The environmental and economic impacts of nitrate leaching from agricultural soils highlight the need for accurate prediction to support effective nitrogen management. This study evaluated six machine learning (ML) models—Multiple Linear Regression, Elastic Net, K-Nearest Neighbors, Decision Trees, Extra Trees, and Multi-Layer Perceptron—using 2993 field observations to predict nitrate leaching. The models incorporated crop sequence, manure and fertilizer inputs, soil, and percolation and are benchmarked against the empirical NLES5 model used in Danish nitrogen regulation. Four ML models achieved higher predictive accuracy than NLES5 when evaluated on independent test data, with the Extra Trees model achieving the best performance (R2 = 0.63, RMSE = 23.3 kg N ha−1), exceeding NLES5 (R2 = 0.39, RMSE = 29.8 kg N ha−1). Model interpretability analyses identified winter percolation, winter vegetation cover, and soil type as key drivers of nitrate leaching. The Extra Trees model was further evaluated using scenario analyses of long-term leaching trends, marginal responses to spring-applied mineral nitrogen, and spatial patterns within the Bolbro Bæk catchment. Findings highlight the potential of ML models to improve nitrate leaching predictions in ungauged areas. Future research could incorporate additional variables, including crop yield and tillage practices, to enhance model accuracy and support sustainable agricultural practices that maintain productivity while reducing nitrate leaching.

Context: Nitrogen is an essential macronutrient in agriculture, affecting both crop yields and soil health. In Denmark, one of the most densely farmed regions in the world, excess reactive nitrogen (Nr) compounds are lost to the environment along gaseous and hydrological pathways in forms such as nitrate, ammonia, nitrogen oxides and dinitrogen. Objectives: Here, we aim to assess the effect of different field management practices (fertilisation, crop residue management or cultivation of catch crops) on environmental Nr losses and the field scale soil net GHG balance (i. e., sum of soil C stock changes and direct and indirect N2O emissions). Methods: For this purpose, highly detailed data from the Danish Agricultural Watershed Monitoring Program (LOOP-program; 2013-2019) were used in combination with the process-based model LandscapeDNDC. Results and conclusions: The results indicate that a mixture of organic and synthetic fertilisers turns soils to a stronger net sink of GHGs (similar to 70 - similar to 514 kgCO(2-)eq ha(-1) yr(-1)) compared to exclusive use of only one type of fertiliser. In addition, incorporating crop residue and cultivation of catch crops increases the nitrogen use efficiency (NUE) by 3-11 % on average and decreases environmental Nr losses. Significance: These findings emphasize the potential of targeted fertiliser, residue and catch crop management to increase the sustainability of crop production systems in Denmark.

Globally, food production for an ever-growing population is a well-known threat to the environment due to losses of excess reactive nitrogen (N) from agriculture. Since the 1980s, many countries of the Global North, such as Denmark, have successfully combatted N pollution in the aquatic environment by regulation and introduction of national agricultural one-size-fits-all mitigation measures. Despite this success, further reduction of the N load is required to meet the EU water directives demands, and implementation of additional targeted N regulation of agriculture has scientifically and politically been found to be a way forward. In this paper, we present a comprehensive concept to make future targeted N regulation successful environmentally and economically. The concept focus is on how and where to establish detailed maps of the groundwater denitrification potential (N retention) in areas, such as Denmark, covered by Quaternary deposits. Quaternary deposits are abundant in many parts of the world, and often feature very complex geological and geochemical architectures. We show that this subsurface complexity results in large local differences in groundwater N retention. Prioritization of the most complex areas for implementation of the new concept can be a cost-efficient way to achieve lower N impact on the aquatic environment.

A complete understanding of the nexus between productivity and sustainability of agricultural production systems calls for a comprehensive assessment of the nitrogen budget (NB). In our study, data from the well-monitored Danish Agricultural Watershed Monitoring Program (LOOP-program; 2013–2019) is used for a quantitative inter-comparison of three different approaches to drive the process-based model LandscapeDNDC on the regional scale. The aim is to assess how assumptions and simplifications about farm management activities at a regional scale induce previously unquantified uncertainties in the simulation of yields and the NB of cropping systems. Our findings reveal that the approach based on detailed field-level management data (A) performs the best in simulation of yield (r2 = 0.93). In contrast, the other two different data aggregation approaches (B: Sequential mono-cropping of six major crops with simulation results averaged according to proportional area, and C: simulation of 20 most frequent crop rotations) have lower correlations to the observed yields (r2 = 0.92 and 0.77, respectively) but are still statistically significant at p < 0.05 level. Notable differences arise between detailed and more aggregated crop system simulations concerning the NB, particularly concerning N losses to the environment. Compared to the detailed approach (A) (gaseous N fluxes: 24.3 kg-N ha−1 year−1; nitrate leaching: 14.7 kg-N ha−1 year−1), the aggregation approach B leads to a 31.4% over-estimation in total gaseous N fluxes (+7.6 kg-N ha−1 year−1), while nitrate leaching shows a similar average with a distinct pattern. Conversely, employing aggregation approach C leads to a 17.6% over-estimation in total gaseous fluxes (+4.3 kg-N ha−1 year−1) and a 204.9% over-estimation in nitrate leaching (+30.2 kg-N ha−1 year−1). These findings suggest that management representation should be chosen carefully because it can induce large uncertainties, especially when simulating large-scale NBs or assessing the environmental impact of cropping management. This may compromise the accuracy of national and international nutrient budgets, and preclude comparisons among different sources when the approaches for management representation differ.

Modeling of nitrate transport and retention in agricultural land use areas provides useful information to support water quality assessment and management. The accuracy and precision of model simulations are highly dependent on model input factors for which the appropriate values are generally difficult to determine and from which various uncertainties are induced into the modeling procedure. In this study, we applied a Distance-based Generalized Sensitivity Analysis (DGSA) to a high-resolution (25 × 25 m) nitrate transport and retention model for a tile-drained agricultural catchment (4.4 km2) to investigate the extent to which model input factors affect the spatially distributed nitrate retention. The input factors included the nitrate leaching from the root zone, the partitioning of nitrate into tile drainage and groundwater flux, the groundwater flux out of the catchment, the hydrogeological properties, and the denitrification rates in groundwater. The DGSA results were examined in both spatially lumped and distributed perspective. We found that the partitioning of nitrate into tile drainage and groundwater flux was the most important factor for modeling nitrate retention while the hydrogeological properties were secondary but also important. Conversely, the nitrate leaching from the root zone and denitrification rates in groundwater were noninfluential. By increasing the resolution of the DGSA analysis from catchment to model pixel, we found that input factors noninfluential on catchment scale were influential on pixel scale in discrete areas, and, as a general take-home-message, input factors influential on nitrate retention in at least 25 % of the model pixels were sensitive on catchment scale as well. Improved understanding of sensitivity of modelling nitrate retention may help the modelers and water managers to decide which input factors to prioritize in the modelling and data collection to improve the accuracy and precision of the model responses.

We investigated the utility of using synchronous measurements to create nitrogen (N) emission and retention maps of agricultural areas. Total N (TN) emissions from agricultural areas in three different Danish pilot catchments (1800–3737 ha) and within sub-catchments (100–1200 ha) were determined by a source apportionment approach. Intensive daily (main gauging stations) and fortnightly (synchronous stations) monitoring of discharge, TN, and nitrate-N (NO3-N) concentrations was conducted for two years. The groundwater N retention was calculated as the difference between a model-calculated NO3-N leaching from agricultural fields and the calculated agricultural N emission. The average annual N leaching and N emission in the three catchments amounted to 68, 48, and 58 kg N/ha and 6, 30, and 40 kg N/ha, respectively. The N retention in groundwater in the three catchments, calculated based on either TN or NO3-N emissions, amounted to 26 and 44%, 44 and 57%, and 93 and 97%, respectively, with large variations within two of the main catchments. From this study, we conclude that synchronous measurements in streams provide a good opportunity for developing local N emission and N retention maps. However, NO3-N should be used when dealing with N retention calculation at the finer resolution scale of 100–300 ha catchments.

Future development of bioeconomy is expected to change land use in the Nordic countries in agriculture and forestry. The changes are likely to affect water quality due to changes in nutrient run-off. To explore possible future land-use changes and their environmental impact, stakeholders and experts from four Nordic countries (Denmark, Finland, Norway and Sweden) were consulted. The methodological framework for the consultation was to identify a set of relevant land-use attributes for agriculture and forestry, e.g. tillage conservation effort, fertiliser use, animal husbandry, biogas production from manure, forestry management options, and implementation of mitigation measures, including protection of sensitive areas. The stakeholders and experts provided their opinions on how these attributes might change in terms of their environmental impacts on water quality given five Nordic bioeconomic scenarios (sustainability, business as usual, self-sufficiency, cities first and maximizing economic growth). A compilation methodology was developed to allow comparing and merging the stakeholder and expert opinions for each attribute and scenario. The compiled opinions for agriculture and forestry suggest that the business-as-usual scenario may slightly decrease the current environmental impact for most attributes due to new technologies, but that the sustainability scenario would be the only option to achieve a clear environmental improvement. In contrast, for the self-sufficiency scenario, as well as the maximum growth scenario, a deterioration of the environment and water quality was expected for most of the attributes. The results from the stakeholder consultations are used as inputs to models for estimating the impact of the land-use attributes and scenarios on nutrient run-off from catchments in the Nordic countries (as reported in other papers in this special issue). Furthermore, these results will facilitate policy level discussions concerning how to facilitate the shift to bioeconomy with increasing biomass exploitation without deteriorating water quality and ecological status in Nordic rivers and lakes.

Intensive agriculture has been linked to increased nitrogen loads and adverse effects on downstream aquatic ecosystems. Sustained large net nitrogen surpluses have been shown in several contexts to form legacies in soil or waters, which delay the effects of reduction measures. In this study, detailed land use and agricultural statistics were used to reconstruct the annual nitrogen surpluses in three agriculture-dominated watersheds of Denmark (600–2700 km 2 ) with well-drained loamy soils. These surpluses and long-term hydrological records were used as inputs to the process model ELEMeNT to quantify the nitrogen stores and fluxes for 1920–2020. A multi-objective calibration using timeseries of river nitrate loads, as well as other non-conventional data sources, allowed to explore the potential of these different data to constrain the nitrogen cycling model. We found the flux-weighted nitrate concentrations in the root zone percolate below croplands, a dataset not commonly used in calibrating watershed models, to be critical in reducing parameter uncertainty. Groundwater nitrate legacies built up in all three studied watersheds during 1950–1990 corresponding to ∼2% of the surplus (or ∼1 kg N ha yr −1 ) before they went down at a similar rate during 1990–2015. Over the same periods active soil nitrogen legacies first accumulated by approximately 10% of the surplus (∼5 kg N ha yr −1 ), before undergoing a commensurate reduction. Both legacies appear to have been the drivers of hysteresis in the diffuse load at the catchments’ outlet and hindrances to reaching water quality goals. Results indicate that the low cropland surpluses enforced during 2008–2015 had a larger impact on the diffuse river loads than the European Union’s untargeted grass set-aside policy of 1993–2008. Collectively, the measures of 1990–2015 are estimated to have reset the diffuse load regimes of the watersheds back to the situation prevailing in the 1960s.

Subsurface drainage systems are a dominant flow and nitrate transport pathway from fields to surface waters. Existing methods to estimate tile flow are either expensive, complex or need diverse data types. In this study, a parsimonious statistical model was derived and validated for use to obtain estimates of annual tile flow at field scale (1-10 ha) and catchment scale (200-1400 ha) in tile-drained minerogenic soils. The model was developed from tile flow and precipitation data from 38 drainage stations distributed all over Denmark. Firstly, a significant linear relationship between tile flow and precipitation was derived (R2 =0.57; P < 0.0001). The threshold value in the linear regression model (543 mm) can be viewed as a value from where tile flow is initiated, and the slope value (0.48) indicates that when the threshold value is reached, precipitation is partitioned into recharge (52%) and tile flow (48%). Secondly, the developed model was evaluated at catchment scale in 14 smaller Danish catchments and in more detail in a 15th catchment (Lilleb AE k). The evaluation showed that the model was more accurate and precise at catchment scale than at field scale. This difference might appear because the natural spatial variability in tile flow generation among fields within a catchment is averaged when evaluating the mean of multiple fields at catchment scale. Thirdly, the capability of the model to estimate transport of nitrate-N from tile drains was evaluated in Lilleb AE k at catchment scale, and here the model performed satisfactorily against the measured transport of nitrate-N (NSE=0.72 and PBIAS=-20%). We suggest that the model can be used to aid agricultural water management.

Nitrogen (N) and phosphorus (P) losses via agricultural drainage water have negative impacts on receiving water bodies and large-scale programmes to reduce nutrient losses have been established in the Nordic and Baltic countries, together with agricultural catchment monitoring programmes. This study evaluated time series (9-40 years) of data from 34 selected Nordic-Baltic catchments for spatial and temporal variations in area-specific water discharge (mm) and in concentrations and transport of total nitrogen (TN) and total phosphorus (TP).Water discharge from the catchments varied from 125 mm (Denmark) to > 1000 mm (Norway). Catchments with low TN concentrations (& LE;3 mg L-1) were dominated by clay or grass leys or were undrained with reduction of nitrate (NO3) in shallow groundwater. Catchments with high TN concentrations (& GE;10 mg L-1) had loams and cereal crops. TP concentrations were highest (& GE;0.45 mg L-1) in catchments with erosive soils, relatively high water discharge and cereal crops, and lowest (& LE;0.07 mg L-1) in catchments with permeable soils.Generalised additive mixed model (GAMM) analysis of time series of transport and flow-weighted concen-trations of TN and TP for temporal patterns revealed decreases in TN concentrations in seven catchments and increases in eight, while four had periods with opposing trends. TN concentrations decreased in Denmark and Sweden in 1990-2010, following introduction of mitigation programmes. TP concentrations decreased in eight catchments and increased in six, while one showed opposing trends. Decreases in TP coincided with improved P balance in catchments with sand and loam. To further reduce N and P losses, a tailored set of mitigation measures is needed for each combination of soil, climate, geohydrology and agricultural production. Intensive monitoring of small catchments can reveal how N and P losses relate to natural conditions and to changes in agricultural production.

The links between nitrate-nitrogen (N) leaching from agricultural fields and N measured in streams can in general terms be divided in two pathways: groundwater and a more surface-near transport (e.g. tile drains). The former with a typical slower hydrological response than the latter. Therefore, catchments with quick hydrologic pathways respond also quickly on programme of measures for N. On the other hand, catchments having bot high N-attenuation or longer N time lags makes it complicated for managers and policy makers as the response of the implemented programme of measures might both be dampened and delayed. As River Basin Management Plans (RBMPs) under the Water Framework Directive (WFD) runs in 6 years periods – such time lags might end up as an overdosing of measures. Therefore, attenuation and time lags needs to be mapped as they have major effects on the expected effects of RBMPs and its legacy for water quality. The aim of this study is to improve our understanding of N lags mapped based on 30 years of data from 160 Danish stream monitoring stations. A national wide screening for trends in annual flow-weighted total nitrogen (TN) concentrations at 163 river monitoring stations shows in most cases a downward trend (average: 30% ± 17%) during the last 30 years 1990-2019). The N-surplus has been reduced (farm gate: -44%; field: -45%) during the same period. Before 1990, the N-surplus in agriculture was increasing and started at first levelling off in the mid 1980ies. Diffuse N-sources and mostly agriculture contributed the most to TN in streams (93% ±8%) during the period 1990-2019). The reduction in the diffuse N loadings are paralleling the development of the N surplus for most Danish streams. However, in certain parts of Denmark several river monitoring stations shows a much different response, which in some cases is no response at all. Such a pattern can only be explained by N-flows in the catchments to be delayed in groundwater aquifers. Using long term data for national N-surplus a simple lag-time analysis shows that the time lags for N are long for 21 catchments (up to 20 years), medium long for N in 62 catchments and with nearly no delay for N in 80 catchments (Fig. 1). Moreover, all the stream stations experiencing long time lags are situated in the chalk and partly karstic landscapes of Denmark from the Danien period. The catchments having long delays for N shows in most cases also a very low attenuation of N in groundwater as measured N-concentrations are substantially higher than found in the streams having nearly no time lags. Therefore, we conclude that incorporation of biogeochemical and hydrologic time lag principles into water quality regulations will be necessary for providing managers and regulators with realistic expectations when implementing new policies for N. Figure 1: Map of Denmark showing catchments with short (< 20% older than 10 years), medium (20-40 % older than 10 years) and long (> 40% older than 10 years) for nitrogen in groundwater.

Nitrate leaching from agricultural soils is a considerable environmental concern related to nitrogen (N) appli-cation, and in some countries N fertilizer application is regulated by legislation to reach environmental goals. Information about the change in nitrate leaching with changes in mineral N fertilizer rates is important under such conditions. The increase in nitrate leaching due to extra mineral N application near the economic optimal N application for crop production, called marginal nitrate leaching, depends on a number of factors like soil type, crop types, climatic conditions and N fertilization rates. In this study, we collected published experimental data from 44 site-years from Denmark and 31 site-years from other European countries (Germany, Sweden, UK) with measured nitrate leaching at increasing mineral N fertilizer rates. We focused on obtaining marginal nitrate leaching around optimal N rates based on available information from the different field experiments. The measured nitrate leaching varied from 3 to 92 kg N ha(-1) in the Danish dataset. For the European data set, it varied from 1 to 124 kg N ha(-1). The median and mean of the estimated marginal nitrate leaching at the optimal N rates were 17.0% and 20.5%, respectively, in the Danish dataset and 9.0% and 14.9%, respectively, in the European dataset. At the optimal N rate, a positive relationship was found between yearly nitrate leaching and marginal nitrate leaching. Both nitrate leaching and marginal nitrate leaching at the optimal N rate were positively correlated with precipitation during the hydrological year and winter periods, but not to any other environ -mental factors tested. There was no significant difference in either marginal nitrate leaching or nitrate leaching between growing spring cereals versus winter cereals. No significant effect of winter vegetation cover on marginal nitrate leaching was detected. Furthermore, the marginal nitrate leaching at the optimal N rate in the first year of the experiments with increasing N rates was significantly lower than the accumulated effect of two or more years with low or high N rates. Therefore, the long-term effects of N rates should be accounted for when estimating marginal nitrate leaching. The marginal nitrate leaching is highly variable between years and only a part of this variation was explained in this study. The positive correlation between marginal nitrate leaching and nitrate leaching at optimum N rates implies that cropping system and management factors reducing nitrate leaching also reduce the marginal nitrate leaching.

In Denmark, eutrophication of coastal areas is one of the major challenges in meeting the requirements from the European Water Framework Directive (WFD). This eutrophication is mainly a cause of excessive nitrogen loads from agricultural production, which can be reduced by implementing abatement measures. The cost-effectiveness of a land use policy may vary depending on the abatement measures used and where they are implemented. Taking account of the spatial heterogeneity of costs and effects in the choice of abatement measures significantly reduces the cost of meeting WFD targets. Targeting of regulation to identify least cost options to reduce nitrogen has for many years focused on agricultural abatement measures. However, the marginal costs of reductions in agriculture have now increased to an extent, where it has become relevant also to reconsider other sectors. We set out to compare nitrogen abatement measures across agriculture and waste water treatment at a national level, acknowledging heterogeneity in marginal costs across spatial location as well as differences in reduction requirements across catchments. In the analysis, we combine data on costs and effects of abatement measures at a fine spatial scale to estimate marginal costs within the two sectors at a national level for Denmark. The model minimizes the costs of meeting the specific required nitrogen load reduction targets for all catchments and finds the composition and spatial location of the optimal abatement effort. We find that waste water treatment abatement measures are only relevant in two coastal catchments, where they mainly serve as a supplement due to insufficient potential for agricultural land in rotation to provide all the N load reductions. Furthermore, we find that the pressure on agricultural land to reduce nutrient loads is very high in some catchments, implying that abatement measures such as land retirement where abatement costs are high enters the optimal solution due to their high level of effectiveness. Recommendations and average costs vary across catchments, indicating that results are not easily translated to generic national-level policy design.

Crop failure detection using UAV images is helpful for precision agriculture, enabling the precision management of failure areas to reduce crop loss. For wheat failure area detection at the seedling stage using UAV images, the commonly used methods are not sufficiently accurate. Thus, herein, a new tool for precision wheat management at the seedling stage is designed. For this purpose, field experiments with two wheat cultivars and four nitrogen (N) treatments were conducted to create different scenarios for the failure area, and multispectral UAV images were acquired at the seedling growth stage. Based on the above data, a new failure detection method was designed by assimilating prior knowledge and a filter analysis strategy and compared with classical filter-based methods and Hough transform-based methods for wheat failure area detection. The results showed that the newly proposed assimilation method had a detection accuracy between 83.86% and 97.67% for different N levels and cultivars. In contrast, the filter-based methods and Hough transform-based methods had detection accuracies between 53.73% and 83.95% and between 20.71% and 75.79%, respectively. Thus, the assimilation method demonstrated the best failure detection performance.

NLES5 is the fifth version of an empirical model for predicting annual nitrate leaching from the root zone (1-meter depth), accounting for effects of nitrogen (N) inputs, crop sequences, autumn and winter crop cover, soil types, and weather conditions. It was developed and calibrated based on a comprehensive nitrate leaching dataset, primarily from Denmark. The model is used for quantifying annual nitrate leaching under Danish soil, weather, and field management practice. The model simulates the effects of N application rate, the presence of a cover crop, and the effects of crop management targeting measures to reduce agricultural nitrate leaching for the improvement of the quality of groundwater and surface water systems. The model takes into account crop and N management effects in the year of nitrate leaching and the two previous years, while long-term effects of N inputs are accounted for via total N in topsoil. The model provides estimates of nitrate leaching for the most important crops grown in Denmark and their management at cropping system level considering effects of soil and climate. The prediction of average annual nitrate leaching following the 13 main crop classes in the model varied from 25 to 170 kg N ha(-1)& nbsp;with a model performance (independent validation data (856 observations)) of Root Mean Square Error (RMSE) of 30.8 kg N ha(-1) and a coefficient of determination (R-2) value of 0.40. The RMSE obtained for the calibration data (2053 observations) was 29.6 kg N ha(-1) and the R-2 was 0.53.

Diffuse pollution of nitrate from agricultural fields is a critical environmental problem around the world. Sources and sinks of nitrate are heterogeneously distributed over various spatial scales, and the connectivity and transport pathways between them also change at different temporal scales. Therefore, understanding the impact of these variabilities in nitrate transport to the aquatic environment is fundamental for a correct numerical modelling of nitrate transport within a catchment. This study, hence, investigated controls on the spatiotemporal variability of nitrate in a glacial landscape and upscaled it to catchment scale by synthesizing geological, hydrogeochemical, and geophysical information. We found that different parts of the sedimentary succession define the locations of nitrate sinks in this catchment. Denitrification mainly may occur mainly around small patches of postglacial sediments on the outwash plain, which is the youngest formation covering the top layer of the catchment. In contrast, in older geological elements, which constitute the hill and the layers below the outwash plain, oxic, nitrate-containing groundwater was found, probably because of depletion of reduced compounds over the long exposure time. We also found that the boundary between these two formations may govern the seasonal shift of this oxic groundwater's connectivity to the stream consequently nitrate export from the catchment. This conceptual understanding of nitrate transport and sinks then was transformed into a 3D hydrogeochemistry model based on a high-resolution resistivity model of the catchment. We propose that such a basic understanding of how a catchment hydrogeochemically operates should be the first step toward setting up a catchment scale hydrological model with reactive N transport.

Excess nitrogen (N) losses from intensive agricultural production are a world-wide problem causing eutrophication in vulnerable aquatic ecosystems such as estuaries. Therefore, Denmark as one of the most intensively farmed countries in the world has enforced mandatory regulations on agricultural production since the late 1980s. We demonstrate the outcome of the regulations imposed on agriculture by analyzing decadal trends in nitrate (NO3-) concentrations and loads in streams using 29 years of detailed monitoring data and survey information on agricultural practices at field level from five intensively cultivated headwater catchments. The analysis includes the importance of four main drivers (climate, land use, agricultural practices, and biogeophysical properties of catchments), each divided into different factors that may influence stream NO3- loads during three subperiods defined by the time of introduction of different mitigation measures: i) 1990-1998, ii) 1999-2007, and iii) 2008-2018. Significant correlations with annual flow-weighted stream NO3- concentrations and/or loads were found for factors representing all of the four main drivers including precipitation, large scale climate fluctuations, runoff, previous year's runoff, baseflow index, number of annual frost days, agricultural area, livestock density, field N surplus, catch crop cover, manure storage capacity, method and time of manure spreading, and time of soil tillage. Changes in the four drivers were reflected by the load-runoff (L-Q) relationships for each of the three subperiods within each of the five headwater catchments. The five catchments experienced large but catchment-specific downward shifts in the L-Q relationship attributable to changes in land use and agricultural management within the catchments. The documented large downward shifts in NO3- loads demonstrated for the five catchments (30-52%) as a consequence of mandatory regulation over a period of nearly three decades are a unique example of how agriculture can reduce its environmental impact. (C) 2021 The Authors. Published by Elsevier B.V.

In this paper, we investigate the potential gains in cost-effectiveness from changing the spatial scale at which nutrient reduction targets are set for the Baltic Sea, with particular focus on nutrient loadings from agriculture. The costs of achieving loading reductions are compared across five levels of spatial scale, namely the entire Baltic Sea; the marine basin level; the country level; the watershed level; and the grid square level. A novel highly-disaggregated model, which represents decreases in agricultural profits, changes in root zone N concentrations and transport to the Baltic Sea is used. The model includes 14 Baltic Sea marine basins, 14 countries, 117 watersheds and 19,023 10-by-10 km grid squares. The main result which emerges is that there is a large variation in the total cost of the program depending on the spatial scale of targeting: for example, for a 40% reduction in loads, the costs of a Baltic Sea-wide target is nearly three times lower than targets set at the smallest level of spatial scale (grid square). These results have important implications for both domestic and international policy design for achieving water quality improvements where non-point pollution is a key stressor of water quality.