. Evol. 3 (12), 4197–4214. http://dx.doi.org/10.1002/ece3.782. Ruosteenoja, K., Tuomenvirta, H., Jylhä, K., 2007. GCM-based regional temperature and precipitation change estimates for Europe under four SRES scenarios applying a super-ensemble pattern scaling method. Clim. Chang. 81, 193–208. Ruosteenoja, K., Räisänen, J., Pirinen, P., 2010. Projected changes in thermal seasons and the growing season in Finland. Int. J. Climatol. 31, 1473–1487. http://dx.doi.org/10. 1002/joc.2171. Salo, T., Lemola, R., 2014. Typpi- ja fosforitaseet. Luku 3.3. raportissa: Jyrki Aakkula & Johanna Leppänen (toim.) 2014. Maatalouden ympäristötuen seurantatutkimus (Mytvas3). Loppuraportti. Maa- ja metsätalousministeriö 3/2014 (ISBN 978-952453-851-0 (Painettu), ISBN 978-952-453-852-7 (Verkkojulkaisu), ISSN 1238–2531 (Painettu), ISSN 1797-397X (Verkkojulkaisu). 267 pp.). Salo, T., Lemola, R., Esala, M., 2007. National and regional net nitrogen balances in Finland in 1990–2005. Agric. Food Sci. 16 (4), 366–375. Schoumans, O.F., Chardon, W.J., Bechmann, M.E., Gascuel-Odoux, C., Hofman, G., Kronvang, B., Rubæk, G.H., Ulén, B., Dorioz, J.-M., 2014. Mitigation options to reduce phosphorus losses from the agricultural sector and improve surface water quality: a review. Sci. Total Environ. 468–469, 1255–1266. Soil Conservation Service, 1972. Hydrology. Section 4. Soil Conservation Service National Handbook. U.S. Department of Agriculture, Washington, DC. Tattari, S., Bärlund, I., Rekolainen, S., Posch, M., Siimes, K., Tuhkanen, H.-R., Yli-Halla, M., 2001. Modeling sediment yield and phosphorus transport in Finnish clayey soils. Trans. ASAE 44 (2), 297–307. Turtola, E., Kemppainen, E., 1998. Nitrogen and phosphorus losses in surface runoff and drainage water after application of slurry and mineral fertilizer to perennial grass ley. Agric. Food Sci. Finland 7, 569–581. van der Linden, P., Mitchell, J.F.B. (Eds.), 2009. ENSEMBLES: Climate Change and Its Impacts: Summary of Research and Results from the ENSEMBLES Project. Met Office Hadley Centre, FitzRoy Road, Exeter EX1 3 PB, UK (160 pp.). van der Salm, C., Chardon, W.J., Koopmans, G.F., van Middelkoop, J.C., Ehlert, P.A.I., 2009. Phytoextraction of phosphorus-enriched grassland soils. J. Environ. Qual. 38, 751–761. Wade, A.J., Durand, P., Beaujouan, V., Wessel, W.W., Raat, K.J., Whitehead, P.G., Butterfield, D., Rankinen, K., Lepisto, A., 2002. A nitrogen model for European catchments: INCA, new model structure and equations. Hydrological Earth System Science 6, 559–582. Ylätalo, M. (Ed.), 1996. Maatalousyritysten sopeutuminen EU:ssa vallitseviin hintasuhteisiin. Julkaisuja 12. Helsingin yliopisto, Taloustieteen laitos, Maatalousekonomia, Helsinki (258 pp.). Yli-Halla, M., Tattari, S., Bärlund, I., Tuhkanen, H.-R., Posch, M., Siimes, K., Rekolainen, S., 2005. Simulating processes of soil phosphorus in geologically young acidic soils of Finland. Trans. ASAE 48 (1), 101–108.

Variability in climate change simulations affects needed long-term riverine nutrient reductions for the Baltic Sea.

Projected future climate change and Baltic Sea ecosystem management.

Modelling the influence of climate change on the chemical concentrations in the Baltic Sea region with the POPCYCLING-Baltic model.

Nutrient loadings from urban catchments under climate change scenarios: case studies in Stockholm, Sweden.

Online participation in climate change adaptation: A case study of agricultural adaptation measures in Northern Italy.

Baltic Sea ecosystem-based management under climate change: Synthesis and future challenges.

Spatial properties of nutrient concentrations in the Baltic Sea.

The impact of high-end climate change on agricultural welfare.

Uncertainty assessments and hydrological implications of climate change in two adjacent agricultural catchments of a rapidly urbanizing watershed.

Changing climate and nutrient transfers: Evidence from high temporal resolution concentration-flow dynamics in headwater catchments.

Managing Climate Change Refugia for Climate Adaptation.

Understanding nutrient biogeochemistry in agricultural catchments: the challenge of appropriate monitoring frequencies.

Baltic Sea nutrient reductions--what should we aim for?

Climate-change adaptation: Designer reefs.

Operational resilience of reservoirs to climate change, agricultural demand, and tourism: A case study from Sardinia.

Contrasting climate change impact on river flows from high-altitude catchments in the Himalayan and Andes Mountains.

Adapting to the effects of climate change on Inuit health.

Climate change mitigation and adaptation in the land use sector: from complementarity to synergy.

A sea change in exoplanet climate models?

Climate change: Pakistan must invest in adaptation.

Greenhouse gas emissions from agricultural food production to supply Indian diets: Implications for climate change mitigation.

Handling preference heterogeneity for river services' adaptation to climate change.

Climate change adaptation: putting principles into practice.

Nutrient content in macrophyta collected from southern Baltic Sea beaches in relation to eutrophication and biogas production.