Direct and indirect effects of climate change on herbicide leaching — A regional scale assessment in Sweden

Science of the Total Environment 514 (2015) 239–249

Contents lists available at ScienceDirect

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

Direct and indirect effects of climate change on herbicide leaching — A regional scale assessment in Sweden Karin Steffens a,⁎, Nicholas Jarvis a, Elisabet Lewan a, Bodil Lindström b, Jenny Kreuger b, Erik Kjellström c, Julien Moeys a a b c

Department of Soil and Environment, Swedish University of Agricultural Sciences, Box 7014, 75007 Uppsala, Sweden Department of Aquatic Sciences and Assessment, Swedish University of Agricultural Sciences, Box 7050, 75007 Uppsala, Sweden Rossby Centre, Swedish Meteorological and Hydrological Institute, 60176 Norrköping, Sweden

H I G H L I G H T S • • • • •

67 crop-herbicide combinations were simulated with a regionalized version of MACRO. MACRO-SE successfully distinguished leachable and non-leachable herbicides. Direct effects of climate change led to small reductions in leachate concentration. Indirect effects doubled the area at risk of groundwater contamination. Indirect effects of climate change should be investigated alongside the direct.

a r t i c l e

i n f o

Article history: Received 22 October 2014 Received in revised form 17 December 2014 Accepted 17 December 2014 Available online 7 February 2015 Editor: D. Barcelo Keywords: Regional scale Pesticide modelling Climate change Direct effects Indirect effects MACRO

a b s t r a c t Climate change is not only likely to improve conditions for crop production in Sweden, but also to increase weed pressure and the need for herbicides. This study aimed at assessing and contrasting the direct and indirect effects of climate change on herbicide leaching to groundwater in a major crop production region in south-west Sweden with the help of the regional pesticide fate and transport model MACRO-SE. We simulated 37 out of the 41 herbicides that are currently approved for use in Sweden on eight major crop types for the 24 most common soil types in the region. The results were aggregated accounting for the fractional coverage of the crop and the area sprayed with a particular herbicide. For simulations of the future, we used projections of five different climate models as model driving data and assessed three different future scenarios: (A) only changes in climate, (B) changes in climate and land-use (altered crop distribution), and (C) changes in climate, land-use, and an increase in herbicide use. The model successfully distinguished between leachable and non-leachable compounds (88% correctly classified) in a qualitative comparison against regional-scale monitoring data. Leaching was dominated by only a few herbicides and crops under current climate and agronomic conditions. The model simulations suggest that the direct effects of an increase in temperature, which enhances degradation, and precipitation which promotes leaching, cancel each other at a regional scale, resulting in a slight decrease in leachate concentrations in a future climate. However, the area at risk of groundwater contamination doubled when indirect effects of changes in land-use and herbicide use, were considered. We therefore concluded that it is important to consider the indirect effects of climate change alongside the direct effects and that effective mitigation strategies and strict regulation are required to secure future (drinking) water resources. © 2014 Elsevier B.V. All rights reserved.

1. Introduction

Abbreviations: GCM, global climate model; FST, FOOTPRINT soil type; PAS, pesticide application scenario, a unique combinations of a certain pesticide compound used on a certain crop at a certain time with a certain dose; LOD, limit of detection; GSS, Southern plains of Götaland, in Swedish: Götalands Södra Slättbygder. ⁎ Corresponding author. E-mail address: [email protected] (K. Steffens).

http://dx.doi.org/10.1016/j.scitotenv.2014.12.049 0048-9697/© 2014 Elsevier B.V. All rights reserved.

The conditions for crop growth and productivity are likely to improve in Sweden and other Nordic countries in a changing climate due to an expected northward shift of thermally suitable crop production (e.g. Trnka et al., 2011). Adaptation to changes in climate that influence crop production is highly probable including, for example, changes in the timing of crop cultivation and selection of other crop types or cultivars (Olesen et al., 2011, 2012). Climate not only affects crop growth

240

K. Steffens et al. / Science of the Total Environment 514 (2015) 239–249

and productivity, but also the spatial and temporal distribution and proliferation of weeds, insect pests and pathogens (e.g. Patterson et al., 1999). A northward shift of weeds, pests and diseases might be expected (e.g. Olesen et al., 2011) due to faster development, reproduction and increased survival rates (Patterson et al., 1999), which might lead to dramatic changes in crop health in Sweden (Roos et al., 2011). Thus, there is an increasing concern amongst scientists, regulatory authorities, stakeholders and the general public about water quality and contamination due to increased pesticide use as a consequence of climate change (Bloomfield et al., 2006; Delpla et al., 2009; Solheim et al., 2010; Kattwinkel et al., 2011; Henriksen et al., 2013). Changes in climate will influence pesticide fate directly by changes in climate variables such as temperature and precipitation (Nolan et al., 2008; Lewan et al., 2009). These direct effects of climate variables may sometimes counteract one another. For example, higher temperatures or higher soil moisture contents will increase degradation rates, whereas higher rainfall will generally increase leaching, especially if macropore flow is triggered more often (see e.g. Bloomfield et al., 2006; Beulke et al., 2007). Pesticide fate and behaviour will also be influenced by many indirect effects of climate change including, for example, changes in cropping patterns and crop growth (Olesen et al., 2011; Fogelfors et al., 2009), pesticide application rates (Koleva et al., 2009; Kattwinkel et al., 2011), and soil conditions affecting fate processes such as changes in soil organic carbon content or climate induced freezing/thawing cycles (Stenrød et al., 2008). Although the range of possible influencing factors is quite well understood, the impact of climate change on pesticide fate and transport in the environment has only rarely been assessed quantitatively. Beulke et al. (2007) performed a modelling study assessing the effects of climate change on the transport to groundwater and surface water of several different representative pesticides. Steffens et al. (2013, 2014) demonstrated the effect of model structural, parameter and climate uncertainty on predictions of pesticide losses to tile drains from a heavy clay soil under present and future climate conditions. Ahmadi et al. (2014) modelled changes in atrazine losses to surface water via surface run-off at the watershed scale for a large ensemble of climate model projections. Most of these studies only considered the potential direct impacts of climate change. Bloomfield et al. (2006), Beulke et al. (2007), as well as Steffens et al. (2013, 2014) hypothesized that the indirect effects of climate change might be more significant for future pesticide losses than direct effects, but to our knowledge only a few studies have explicitly attempted to quantify such effects. Kattwinkel et al. (2011) assessed the effects of climate change on the exposure of surface water to agricultural insecticides and concluded that the combined effect is likely to be stronger than the direct or indirect effects of climate change (i.e. changes in land-use and insecticide use) alone. A report issued by the Danish Environmental Protection Agency also assessed the direct and indirect effects of climate change on pesticide leaching at two different sites and for two different agricultural production systems in Denmark for the year 2050 (Henriksen et al., 2013). The indirect effects accounted for changes in crop rotation as well as crop and pest management. They found that the direct and indirect impacts of climate change on leaching risks were small on sandy soils, but more significant for loamy soils prone to macropore flow. Groundwater supplies half of the drinking water in Sweden. These drinking water resources are highly valuable and also very slow and expensive to remediate (Vonberg et al., 2014), so their protection is of paramount importance both today and in the future. Herbicides pose the biggest threat for groundwater contamination by pesticides as they are usually much more mobile than fungicides or insecticides. The aims of this study were therefore to assess and contrast the direct and indirect effects of climate change on herbicide leaching to groundwater in a major crop production region of south-west Sweden. The indirect effects included changes in land-use (crop distribution) and herbicide use in a future climate. We used MACRO-SE, a regionalized version of MACRO 5.2 (Larsbo et al., 2005), to simulate the leaching of herbicide

compounds presently registered for use in Sweden under both present (1970–1999) and future conditions (2070–2099). As a reality check, the simulations for present conditions were compared with monitoring data for herbicides in groundwater. For the future, we defined three herbicide use scenarios that were driven by five different climate scenarios for the end of the century to account for climate uncertainty.

2. Material and methods 2.1. The modelling tools MACRO and MACRO-SE MACRO 5.2 is a one-dimensional physically based model of water flow and solute transport in soil based on a dual-permeability approach. It is used for pesticide registration, both for active ingredients within the European Union (FOCUS, 2000, 2001), and in Sweden for product registration. Richard's equation is used to calculate water flow in the soil matrix and a kinematic wave equation for preferential water flow via macropores. The saturated hydraulic conductivity of the soil matrix governs the partitioning of water flow between matrix and macropore systems. Solute transport in the matrix follows the convection–dispersion equation. A proxy parameter for the unknown geometry of soil macropore structure (Gerke and Van Genuchten, 1996) controls the rate for exchange of water and solutes between the two pore systems via diffusion and convection. A complete water balance is simulated: root water uptake is calculated using the model described by Jarvis (1989), flow and transport to drainage systems is calculated by the Hooghoud equation and seepage potential theory, and the potential evapotranspiration is estimated based on the Penman–Monteith equation (Larsbo and Jarvis, 2003). First-order kinetics for pesticide degradation are calculated with the rate coefficient given as a function of soil temperature and moisture content. Sorption is simulated with a Freundlich sorption isotherm and assumed to be proportional to the organic carbon content of the soil. For a detailed description of the model, the reader is referred to Larsbo et al. (2005) and Jarvis and Larsbo (2012). MACRO-SE is a regionalized version of MACRO 5.2, currently under development by the Centre for Chemical Pesticides (CKB), Swedish University of Agricultural Sciences (SLU). It combines soil maps, detailed information on land-use (arable land), crop area and climate data with a set of empirical pedotransfer functions (Moeys et al., 2012) and other parameter estimation routines to provide a complete parameterization of MACRO 5.2 at regional scales. The soil maps are based on the FOOTPRINT soil type (FST) classification (Centofanti et al., 2008). The original classification, designed to characterize a limit number of soil types to support European wide modelling of pesticide leaching to groundwater and surface waters, was adapted to Swedish conditions. Each FST is defined by a hydrological class, topsoil and subsoil texture and topsoil organic matter content. More details on MACRO-SE are given in the Supplementary material.

2.2. Study region We focused our study on the southern part of a major crop production region in Sweden (GSS, the southern plains of Götaland, see Fig. 1) located in the county of Scania in Southern Sweden. The total land area in Scania is 1100 Mha of which 46% is agricultural land (40% arable land, 6% permanent grassland), 37% forest, and 9% urban land. The proportion of agricultural land is even higher in the GSS region (61%). In a national context, this is very high as only 8% of the total land area of Sweden is agricultural land, whereas 69% is forested (SJV, 2014). Agriculture in Scania is also more intensive than in the rest of Sweden: it contributes almost 50% of the total production on less than 20% of the agricultural area (excluding grasslands; based on SJV, 2014) with 60% of the national pesticide usage (SCB, 2011).

K. Steffens et al. / Science of the Total Environment 514 (2015) 239–249

Fig. 1. Map of Sweden with lakes and surrounding coastal water (light grey). The counties of Scania and Halland are outlined with the simulated region presented in dark grey. The two pesticide monitoring catchments relevant for our study are marked with a black point (Scania) and a black triangle (Halland). Background data (lakes, coastlines, county boarders): ©Lantmäteriet 2014.

2.3. Input data, data support and model parameterization 2.3.1. Climate Present climate (1970–1999). For present climate conditions, we used an observed time series for the period 1970–1999 from the weather station in Barkåkra (5629′ N, 1285′ E). This weather station was chosen by the Swedish Meteorological and Hydrological Institute (SMHI) to be representative for the part of the region GSS we simulated. The climate in the simulated region is humid temperate with an annual precipitation of 706 mm and an annual average temperature of 7.8 °C (see Table 1). Monthly mean winter temperatures are around 0 °C, while they reach up to 16 °C during summer months. Precipitation is relatively evenly distributed over the year with smallest amounts during April and May (Fig. S1). Daily time series of potential evapotranspiration were internally calculated using the Penman–Monteith equation (Larsbo and Jarvis, 2003) from daily average values of temperature, solar radiation, wind speed and vapour pressure deficit. Daily precipitation was disaggregated to hourly values according to Olsson (1998) to obtain hourly driving data for precipitation.

greenhouse gas emission scenario (A1B), see Nakićenović and Swart (2000). The future climate time series were generated by changing the observed present time series with monthly change factors, the so-called delta change method (e.g. Fowler et al., 2007). We followed the approach and reasoning described in Steffens et al. (2014). The change factors were derived by comparing the projected present (1970–1999) and future climate periods (2070–2099) for each member of the ensemble of climate scenarios. In the case of precipitation, we used relative change factors, whereas additive change factors were used for temperature and solar radiation. No changes were made for wind speed and relative humidity. Wind speed was kept unchanged because no systematic patterns are projected towards the end of the century (Kjellström et al., 2011), while according to Bengtsson (2010), both observations and model results agree that relative humidity would hardly be affected in a changing climate. The projections for annual mean temperature and annual precipitation differ strongly amongst the climate models (see Table 1) and range from an increase of 2 to 3.5 °C for temperature and 12 to 25% in annual precipitation. Seasonal variations in these projected changes are presented in Fig. 2 for all five climate scenarios. Towards the end of the 21st-century, all models project a temperature increase all year round and an increase in winter precipitation for this region. Projections of summer precipitation are uncertain and there is, for instance, no consensus amongst the climate models on the direction of change. Our RCA3-ensemble is small, which might lead to biased results as larger ensembles might show larger spreads. We therefore compared our climate scenarios to an ensemble of more than 20 GCMs taken from CMIP3 (Coupled Model Intercomparison Project, Phase 3; Christensen et al., 2007), for which Lind and Kjellström (2008) provided monthly change factors for southern Sweden comparing 2071–2100 with 1961–1990 (see dark grey area in Fig. 2). Our RCA3-ensemble reflected both the seasonal patterns and the spread of the larger CMIP3-ensemble very well for precipitation. For temperature, the seasonal signatures were captured but our RCA3-ensemble showed a narrower range of projected increases than the larger CMIP3ensemble. This seems, however, reasonable as the area covered by the CMIP3-ensemble extends further north, where projected (winter) temperatures increase more. 2.3.2. Soils The dominant soils in the GSS region are developed in quaternary moraine (till) deposits and have clay contents between 10 and 20% in the topsoil (Fig. 3) as well as low to medium organic matter contents (58% with b3%, 39% with 3–5% and 3% with N 5% organic matter content). 34% of the arable land in the GSS region is classified as a discharge area, either because the subsoil is impermeable or because the area is low-lying in the landscape. These soils are assumed to have no direct recharge to groundwater and are therefore not relevant for our study (“U”-soils in Fig. 3; see also the Supplementary material). Of the remain-

Table 1 Summary of the climate input data: annual average temperature and annual precipitation amounts for present and for future climate conditions derived from the five different climate model projections (see also Kjellström et al., 2011). GCM stands for global climate model. GCM

Future climate (2070–2099). We focused on the end of this century as the climate change signal would then be most pronounced and because scenarios have been developed for future herbicide use in Sweden for the period around 2085 (Wivstad, 2010). In order to account for uncertainty in climate input data, we used an ensemble of climate scenarios (see Table 1) generated by downscaling five different global climate model (GCM) projections using the same regional climate model (RCA3; Samuelsson et al., 2011). All GCMs were run with the same

241

Present/reference BCM CCSM3 HADCM3Q0 IPSL ECHAM5

Temperature Precipitation [°C] [mm] 7.8 9.9

706 839

9.8 11.3 10.8 10.5

819 826 883 792

Reference for the GCM – Bleck et al. (1992) and Déqué et al. (1994) Collins et al. (2006) Gordon et al. (2000) Hourdin et al. (2006) Roeckner et al. (2006) and Jungclaus et al. (2006)

242

K. Steffens et al. / Science of the Total Environment 514 (2015) 239–249

ing area with recharge to groundwater, we simulated the 24 most common soil types, which covered 97% of the area (see also Fig. S2). 2.3.3. Crops Nine major crop types were simulated: grass, spring cereals, winter cereals, sugar beets, winter rape, spring rape, potatoes, peas and maize. We assumed that grassland was not treated with herbicides, so that the simulated water percolation under grass only served to dilute herbicide leaching loads from arable land. These nine crop types cover 80% of the arable land area in the region. The remaining area is covered with horticultural crops (vegetables, field fruits and berries, flowers), fruit trees, tree plantations and fallow land. Cereals are the dominant arable crops in the GSS region of Scania, with winter cereals accounting for 25.4% of the arable land and spring cereals for 20.8%. Sugar beets account on average for 9.6% of the arable area, winter rape for 7.1% and spring rape for 0.7%. Peas (and beans), potatoes and maize are minor crops covering 2.0, 1.4, and 0.5% of the arable land on average, but locally they are more widely grown. Regional average values and the spatial distribution of the crops are presented in the Supplementary material (Table S1, Fig. S3). Default crop growth parameters in MACRO-SE for the simulated crops (see Table S1) were based on typical farming practices identified for the GSS region (Myrbeck, 1998). 2.3.4. Current herbicide usage We simulated all of the 41 herbicides currently approved for use in Sweden on the 8 major arable crop types with the exception of glyphosate, and 3 others for which insufficient use data was available (bifenox, clomazone, and picloram). The required environmental fate properties of the compounds simulated (i.e. the degradation half-life, DT50, the Freundlich sorption coefficient normalized for organic carbon, K f oc , and the Freundlich exponent, nf) were taken from the Pesticide Properties DataBase (PPDB) of the University of Hertfordshire, U.K. (PPDB, 2013, see Fig. S4 and Table S2). We assumed constant values for these parameters on all soils, even though they can show considerable spatial variation (e.g. Ghafoor et al., 2011, 2013). Eight of the 37 compounds are weak acids for which sorption is pH-dependent. However, a comparison of the pKa-values for these compounds with soil pH-values in this region (mean pH of 6.7, standard deviation of 0.64) suggests that the effects of pH should be negligible in most cases. Combining the 37 compounds with the crops they are approved for use on gave 67 so-called pesticide application scenarios (PAS), defined as unique combinations of a certain pesticide compound used on a certain crop at a certain time with a certain dose. Data from two catchment areas in south-west Sweden (marked in Fig. 1) were used to support the parameterization of herbicide use in our simulations. These catchments are part of the Swedish national environmental monitoring programme for pesticides (Adielsson and Kreuger, 2007) run by SLU on behalf of the Swedish Environmental Protection Agency. In addition to monitoring pesticide residues in surface water, groundwater, stream sediment, and rain water, farmers in the catchments are interviewed each year to obtain information about field size, crops grown on the field and pesticide usage, which includes the specific substances applied, their application rate and the day of application. Data were available from 1993 to 2012 for one of the catchments and from 2005 to 2012 for the other. A comparison with official statistics for the region as a whole (SCB, 2011) suggests that the two catchments are representative for the whole region in terms of the overall fractional coverage of crops grown in the catchments, the fraction of each crop sprayed with herbicides and average overall herbicide doses. Further details on the two catchments and the data available are given in the Supplementary material. The median application dose, median application date and the crop area sprayed with a certain herbicide were calculated for each PAS. A two-week application “window” was assumed in MACRO-SE centred on the median application date. Within this window, the day closest

Fig. 2. Projected monthly change factors for (A) temperature and (B) precipitation derived for five different climate scenarios (Table 1). The change was calculated for the period 2070–2099 compared to 1970–1999 and presented as absolute difference for temperature and relative change for precipitation. The dark-grey background area represents the spread in the change factors for southern Sweden derived from an ensemble of 23 different GCMs all under the SRES A1B emission scenario (taken from Lind and Kjellström, 2008). The dotted line represents “no change”.

to the median application date that satisfied the criteria that the rainfall is less than 2 mm on the day of application and less than 10 mm the day before was chosen for each simulated year. The mean number of herbicide applications per year was also estimated from the available dataset for those compounds with multiple applications on one crop. For such compounds (mainly herbicides used in sugar beets and potatoes, but also some for maize and spring cereals), two simulations were run, one for the rounded number of applications above the average and one for the rounded number below. In a post-processing step, an average herbicide leachate concentration was calculated from these two simulations. Note that multiple applications on winter cereals and winter rape (i.e. applications both in autumn and spring to the same crop) were indirectly taken care of by having different factors for the fraction of the crop sprayed with the herbicides in spring and in autumn, respectively. In the Supplementary material, we present the parameter values used to define each PAS (Table S2), a summary of the data sources used to estimate them (Table S3) and some background information concerning variability and uncertainty in the input data (Figs. S5 and S6).

K. Steffens et al. / Science of the Total Environment 514 (2015) 239–249

243

Fig. 3. Topsoil clay content of the simulated area. The legend shows the range of clay contents in % for the simulated soils. The non-arable soils are shown transparent (background colour), the arable soils outside the simulated area are shown in light-grey, soils with no recharge to groundwater (“U”-soils) are shown in dark-grey and the soils with recharge to groundwater that were not simulated are shown in cyan blue.

2.4. Simulations and post-processing We simulated all combinations of the 67 PAS (see Table S2), the 24 soil types (see Fig. S2) and the 6 different climate time series (present conditions and five future scenarios, Table 1), which gave 9504 simulations in total. Each simulation was run for 30 years to capture climatic variations, with a preceding six year spin-up period in order to minimize the effects of unknown initial conditions, as is commonly done in simulations for pesticide registration purposes (FOCUS, 2000, 2001). From each of the simulations, we obtained predictions of accumulated percolation [m] at a depth of 2 m and the accumulated pesticide mass leaching past that depth [mg m−2]. In a subsequent post-processing step, we calculated area-weighted average total concentrations of herbicides, C pðiÞ [μg l−1], in water percolating to groundwater at a given location (map pixel p) with a soil type i: Pncrops Xnsubs  C pðiÞ ¼

j¼1

k¼1

 f cropð j;pÞ f subðk; jÞ f indð jÞ Li; j;k

Xncrops j¼1

f cropð j;pÞ W i; j

ð1Þ

where Li,j,k is the accumulated pesticide mass transported with percolating water simulated for soil i, crop j and herbicide k and Wi,j is the corresponding accumulated amount of water percolating under soil type i and crop j. fcrop(j,p) denotes the fractional coverage of crop j at location p, fsub(k,j) represents the fraction of the area of crop j sprayed with herbicide k, and find(j) is a crop specific factor that describes the change in herbicide use in the future compared to present conditions, i.e. it is equal to

1 for present conditions (for values assumed in the future scenarios, see Table S1). In order to calculate the leachate concentrations for any given crop, Eq. (1) was adapted to only account for one crop and all the herbicides used on that crop. Likewise, to calculate the leachate concentration for a specific herbicide, only that herbicide and all crops on which it is applied were included. Thus, the sum of all concentrations for single herbicides or single crops equals the total concentration calculated with Eq. (1). 2.5. Comparing simulations with measurements For various reasons, it is difficult to quantitatively compare our simulations with groundwater monitoring data. For example, we only simulate leaching at 2 m depth, whereas groundwater samples are commonly taken much deeper and we do not account for any dissipation processes in the deeper vadose zone or in groundwater that reduce the concentration of pesticides. Furthermore, we have neither information on historical land use and pesticide applications in groundwater sampling locations nor the recharge area that could influence samples taken from a groundwater well. In this respect, most of the monitoring data is available for drinking water abstraction areas, for which the regulations for pesticide usage in Sweden are usually stricter to ensure groundwater protection. Many of these wells may be sited in nonagricultural land (e.g. forested areas) where pesticides are not applied. The likelihood of detecting herbicides also depends on how long the compound has been in use in relation to the age of the sampled groundwater, which can be rather old even for shallow groundwater (Åkesson

244

K. Steffens et al. / Science of the Total Environment 514 (2015) 239–249

et al., 2014). The above considerations suggest that our simulations should generally overestimate measured concentrations in groundwater, although monitoring data may also be affected by the occurrence of local point sources (see Kværner et al., 2014). However, we have compared our results qualitatively and in an aggregated way with available groundwater monitoring data to test the ability of MACRO-SE to distinguish between leachable and nonleachable herbicides. Results from four different groundwater monitoring campaigns in the regions of Scania and Halland were used for the comparison with modelling results. For the two SLU-monitoring catchments (see Fig. 1), data from groundwater samples were available for the years 2004–2012 (CKB, 2014). In both catchments, groundwater samples were collected 4 times per year from four shallow monitoring wells in two clusters in Halland (2 and 3 m deep) and Scania (3–6 m deep) and in a deeper well (3 and 4.5–7 m in Halland and Scania, respectively). In 2012, all but 5 of the substances that we modelled were analysed (Lindström et al., 2013). The limits of detection (LOD) are rather low for all substances (between 0.001 and 0.01 μg l−1). The county board of Scania analysed 33 pesticides, including 11 of the herbicides we simulated, from 141 wells between 2007 and 2010 (Virgin, 2012). The depth of the sampled groundwater was between b2 m and N65 m. The county board of Halland analysed for 46 different pesticides in 64 groundwater samples taken between 2008 and 2011 (Löfgren and Tollebäck, 2012). The typical LOD for both sets of samples was 0.01 μg l−1. The fourth source of pesticide analyses in groundwater samples was a survey of the drinking water quality in 19 private groundwater wells in an area of Halland dominated by agriculture (Larsson et al., 2013). The wells were between 3–4 m and 110 m deep. The study was performed in 2011 and 104 pesticides were analysed with the same methods and detection limits as described by Lindström et al. (2013). If a herbicide was found at least once, it was classified as detected. In order to compare with measurements, we censored the simulation results based on a typical LOD of 0.01 µg l-1 used for the monitoring, i.e. simulations for any location in the region with concentrations above this LOD were treated as virtual detects and the others as virtual non-detects. 2.6. Scenarios for herbicide use in the future For future conditions, three different scenarios were defined. Scenario (A) was based on current crop distributions and herbicide use and only assessed direct effects, while the other two scenarios accounted for both direct and indirect effects of climate change due to changes in crop distribution and herbicide use. In each case, the five future climate model scenarios were used to represent climate input uncertainty. Scenario (B): “change in climate and land-use”. The indirect effects were represented as changes in land-use and an expected shift in cropping patterns from spring cereals and spring rape to autumnsown cereals and winter rape and from grassland to maize that would follow farmers' attempts to maximize their net income (Eckersten et al., 2008). In Scania in 2013, yields of autumn sown wheat were 30% higher than yields of spring wheat, while yields of autumn-sown rape exceeded spring rape yields by 42% for the same region (SCB, 2013b). We based our calculations on the assumptions described in Wivstad (2010) and reduced the area of spring cereals and spring rape by 60% and 100%, respectively. To compensate, the area of winter cereals and winter rape were increased in the proportion of 3:2. We also decreased the area of grassland by 50% and compensated by an equivalent increase in the area of maize. As the cultivation of the C4plant maize is temperature limited in Sweden under present climatic condition, it has the potential to increase significantly in a warmer climate (Trnka et al., 2011) to partly replace grass silage as a fodder crop. This scenario follows current agro-economic conditions and

assumes a constant area of arable land, which is a worst-case assumption from the point of view of herbicide leaching risks to groundwater, as the current long-term trend in this region is a loss of arable land due to re-afforestation and urbanization (SCB, 2013a). Scenario (C): “change in climate, land-use and herbicide usage”. In addition to a change in crop distribution, we assumed that the likely increase in weed pressure due to more favourable conditions for weed growth in a warmer climate (e.g. Patterson et al., 1999; Olesen et al., 2012) would be met by an increase in herbicide use. This increase in herbicide use can either be due to an increase in the proportion of the crop area sprayed or due to more frequent spraying on the same crop. We estimated change factors (see find(j) in Eq. (1)) based on the herbicide use indicators presented by Wivstad (2010), which were derived from literature information and today's herbicide use further south in Europe, as well as from the expert judgement of researchers, agricultural extension advisors and other plant protection specialists. These change factors vary from 1.035 to 1.59 for the individual crops (see Table S1) and correspond to an overall increase in the use of herbicides in the whole region by 45%. This scenario is therefore based on the idea of business as usual in terms of socio-economic factors influencing herbicide use, and includes the direct effects of climate change and indirect agronomic effects that influence herbicide use in the future through changes in crop distribution and weed pressure. Not all potential indirect effects were (or can be) assessed in our study. The factors that we included here were chosen based on a combination of high likelihood of occurrence, ease of parameterization of the model, a high sensitivity for herbicide leaching and computational capacities. We did not account for changes in plant development although there are several studies that suggest an earlier start of the growing season in Sweden (Olesen et al., 2012; Trnka et al., 2011; Fogelfors et al., 2009). This is partly because it is difficult to assess when sowing and herbicide applications would occur in the future, as not only temperature but also soil moisture affects the timing of such management practices (Trnka et al., 2011; Fogelfors et al., 2009). In addition, we carried out a preliminary sensitivity analysis, which showed that the effects of changes in sowing dates and application timings of up to 2 weeks would be minor compared to the indirect effects that were included in the study (results not shown). These analyses also showed that the effects of expected changes in soil organic carbon content of up to 3% of the current value (Lugato et al., 2014) would be negligible. Other factors that might influence pesticide fate and transport in the future include changes in crop production and soil management systems due to political and socio-economic factors and new technologies (e.g. new crop varieties). In this respect, our simulations, which are based on the herbicides in use in Sweden today, can be considered to represent a worst-case scenario for the future. This is because it seems unlikely that the registration procedures, which control the compounds approved for use, would become less strict than they are today, given the present state of societal concern for the environment. Likewise, the current regulatory climate in both Sweden and the EU (e.g. the Water Framework Directive and the Directive for the Sustainable Use of Pesticides) is putting increasing pressure on farmers to reduce their use of pesticides (e.g. through adoption of Integrated Pest Management (IPM) practices). 3. Results and discussion 3.1. Simulated percolation — present and future conditions On average over the region we simulated an annual percolation of 97 mm under present conditions, which is 13.7% of the average annual precipitation (cf. Table 1 with Table S4). For future conditions, the rainfall increased by 12–25% depending on the climate scenario, while annual percolation increased slightly less (by 8–23%) and percolation as a fraction of the precipitation was slightly reduced (13.2–13.6% for the different climate scenarios). More details on the simulated water

K. Steffens et al. / Science of the Total Environment 514 (2015) 239–249

balance components are given in the Supplementary material (Fig. S7 and Table S4).

245

in the relative contributions of the different crops. Thus, concentrations are dominated by spring cereals, peas and winter rape, followed by sugar beets and winter cereals (see Fig. S10).

3.2. Simulated herbicide leaching to groundwater — present conditions 3.3. Comparison of modelled results with monitoring data Fig. 4A shows the simulated herbicide concentrations leaching to groundwater, which range from zero to 1.4 μg l− 1 with a mean of 0.45 μg l−1. It also illustrates spatial differences in leaching risks and identifies potential hot-spots for herbicide leaching towards groundwater at the regional scale. The average concentration of herbicides simulated in leachate from different soils is larger for soils with lower organic carbon contents due to weaker sorption and also shows a general increase with increasing topsoil clay contents (compare Fig. 3 with Fig. 4A; and see Fig. S8). This is because the parameter estimation routines in MACRO-SE predict stronger macropore flow in clay soils and sorption is depending on organic carbon content (Moeys et al., 2012). Of the 37 simulated herbicides, only 7 compounds gave simulated concentrations above an assumed LOD of 0.01 μg l−1 on more than 5% of the entire area (Fig. 5). Total leachate concentrations are clearly dominated by 4 herbicides: clopyralid, bentazone, metazachlor, and metamitron (see Fig. S9). These compounds had simulated concentrations above the EU drinking water guideline value for single compounds of 0.1 μg l−1 on 95%, 57%, 3.7% and 2%, respectively, of the arable land in the GSS region (not shown here). Clopyralid is widely used on both spring and autumn sown cereals and rape, as well as on maize. Bentazone is used on peas and maize, metazachlor is used on spring and winter rape and metamitron is used on sugar beets. The contribution of these herbicides to the overall leaching is of course also reflected

Table 2 shows that 10 herbicides were correctly classified as “leachable” (i.e. detected within monitoring or measuring campaigns and simulated with leachate concentrations above the LOD) and 19 were correctly classified as “non-leachable”. Thus, MACRO-SE was able to distinguish rather successfully between leachable and non-leachable compounds, with 88% correctly classified (Table 2). Even if a lower LOD was considered (as used in the SLU-monitoring), only 2 additional compounds would be wrongly classified as leachable although they were not detected, which would slightly reduce the proportion of correctly classified compounds to 82%. Of the “leachable” compounds, bentazone was most frequently detected in groundwater samples (e.g. in 10% of the samples taken by the Scania county board) and is also regularly detected in groundwater elsewhere in Europe (e.g. Loos et al., 2010). Of the 12 compounds simulated to leach at concentrations above LOD (see Fig. 5) only flurtamone and propoxicarbazone-sodium were incorrectly classified as they were not detected in any of the groundwater samples. One reason may be that these compounds were first approved for use in 2002 and 2005 respectively, which has given little time for leaching, especially to those monitoring wells that abstract deeper groundwater. The remaining 2 herbicides, diflufenican and propyzamide, were incorrectly classified as non-leachable although they were detected at least once in a groundwater sample. Point sources

Fig. 4. Herbicide concentration [μg l−1] in leachate to groundwater for reference conditions and the absolute change in herbicide concentration from the present to the future based on the different climate model projections (scenario A). Negative values denote a reduction in future herbicide concentrations and positive values an increase in predicted future herbicide concentrations.

246

K. Steffens et al. / Science of the Total Environment 514 (2015) 239–249

could explain these detections as both of these compounds are strongly sorbing (see Table S2). Four of the 37 compounds (clethodim, desmedipham, diquat dibromide, and mesotrione) were not analysed in any of the monitoring programmes and were therefore excluded from this comparison.

Table 2 Confusion matrix that summarizes the comparison of our modelling results with measurements. The simulations were censored based on a typical detection limit (LOD) of 0.01 μg l−1. Sim. conc. NLOD denotes the virtual (simulated) detects whereas sim. conc. bLOD denotes the virtual (simulated) non-detects.

Sim. conc. >LOD

3.4. Herbicide concentrations in future scenarios Fig. 6 summarizes the leachate concentrations projected for the three future scenarios compared with the reference simulations for current conditions. Under present conditions, 35% of the land area in the studied region had simulated concentrations above the EU drinking water limit of 0.5 μg l− 1. For scenario A, the results from four of the five climate model projections were very similar, showing only small changes from present climate conditions (Figs. 4 and 6). Thus, for these models, increased degradation rates due to higher temperatures were more or less balanced by increased leaching risks due to higher rainfall (Fig. 2). In contrast, the ECHAM5 model gave significant reductions in leachate concentrations, probably because it projects the smallest increases in annual rainfall and decreases in summer rainfall (see Table 1 and Fig. 2). These differences amongst climate projections emphasize the importance of aggregated or ensemble predictions. The area projected by the ensemble mean to be above the EU drinking water limit decreased slightly from 35% to 31% (Fig. 6A, dark grey line; see also Fig. S11). If indirect effects were considered, the proportion of the land area with leachate concentrations above the EU limit increased to 50% when projected land-use changes were taken into account (scenario B; Fig. 6B, dark grey line) and doubled to 70% if additional increases in herbicide usage were assumed (scenario C; Fig. 6C, dark grey line). Interestingly, the projections based on the five climate models differed much more when indirect effects on leaching were also considered (Fig. 6), which suggests strong interaction effects between climate variables, land-use and herbicide use. Nevertheless, for scenario (C) all climate models showed higher leachate concentrations in the future. Thus, the uncertainty due to climate input data does not overshadow the projected indirect effects of climate change (i.e. land-use changes and increases in herbicide use). It supports the hypothesis put forward by Bloomfield et al. (2006), Beulke et al. (2007) and Steffens et al. (2013, 2014), that indirect effects of climate change might be more important

Fig. 5. Fraction of the simulated arable area for which modelled concentrations of herbicides exceeded the LOD of 0.01 μg l−1.

Detected Not detected sum

Sim. conc.
sum

10

2

12

2

19

21

12

21

33

than the direct effects of climate change. It is also in line with the conclusions of Kattwinkel et al. (2011) for surface water exposure to insecticides. Although the differences in the cumulative probability distributions of projected leachate concentrations are small for four of the five climate models in scenario A (Fig. 6A), Fig. 4 shows that their spatial patterns can differ markedly. As the only difference between the climate scenarios is the difference in monthly temperatures and precipitation amounts, these different spatial patterns reflect the importance of the interplay of mean changes in temperature and precipitation, with

Fig. 6. Cumulative distribution function of the herbicide leachate concentrations [μg l−1] aggregated over the entire simulated area (each simulated pixel). Each curve represents the fraction of the land that is predicted to have total leachate concentrations exceeding a specific value. Each subplot shows one future herbicide use scenario. The black line denotes the reference situation for present conditions (identical in all subplots), the different coloured lines are the results based on the 5 climate model projections and the dark grey line shows the ensemble prediction generated from the different climate model projections.

K. Steffens et al. / Science of the Total Environment 514 (2015) 239–249

other factors such as soil type, crop distributions and compound properties. Leachate concentrations increased in areas where clay soils dominate (compare Fig. 4 with Fig. 3). On these soils, increased precipitation triggers more macropore flow in the model (Moeys et al., 2012) and thus increases leaching losses. Correlations between soil properties and the simulated changes in herbicide concentrations for scenario (A) are presented in Fig. 7A, B. Similarly, Fig. 7C, D relates the projected changes in climate variables to the simulated changes in concentrations. Fig. 7 shows that the change in projected leachate concentrations is most strongly correlated with the clay content of the soil (r = 0.75). It is also positively correlated with the change in precipitation amounts in spring, the average change throughout the year (both with r = 0.41) and with the change in precipitation during the summer months (mainly August, r = 0.38). Changes in winter precipitation were negatively correlated with the changes in concentrations (r = −0.30), probably due to the dilution effect of an increase in percolation. As might be expected, changes in temperature showed negative correlations with the change in leachate concentrations (r = −0.23 for annual temperature, −0.22 for autumn temperatures and −0.21 for summer temperatures).

247

• MACRO-SE is an appropriate modelling tool for such modelling studies as it distinguished successfully between leachable and non-leachable herbicide compounds, despite potentially large uncertainties. • Only a few herbicides and crop types dominate herbicide concentrations in leachate. • It is very important to account for the indirect effects of climate change on the leaching of herbicides alongside the direct effects, as the risks for groundwater contamination can be significantly affected. In our study, the estimated direct effects of climate change were small, while changes in cropping patterns and herbicide use were projected to double the area at risk of groundwater contamination. • There is a need for ensemble modelling as there are strong interactions between the climatic factors and the future scenarios on cropping patterns, herbicide use and leaching risks. • Currently vulnerable areas with medium to high clay contents might become even more vulnerable in the future as changes in herbicide leaching in the future were positively correlated with soil clay content.

4. Conclusions The following conclusions can be drawn from this regional scale assessment of climate change effects on herbicide leaching to groundwater with MACRO-SE:

Although the uncertainty in the estimates of indirect effects of climate change is high and largely unquantifiable, our study clearly implies an urgent need to adopt policies, mitigation practices and

Fig. 7. Relative change in herbicide concentration from the present to the future for scenario (A) in relation to clay content (subplot A), organic carbon content (B), changes in mean annual temperature (C) and annual precipitation (D). For a given climate model (see legend), each dot represents the relative change in concentration between the present and the future for one of the 24 soil types that were simulated. The dotted line marks no-change, values above denote an increase in future concentrations compared to present and values below a reduction.

248

K. Steffens et al. / Science of the Total Environment 514 (2015) 239–249

management strategies that will avoid future contamination of groundwater and drinking water supplies. This could include developments in agricultural practices and production systems (ecological or low-input farming, IPM) or in pesticide chemistry (improved fate properties, reduced application rates) that enable significant reductions in herbicide use and mitigate leaching risks. Acknowledgement This study was financed by the Swedish Farmers' Foundation for Agricultural Research (SLF) (H0833488) and by strategic funding from the Faculty of Natural Resources and Agricultural Sciences at the Swedish University of Agricultural Sciences (SLU), Sweden (5021002 - SoilWater-Landscape-Society). We thank the Centre for Chemical Pesticides (CKB) at SLU for providing the datasets (soil, crop, pesticide use) and the modelling software. We would also like to thank our colleague Elsa Coucheney and three anonymous reviewers for valuable comments on the manuscript. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.scitotenv.2014.12.049. References Adielsson, S., Kreuger, J., 2007. Environmental monitoring of pesticides in Sweden. Establishing a Nordic Pesticide Monitoring Network. Nordic Council of Ministers, pp. 65–74 (Vol. 2007:514 of TemaNord). Ahmadi, M., Records, R., Arabi, M., 2014. Impact of climate change on diffuse pollutant fluxes at the watershed scale. Hydrol. Process. 28 (4), 1962–1972. Åkesson, M., Bendz, D., Carlsson, C., Sparrenbom, C.J., Kreuger, J., 2014. Modelling pesticide transport in a shallow groundwater catchment using tritium and helium-3 data. Appl. Geochem. 50, 231–239. Bengtsson, L., 2010. The global atmospheric water cycle. Environ. Res. Lett. 5, 025002 (8 pp.). Beulke, S., Boxall, A.B.A., Brown, C.D., Thomas, M., Falloon, P., 2007. Impacts of climate change on pesticide transport to groundwater and surface water. In: Del Re, A.A.M., Capri, E., Fraoulis, G., Trevisan, M. (Eds.), Environmental Fate and Ecological Effects. XIII Symposium Pesticide Chemistry, pp. 444–451. Bleck, R., Rooth, C., Hu, D., Smith, L.T., 1992. Salinity-driven thermocline transients in a wind-forced and thermohaline-forced isopycnic coordinate model of the NorthAtlantic. J. Phys. Oceanogr. 22 (12), 1486–1505. Bloomfield, J.P., Williams, R.J., Gooddy, D.C., Cape, J.N., Guha, P., 2006. Impacts of climate change on the fate and behaviour of pesticides in surface and groundwater—a UK perspective. Sci. Total Environ. 369 (1–3), 163–177. Centofanti, T., Hollis, J.M., Blenkinsop, S., Fowler, H.J., Truckell, I., Dubus, I.G., Reichenberger, S., 2008. Development of agro-environmental scenarios to support pesticide risk assessment in Europe. Sci. Total Environ. 407 (1), 574–588. Christensen, J.H., Carter, T.R., Rummukainen, M., Amanatidis, G., MAY, 2007. Evaluating the performance and utility of regional climate models: the PRUDENCE project. Clim. Chang. 81 (1), 1–6. CKB, 2014. http://www.slu.se/ckb. Collins, W.D., Bitz, C.M., Blackmon, M.L., Bonan, G.B., Bretherton, C.S., Carton, J.A., Chang, P., Doney, S.C., Hack, J.J., Henderson, T.B., Kiehl, J.T., Large, W.G., McKenna, D.S., Santer, B.D., Smith, R.D., 2006. The Community Climate System Model version 3 (CCSM3). J. Clim. 19 (11), 2122–2143 (JUN 1). Delpla, I., Jung, A.-V., Baures, E., Clement, M., Thomas, O., 2009. Impacts of climate change on surface water quality in relation to drinking water production. Environ. Int. 35 (8), 1225–1233. Déqué, M., Dreveton, C., Braun, A., Cariolle, D., SEP, 1994. The ARPEGE/IFS atmosphere model — a contribution to the french community climate modeling. Clim. Dyn. 10 (4–5), 249–266. Eckersten, H., Karlsson, S., Torssel, B., 2008. Climate change and agricultural land use in Sweden: a literature review. Tech. Rep. 7. Department of Crop Production Ecology, Swedish University of Agricultural Sciences, Uppsala, Sweden. FOCUS, 2000. FOCUS groundwater scenarios in the EU review of active substances. Report of the FOCUS Groundwater Scenario Workgroup, EC Document Reference Sanco/321/ 2000 rev.2 (202 pp.). FOCUS, 2001. FOCUS surface water scenarios in the EU evaluation process under 91/414/ EEC. Report of the FOCUS Working Group on Surface Water Scenarios, EC Document Reference SANCO/4802/2001-rev.2 (245 pp.). Fogelfors, H., Wivstad, M., Eckersten, H., Holstein, F., Johansson, S., Verwijst, T., 2009. Strategic analysis of Swedish agriculture — production systems and agricultural landscapes in a time of change. Tech. Rep. 10. Department of Crop Production Ecology, Swedish University of Agricultural Sciences. Fowler, H.J., Blenkisop, S., Tebaldi, C., 2007. Review — linking climate change modelling to impacts studies: recent advances in downscaling techniques for hydrological model-

ling. Int. J. Climatol. 27, 1547–1578. Gerke, H.H., Van Genuchten, M.T., 1996. Macroscopic representation of structural geometry for simulating water and solute movement in dual-porosity media. Adv. Water Resour. 19 (6), 343–357. Ghafoor, A., Jarvis, N.J., Thierfelder, T., Stenström, J., 2011. Measurements and modeling of pesticide persistence in soil at the catchment scale. Sci. Total Environ. 409 (10), 1900–1908. Ghafoor, A., Jarvis, N.J., Stenstrom, J., AUG, 2013. Modelling pesticide sorption in the surface and subsurface soils of an agricultural catchment. Pest Manag. Sci. 69 (8), 919–929. Gordon, C., Cooper, C., Senior, C.A., Banks, H., Gregory, J.M., Johns, T.C., Mitchell, J.F.B., Wood, R.A., FEB, 2000. The simulation of SST, sea ice extents and ocean heat transports in a version of the Hadley Centre coupled model without flux adjustments. Clim. Dyn. 16 (2–3), 147–168. Henriksen, H.J., Rosenbom, A., van der Keur, P., Olesen, J.E., Jørgensen, L.J., Kjær, J., Sonnenborg, T., Christensen, O.B., 2013. Prediction of climatic impacts on pesticide leaching to the aquatic environments. Tech. Rep.Miljøstyrelsen Hourdin, F., Musat, I., Bony, S., Braconnot, P., Codron, F., Dufresne, J.-L., Fairhead, L., Filiberti, M.-A., Friedlingstein, P., Grandpeix, J.-Y., Krinner, G., LeVan, P., Li, Z.-X., Lott, F., DEC, 2006. The LMDZ4 general circulation model: climate performance and sensitivity to parametrized physics with emphasis on tropical convection. Clim. Dyn. 27 (7–8), 787–813. Jarvis, N.J., 1989. A simple empirical model of root water uptake. J. Hydrol. 107 (1–4), 57–72. Jarvis, N., Larsbo, M., 2012. MACRO (V5.2): model use, calibration, and validation. Trans. ASABE 55 (4), 1413–1423. Jungclaus, J.H., Keenlyside, N., Botzet, M., Haak, H., Luo, J.-J., Latif, M., Marotzke, J., Mikolajewicz, U., Roeckner, E., AUG, 2006. Ocean circulation and tropical variability in the coupled model ECHAM5/MPI-OM. J. Clim. 19 (16), 3952–3972. Kattwinkel, M., Kühne, J.-V., Foit, K., Liess, M., 2011. Climate change, agricultural insecticide exposure, and risk for freshwater communities. Ecol. Appl. 21 (6), 2068–2081. Kjellström, E., Nikulin, G., Hansson, U., Strandberg, G., Ullerstig, A., 2011. 21st century changes in the European climate: uncertainties derived from an ensemble of regional climate model simulations. Tellus 63 (1), 24–40. Koleva, N., Schneider, U.A., Tol, R.S.J., 2009. The impact of weather variability and climate change on pesticide application in the US: an empirical investigation. Hamburg University and Centre for Marine and Atmospheric Science, Hamburg (http://www.fnu. zmaw.de). Kværner, J., Eklo, O.M., Solbakken, E., Solberg, I., Sorknes, S., 2014. An integrated approach for assessing influence of agricultural activities on pesticides in a shallow aquifer in south-eastern Norway. Sci. Total Environ. 499, 520–532. Larsbo, M., Jarvis, N.J., 2003. MACRO 5.0. a model of water flow and solute transport in macroporous soil. Technical description. Emergo 2003:6. Tech. Rep.Swedish University of Agricultural Science, Dept. of Soil Science, Div. of Environmental Physics, Uppsala Larsbo, M., Roulier, S., Stenemo, F., Kasteel, R., Jarvis, N.J., 2005. An improved dualpermeability model of water flow and solute transport in the vadose zone. Vadose Zone J. 4, 398–406. Larsson, M., Graaf, S., Nanos, T., Boye, K., Kreuger, J., 2013. Undersökning av växtskyddsmedel i privata dricksvattenbrunnar från ett jordbruksområde i södra halland. CKB Rapport 2013:2. Centre for Chemical Pesticides, Swedish University of Agricultural Sciences, Uppsala (report in Swedish). Lewan, E., Kreuger, J., Jarvis, N.J., 2009. Implications of precipitation patterns and antecedent soil water content for leaching of pesticides from arable land. Agric. Water Manag. 96 (11), 1633–1640. Lind, P., Kjellström, E., 2008. Temperature and precipitation changes in Sweden; a wide range of model-based projections for the 21st century. Reports Meteorology and Climatology, SMHI, SE-60176 Norrköping, Sweden. 113 (50 pp.). Lindström, B., Larsson, M., Nanos, T., Kreuger, J., 2013. Resultat från miljöövervakningen av bekämpningsmedel (växtskyddsmedel) årssammanställning 2012. Tech. Rep.Department of Aquatic Sciences and Assessment, Swedish University of Agricultural Sciences, Uppsala (report in Swedish) Löfgren, H., Tollebäck, E., 2012. Bekämpningsmedel i hallänska yt- och grundvatten. Tech. Rep.Länsstyrelsen i Hallands län, Halmstad, Sweden (report in Swedish) Loos, R., Locoro, G., Comero, S., Contini, S., Schwesig, D., Werres, F., Balsaa, P., Gans, O., Weiss, S., Blaha, L., Bolchi, M., Gawlik, B.M., 2010. Pan-european survey on the occurrence of selected polar organic persistent pollutants in ground water. Water Res. 44 (14), 4115–4126. Lugato, E., Agos, P.P., Bampa, F., Jones, A., Montanarella, L., 2014. A new baseline of organic carbon stock in European agricultural soils using a modelling approach. Glob. Chang. Biol. 20, 313–326. Moeys, J., Larsbo, M., Bergström, L., Brown, C.D., Coquet, Y., Jarvis, N.J., 2012. Functional test of pedotransfer functions to predict water flow and solute transport with the dual-permeability model MACRO. Hydrol. Earth Syst. Sci. 16 (7), 2069–2083. Myrbeck, Å., 1998. Swedish agricultural and horticultural crops. Tech. Rep. 1/98. Kemikalieinspektionen, Solna, Sweden. Nakićenović, N., Swart, R.E., 2000. Special report on emissions scenarios. A Special Report of Working Group III of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, UK. Nolan, B.T., Dubus, I.G., Surdyk, N., Fowler, H.J., Burton, A., Hollis, J.M., Reichenberger, S., Jarvis, N.J., SEP, 2008. Identification of key climatic factors regulating the transport of pesticides in leaching and to tile drains. Pest Manag. Sci. 64 (9), 933–944. Olesen, J.E., Trnka, M., Kersebaum, K.C., Skjelvåg, A.O., Seguin, B., Peltonen-Sainio, P., Rossi, F., Kozyra, J., Micale, F., 2011. Impacts and adaptation of european crop production systems to climate change. Eur. J. Agron. 34 (2), 96–112. Olesen, J.E., Börgesen, C.D., Elsgaard, L., Palosuo, T., Rötter, R.P., Skjelvåg, A.O., PeltonenSainio, P., Börjesson, T., Trnka, M., Ewert, F., Siebert, S., Brisson, N., Eitzinger, J., van

K. Steffens et al. / Science of the Total Environment 514 (2015) 239–249 Asselt, E.D., Oberforster, M., van der Fels-Klerx, H.J., 2012. Changes in time of sowing, flowering and maturity of cereals in Europe under climate change. Food Addit. Contam. Part A: Chem. Anal. Control Expo. Risk Assess. 29, 1527–1542. Olsson, J., 1998. Evaluation of a scaling cascade model for temporal rainfall disaggregation. Hydrol. Earth Syst. Sci. 2 (1), 19–30. Patterson, D.T., Westbrook, J.K., Joyce, R.J.V., Lingren, P.D., Rogasik, J., 1999. Weeds, insects, and diseases. Clim. Chang. 43 (4), 711–727. PPDB, 2013. The Pesticide Properties DataBase (PPDB) developed by the Agriculture & Environment Research Unit (AERU). University of Hertfordshire (funded by UK national sources and through EU-funded projects, 2006–2013. http://sitem.herts.ac.uk/aeru/ iupac/index.htm). Roeckner, E., Brokopf, R., Esch, M., Giorgetta, M., Hagemann, S., Kornblueh, L., Manzini, E., Schlese, U., Schulzweida, U., AUG, 2006. Sensitivity of simulated climate to horizontal and vertical resolution in the ECHAM5 atmosphere model. J. Clim. 19 (16), 3771–3791. Roos, J., Hopkins, R., Kvarnheden, A., Dixelius, C., JAN, 2011. The impact of global warming on plant diseases and insect vectors in Sweden. Eur. J. Plant Pathol. 129 (1), 9–19. Samuelsson, P., Jones, C.G., Willén, U., Ullerstig, A., Gollvik, S., Hansson, U., Jansson, C., Kjellström, E., Nikulin, G., Wyser, K., 2011. The Rossby Centre Regional Climate model RCA3: model description and performance. Tellus 63A, 4–23. SCB, 2011. Växtskyddsmedel i jord- och trädgårdsbruket 2010 - användning i grödor (Plant Protection Products in Agriculture and Horticulture — Use in Crops). Statistics Sweden (MI 31 SM 1101, report in Swedish with English summary, http://www.scb. se/statistik/MI/MI0502/2009I10/MI0502_2009I10_SM_MI31SM1101.pdf). SCB, 2013 aa. Markanvändningen i sverige, sjätte utgåvan (Land Use in Sweden, Sixth Edition). Statistics Sweden (Report in Swedish with English summary, http://www.scb. se/Statistik/_Publikationer/MI0803_2010A01B_BR_00_MI03BR1301.pdf). SCB, 2013b. Skörd av spannmål, trindsäd och oljeväxter 2013 — preliminär statistik för län och riket (Production of Cereals, Dried Pulses and Oilseeds in 2013 — Preliminary Statistics for Counties and the Whole Country). Statistics Sweden (JO 19 SM 1302, report in Swedish with English summary, http://www.scb.se/statistik/MI/MI0502/ 2009I10/MI0502_2009I10_SM_MI31SM1101.pdf).

249

SJV, 2014. Skörd av jordbruksväxter (Harvest of arable crops). Statistical Database of the Swedish Ministry of Agriculture, http://statistik.sjv.se/PXWeb/Menu.aspx?px_ tableid=JO0601M1.px&px_path=Jordbruksverkets. Solheim, A.L., Austnes, K., Eriksen, T.E., Seifert, I., Holen, S., 2010. Climate change impacts on water quality and biodiversity. Background Report for EEA European Environment State and Outlook Report 2010. European Environment Agency/European Topic Centre on Water (ETC/W), Prague, Czech Republic. Steffens, K., Larsbo, M., Moeys, J., Jarvis, N., Lewan, E., 2013. Predicting pesticide leaching under climate change: importance of model structure and parameter uncertainty. Agric. Ecosyst. Environ. 172, 24–34. Steffens, K., Larsbo, M., Moeys, J., Kjellstöm, E., Jarvis, N., Lewan, E., 2014. Modelling pesticide leaching under climate change: parameter vs. climate input uncertainty. Hydrol. Earth Syst. Sci. 18, 479–491. Stenrød, M., Perceval, J., Benoit, P., Almvik, M., Bolli, R.I., Eklo, O.M., Sveistrup, T.E., Kværner, J., 2008. Cold climatic conditions: effects on bioavailability and leaching of the mobile pesticide metribuzin in a silt loam soil in Norway. Cold Reg. Sci. Technol. 53, 4–15. Trnka, M., Olesen, J.E., Kersebaum, K.C., Skjelvag, A.O., Eitzinger, J., Seguin, B., PeltonenSainio, P., Rotter, R., Iglesias, A., Orlandini, S., Dubrovsky, M., Hlavinka, P., Balek, J., Eckersten, H., Cloppet, E., Calanca, P., Gobin, A., Vucetic, V., Nejedlik, P., Kumar, S., Lalic, B., Mestre, A., Rossi, F., Kozyra, J., Alexandrov, V., Semeradova, D., Zalud, Z., 2011. Agroclimatic conditions in Europe under climate change. Glob. Chang. Biol. 17 (7), 2298–2318. Virgin, H., 2012. Grundvattenkvalitet i skåne län - utvärdering av regional provtagning 2007–2010. Tech. Rep.Länsstyrelsen i Skåne län, Malmö, Sweden (report in Swedish) Vonberg, D., Vanderborght, J., Cremer, N., Pütz, T., Herbst, M., Vereecken, H., 2014. 20 years of long-term atrazine monitoring in a shallow aquifer in western Germany. Water Res. 50, 294–306. Wivstad, M., 2010. Klimatförändringarna - en utmaning för jordbruket och giftfri miljö. Tech. Rep.Kemikalieinspektionen (report in Swedish).