Assessing and Mapping the Environmental Impacts of Best Management Practices in Nitrate-Vulnerable Areas

Abstract

This investigation explores the impact of various management practices on nitrate leaching and crop yield in two specific areas in Slovenia. The issue of nitrate leaching from agricultural land is a worldwide threat to drinking water, and despite years of research efforts, universal solutions are still unknown. The two chosen study sites are significant because of their importance for agricultural production and because groundwater aquifers beneath are main sources of drinking water, which makes imposing mitigation measures challenging. One of the areas was reported to be of “bad” status according to the Water Framework Directive criteria, while the other is at risk of reaching this status if nitrate concentrations in groundwater continue to rise. This research used the SWAT model to simulate nitrate leaching and crop yield changes under different agricultural scenarios on different soil types. It aimed to accomplish two objectives: first, to identify parts of the case study areas where the existing combination of soil conditions and agricultural practices enables a high potential for nitrate leaching; second, to identify agricultural practices that decrease nitrate leaching from various soil types while maintaining crop yields in each area. By identifying the most vulnerable locations and the most promising practices, we generated a chart of best management practices for specific soil types as a guide that extension services can use to advise farmers on potential management improvements. The main findings demonstrate that reducing fertilizer application, both organic and inorganic, in areas where the rates exceed crop requirements may not have a significant impact on crop production. However, these reductions often resulted in a noticeable decrease in nitrate exports. The results also showed that soil type is crucially important and should always be considered when evaluating the effects of agricultural management on crop yields and nitrate leaching.

1. Introduction

Nitrogen pollution of groundwater in the form of nitrate leaching from agricultural land is a phenomenon that has been known for decades [1,2,3,4,5]. Numerous scientific studies and legislation have been written to mitigate the problem, but with limited success [6]. In many parts of the world, nitrate leaching resulting from agricultural activities is still a threat to the state of groundwater aquifers.

Many environmental and human factors, such as soil depth and permeability, precipitation patterns, fertilization, and other agricultural practices, influence nitrate leaching. As a result of all those combinations, it is practically impossible to propose a set of best management practices (BMPs) that would work universally. In addition, legislative approaches usually favor more generalized solutions, which is why there are still areas where mitigation efforts are unsuccessful [7,8].

Previous studies have discussed strategies for addressing nitrate leaching and found that a focused, locally specific approach is the best method [9,10,11,12]. However, since environmental and human factors differ around the globe, it is impossible to determine if what worked somewhere else might also work in our specific case. Thus, it is necessary to study the effects of these factors locally, hopefully allowing us to pin down the most environmentally vulnerable parts of the area and define a list of promising agricultural approaches to deal with excessive nitrate leaching in those vulnerable parts. At the same time, it is necessary to acknowledge the importance of agricultural production for human well-being regarding food and other produced materials. Thus, it is practically impossible to ban agriculture in such areas [13]. Instead, we need to try and find a balance, which is usually overlooked in similar studies. One way to evaluate agricultural productivity is through an economic analysis, but a more straightforward evaluation of change in crop yields can also allow us to obtain salient information on the negative effects those mitigation strategies might have on the productivity of the sector. Therefore, this study will test the hypotheses that to reduce nitrate leaching in a vulnerable area by some mitigation measures, it is unnecessary to reduce agricultural production equally.

In our specific case, the final aim of the study was to create an inventory of such combined effective measures for the two case study areas with nitrate-vulnerable groundwater and intensive agricultural production in Slovenia. The whole of Slovenia is designated as a nitrate-vulnerable zone by national law [14], which limits agricultural production universally across the country despite that only some areas, like alluvial plains and karst catchments, were determined to be very vulnerable based on state groundwater monitoring results [15]. Groundwater is the primary source of drinking water in the country. Therefore, extraction wells are further protected by water protection zones with prescribed mandatory limitations to human activities, including agricultural management practices [14]. However, vulnerable aquifers are mostly not protected by locally specific targeted measures [16]. Consequently, several groundwater bodies in Slovenia are either in a “bad” status or in danger of reaching it, according to state monitoring by EU Water Framework Directive (WFD) standards [15]. Previous studies in the areas found that, while agricultural activities are not the only source of nitrates in groundwater, targeted improvements in the agricultural management could mitigate the nitrate leaching to some extent [17,18]. They also emphasized the need for further investigations of locally tested solutions. This study is presented as a two-step process. First, the anticipated nitrate leaching hot spots were located in the case-study maps based on the business-as-usual simulations by the Soil and Water Assessment Tool (SWAT) model. Second, the environmental (nitrate-nitrogen leaching) and agricultural (crop yields) effects of different alternative scenarios were compared to select those effective at simultaneously reducing nitrate-nitrogen leaching and possibly sustaining crop yields at an acceptable rate. In the end, identified locally effective measures were allocated to the hot spots to model whole area outcomes.

2. Materials and Methods

2.1. Case Study Areas

The case-study (CS) areas (Figure 1) were chosen based on findings of previous studies [17,18], and for their environmental vulnerability and agricultural significance. Environmental vulnerability refers to how susceptible an area is to nitrate leaching because of specific environmental and human factors. Agricultural significance, on the other hand, refers to how favorable the environmental conditions of an area are for intensive agricultural production. Both CS areas are alluvial plains with a temperate climate and relatively shallow groundwater table (locally, even less than 10 m below the surface) and soil. Base rock is mostly sand and gravel above intergranular aquifers, which makes both areas naturally susceptible the nitrate leaching. Since the areas are mostly flat and well-drained, they are suitable for intensive production, resulting in significant agricultural land use. This makes them strong candidates for testing the proposed balanced approach, as they feature a vulnerable groundwater body and economically important agricultural production, both of which are important, and one cannot be easily favored over the other.

The CS Dravska kotlina (DK) is located in Northeast Slovenia and measures almost 450 km² of mostly flat ground. Because of favorable natural factors, more than half of the area is covered by agricultural land-use types: arable (38%), grassland (13%), or orchards (2%). Most of the area is between 200–350 m above sea level, with the hilliest part and more extensive agriculture rising to 1350 m. However, based on the nitrate monitoring in groundwater wells, the status of the groundwater body is classified as “bad,” according to WFD criteria [15].

The second CS Krška kotlina (KK), is located in the southeast of the country and measures 76 km². It is a flat plain with fertile soil and over half of the area is covered by agricultural land-use types: arable (44%), grassland (14%), or orchards (3%). The elevation varies from 235 m to 120 m above sea level. The current status of groundwater is not poor, but the trend in nitrate concentrations is rising, so there is a risk of reaching a “bad” status in the future [15].

In terms of water-protection measures, there are several drinking-water extraction wells in the studied CSs, each protected by several water-protection zones (Figure 2). These are areas intended for the protection of drinking-water sources where different restrictions to human activities, including agricultural production, are defined by state law based on groudwater travel times from the extraction wells’ locations. Three main zones are defined: wide, narrow, and narrowest. However, only 2.6% of the area in the DK and 1% in the KK case study are protected by the strictest “narrowest water protection zone,” which is the only one that enforces high-level limitations for agricultural production. Detailed information on limitations is detailed in Section 2.3.

2.2. Overview of the Research Procedures

In order to evaluate the impact of different soil types/agricultural management combinations on the environment and agricultural production, the SWAT 2012 model [19] was used. In Step 1, the anticipated nitrate–nitrogen leaching hot spots were located on the case-study maps according to the base-scenario results, which are intended to simulate the current common agricultural practices in both areas (business-as-usual situation). In Step 2, both the environmental (nitrate–nitrogen leaching) and agricultural (crop yields) effects of different alternative agricultural scenarios (promising measures) were compared to select those that are effective at simultaneously reducing nitrate–nitrogen leaching and possibly sustaining the crop yields at an acceptable rate. Ultimately, in Step 3, these locally effective measures were allocated to the hot spots to reach a balance between environmental impact and agricultural productivity.

2.3. Scenarios of Agricultural Production

Agricultural production in both areas is highly intensive by Slovenian standards, mainly because of the flat terrain and fertile soils, which allow for more efficient management with larger plots and higher yields. Arable land is the prevalent agricultural land use, with corn, winter wheat, and barley being the most common crops grown. Crop rotations are very dependent on the leading production enterprise on each farm. Cattle-focused farms grow lots of corn for silage, while pig-focused ones depend more on barley and corn for grain. Smaller family farms often have mixed crop rotations, but since the conditions for large-scale production in both CS areas are favorable, there are also farms with no animals that grow only arable crops for sale. It is, therefore, challenging to locate nitrate–nitrogen-leaching hot spots in each CS area without knowing which arable fields have a specific type of crop rotation. The Slovenian land-use map is very exact in showing actual land use based on aerial imagery. Unlike the Corine Land Cover database [20], which aggregates land uses into broader classes, the Slovenian map spatially delineates singular houses, trees, arable parcels, roadways, etc., based on several different classes, thus representing the land use very well. Still, with 50% of the land in each area being designated as arable, it is not possible to determine what the agricultural management of each of many small field parcels look like, and the conventional modeling approach would be to assign the same management scenario to all units with arable land use, which is a big simplification.

To reduce this uncertainty and make the hot-spot mapping more reliable, a database of farms in the area was acquired from the Slovenian Agency for Agricultural Markets and Rural Development. The database included the number of animals from each category in a particular farm and the area of parcels they manage. By determining what the main animal species on the farm are, we could assign a more appropriate crop rotation to each of the arable parcels in the area. For example, if over 90% of the animals on a particular farm were cattle, the crop rotation assigned was cattle-focused (more corn for silage); if over 90% of the animals were pigs, the crop rotation assigned was pig-focused (more corn for grain); and if there were no animals, arable crop production was assumed. The distribution of farm-type-based crop rotations can be seen in Figure 3, where each type is marked with a different color, and later assigned a different crop rotation according to the main enterprise on the farm. Without this detailed approach, all the colorful parts of the map would be one single land-use type: arable, and only a single generic land management could be assigned.

The presented detailed land-use map is crucial for modeling the base scenario (business-as-usual agricultural management), with which we wanted to spatially represent where the critical nitrate–nitrogen leaching hot spots are located in each CS area. With the help of local agricultural advisory services, business-as-usual agricultural management scenarios were prepared based on typical crop rotations and management practices in the area. Business-as-usual crop rotations are presented in more detail in

Table 1. Crop rotations used for modeling the base scenarios.

Main Farm Enterprise	Year in Rotation	Plant	Fertilizer Added (kg N/ha)
			Organic Form	Mineral Form
Cattle	1	Red clover cover crop	-	-
	1	Corn silage	75	144
	1–2	Winter wheat	-	150
	2–3	Grass cover crop	-	80
	3	Corn silage	75	144
	3–4	Red clover cover crop	-	-
	4	Corn silage	75	144
	4–5	Winter barley	-	135
	5–1	Red clover cover crop	-	-
Pigs	1	Red clover cover crop	-	-
	1	Corn for grain	75	90
	1–2	Winter barley	-	150
	3	Corn for grain	75	90
	3–1	Red clover cover crop	-	30
Crops	1	Grass cover crop	-	-
	1	Corn for grain	75	75
	1–2	Winter wheat	-	150
	2–3	Oilseed rape (Canola)	-	183

.

Besides the base scenario, several alternative scenarios were prepared to simulate alternative practices’ impact on nitrate–nitrogen leaching and crop yields. Alternative scenarios were formed by taking the base scenario for each of the four base rotations (cattle, pigs, crops) and changing the crops, fertilizer types, or quantities according to the objectives. This was done by altering the agricultural management (.mgt) files in the model database to feature the appropriate crop or fertilizer type and quantity. Other parameters were not changed. For example, with fertilizer-focused alternatives, we examined how the change from organic to mineral fertilizers, or a decrease in the total amount of fertilizer, impacts nitrate–nitrogen leaching and crop yield. Similarly, we wanted to evaluate the impact of the change from corn-based to soybean-based (nitrogen-fixing plant, no N fertilization needed) cattle and pig rotations.

Finally, with alternative scenarios, where legally binding water-protection restrictions were considered (WPZ), the objective was to present the impact of different water-protection restrictions on nitrate–nitrogen leaching and crop yield. The narrowest zone protects the land in close proximity to the extraction wells. Therefore, the use of liquid manure is prohibited, but there are also some other limits to the use of fertilizers. The narrow and wide zones only restrict the single-dose amounts of fertilizers to encourage several smaller doses to be applied. Grassland scenarios are based mainly on the best fertilization practices manual for Slovenia [21], except for the “Grassland 4-cut intensive” and “Grassland WPZ” scenarios, which are based on intensive management practices used by some of the larger producers in both areas to determine if grassland is an effective use of land, even with high-intensive management. All scenarios are listed in

Table 2. Agricultural management scenarios used for SWAT model simulations.

Abbreviation	Scenario	Description
CTL-base	Cattle base	Business-as-usual crop rotation
CTL-90	Cattle 90% fert.	90% of base fertilizer amount used
CTL-80	Cattle 80% fert.	80% of base fertilizer amount used
CTL-70	Cattle 70% fert.	70% of base fertilizer amount used
CTL-min	Cattle mineral fert.	Mineral instead of organic fertilizers
CTL-soy	Cattle Soybeans	Soybeans replace corn in crop rotation
CTL-broad	Cattle WPZ-broad	Wide-water protection-zone restrictions
CTL-strict	Cattle WPZ-strict	Strict-water protection-zone restrictions
CRP-base	Crops base	Business-as-usual crop rotation
CRP-90	Crops 90% fert.	90% of base fertilizer amount used
CRP-80	Crops 80% fert.	80% of base fertilizer amount used
CRP-70	Crops 70% fert.	70% of base fertilizer amount used
CRP-broad	Crops WPZ-broad	Wide-water protection-zone restrictions
CRP-strict	Crops WPZ-strict	Narrowest-water protection-zone restrictions
PGS-base	Pigs base	Business-as-usual crop rotation
PGS-90	Pigs 90% fert.	90% of base fertilizer amount used
PGS-80	Pigs 80% fert.	80% of base fertilizer amount used
PGS-70	Pigs 70% fert.	70% of base fertilizer amount used
PGS-min	Pigs mineral fert.	Mineral instead of organic fertilizers
PGS-soy	Pigs Soybeans	Soybeans replace corn in crop rotation
PGS-broad	Pigs WPZ-broad	Wide-water protection-zone restrictions
PGS-strict	Pigs WPZ-strict	Narrowest-water protection-zone restrictions
GRS-4C-int	Grassland 4-cut intensive	Intensive practice by some larger farms
GRS-broad	Grassland WPZ-broad	Wide-water protection-zone restrictions
GRS-strict	Grassland WPZ-strict	Narrowest-water protection-zone restrictions
GRS-4C-org	Grassland 4-cut organic fert.	Grass is cut 4 times, organic fert. are used
GRS-4C-min	Grassland 4-cut mineral fert.	Grass is cut 4 times, mineral fert. are used
GRS-3C-org	Grassland 3-cut organic fert.	Grass is cut 3 times, organic fert. are used
GRS-3C-min	Grassland 3-cut mineral fert.	Grass is cut 3 times, mineral fert. are used
GRS-2C-org	Grassland 2-cut organic fert.	Grass is cut 2 times, organic fert. are used
GRS-2C-min	Grassland 2-cut mineral fert.	Grass is cut 2 times, mineral fert. are used
GRS-1C-none	Grassland 1-cut no fert.	Grass is cut only once, no fert. are used

.

2.4. SWAT Model Input Data

Regarding the simulations of environmental impact of different agricultural practices, the Soil and Water Assessment Tool (SWAT) is one of the most commonly used models for simulating impacts of agricultural management on hydrology and water quality in the literature [22]. It is a physical, spatially, semi-distributed hydrological model [23]. Physical structure means that the model’s structure represents the physical characteristics of the modeled system, which requires a large amount of specific input data. On the other hand, the semi-distribution of spatial units greatly reduces the computational time. The core computational units in the SWAT model are called hydrologic response units (HRUs) and are not based on a grid but are lumped units of areas with the same land use, soil type, and slope, meaning the final number of units is significantly reduced. Other than that, its strengths include very effective plant growth and agricultural management representations (based on EPIC model) and a large database of extant studies. This makes the model verification easier and more reliable, as well as the fact that the model is open-sourced and very well-documented [22,24].

The SWAT model requires many input data, ranging from weather, land use and agricultural management, soil properties, slope, and hydrology. A summary of input data is presented in

Table 3. SWAT model input data with source, units, and resolution.

SWAT	Units and Resolution	Data Source *
Weather (at least 10-year period)
Precipitation	mm/day	ARSO
Min and max temperature	°C (daily min & max)	ARSO
Solar radiation	MJ/m2/day	ARSO, calculation
Relative humidity	(daily average)	ARSO
Wind speed	m/s (daily average)	ARSO
Soil
Soil map	.shp (250 m)	CPVO
Depth and number of horizons	mm	CPVO
Rooting depth	mm	CPVO
Bulk density	g/100 mL	Measurements, calculation
Available water capacity	mm H2O/mm soil	Measurements, calculation
Soil hydraulic conductivity	mm/h	Measurements, calculation
Soil albedo		Calculation
Organic carbon	%	Measurements
Texture (sand/silt/clay fractions)	%	Measurements
Soil erosivity (KUSLE)		Calculation—Williams
Slope and land use
Digital elevation model	raster (25 m)	GURS
Land-use map	.shp (25 m)	MKGP
Agricultural-management practices	common in the areas	Farm advisors

Note: * ARSO: [25]; CPVO, MKGP: [26]; GURS: [27].

, while a more detailed explanation of the data preparation and model setup is demonstrated in the following paragraphs.

The model extent for each area was not the typical watershed. Since the CS areas are parts of Slovenia’s two largest river basins, modeling a catchment almost the size of the whole country to get results for two relatively small areas was not considered meaningful. Instead, the model extent was determined based on the groundwater body extent for each CS. In the case of DK CS, the Polskava watershed with reliable discharge data is located mostly inside of the broader groundwater body extent. Therefore, the modeled area was extended to also include the upstream part of this smaller catchment and allow for a conventional discharge calibration. Inclusion of the area outside the Polskava catchment introduces a degree of uncertainty, but the area in question has very similar environmental characteristics to the downstream part of the Polskava catchment. Therefore, we assumed the calibrated values were meaningful for both extents. Since there were no options for including a watershed with reliable data for the KK case, we performed a soil–water-based calibration, where water balance in the model was calibrated by comparing measured soil water content in several field trials with the model equivalent: the output variable SW_END. This procedure was previously described in more detail in [28].

Weather data were acquired from local meteorological stations for the period of 10 years. They were mostly useable in their original form, except for the solar radiation data, which was calculated from the duration of sunshine with Angstrom’s equation:

$${\mathrm{R}}_{\mathrm{s}}={\mathrm{R}}_{\mathrm{a}}\left({\mathrm{a}}_{\mathrm{s}}+{\mathrm{b}}_{\mathrm{s}}\frac{\mathrm{s}}{\mathrm{S}}\right)$$

where: R_a—extraterrestrial radiation (16.83 MJ/m² day); S—max possible number of sunshine hours (11.98 h); s—the actual number of sunshine duration (h); a_s and b_s—Angstrom’s regression constants (a_s = 0.25; b_s = 0.5).

A 10-year simulation period was used because it was long enough to capture major weather events in each area, but short enough to not significantly increase computational times needed for the simulations of different agricultural scenarios.

Land-use and soil-type maps are presented in Figure 4. Soil hydraulic properties were partially measured in the field (for soil–water calibration in KK CS), but, for the most part, they were calculated from the texture with the use of pedo-transfer functions [29]. It is necessary to note that the soil-type map is an important source of uncertainty as its resolution is not as strong as with other spatial datasets. Still, it is the best-known representation of soil types in the areas and was done through extensive survey work over the previous decades. To minimize the impact of this uncertainty, the results were never interpreted spatially specific (i.e., nitrate–nitrogen leaching in this specific field is such and such), but rather environmentally specific (i.e., nitrate–nitrogen leaching in areas with expected land use and soil properties is such and such).

2.5. Model Uncertainty Analysis, Calibration, and Validation

Models were established using the ArcSWAT2012 interface for the periods of 2000–2012 (DK CS) and 2006–2018 (KK CS) in daily time-step. Uncertainty analysis, calibration, and validation of the models were performed in the SWAT–CUP 2012 program [30,31].

Uncertainty analysis is an important step that clarifies which of the input parameters have the most influence on the results. This allows for more effective calibration and updates the modeler with important information about the possible sources of uncertainty. For example, if a small change in a parameter is reflected in a large difference in the simulation results, this means the parameter is very uncertain. Such parameters in our case were SOL_BD, HRU_SLP, and GW_DELAY for DK, as well as SOL_AWC for the KK model area.

As discussed above, two different approaches to calibration were underwent in both CS areas. In DK, a smaller nearby watershed with reliable discharge data was included in the model to calibrate it for discharge for the full years from 2008 to 2011. Validation was done for full years from 2006 to 2007. Nitrate calibration was also performed because total nitrogen data were available for the same sampling site on the river, but only for quarterly periods for the years 2008, 2009, and 2010. Since there were no smaller watersheds in the KK–CS area, the soil–water-based calibration and validation were performed in a way previously described in [28].

2.6. Environmental and Agricultural Effects of Different Agricultural Scenarios

Only the prevalent soil types and land-use classes were selected to analyze environmental and agronomic effectiveness scenarios. Many of the soil types in both areas are marginal (almost three-quarters of them cover less than 3% of the area) and not all are prevalent in agricultural land-use types. In order to optimize the research efforts, only the soil types that cover over 3% of the whole CS area, and are found under agricultural land-use types, were selected. Similarly, only agricultural land-use classes (with emphasis on arable fields and grassland) were included in the analysis. The slope factor was not included in the analysis of the results because we were primarily interested in flat parts of the CS areas. In the case of different circumstances, we believe this methodology would still be valid with only minor adjustments (i.e., including the slope factor and, thus, having more combinations).

Based on the above criteria, five important soil types were selected in DK and three in KK–CS areas. Every agricultural management scenario was simulated on each of these soil types. All these combinations, in the end, provided us information on nitrate–nitrogen leaching and crop-yield differences for the final environmental and agricultural effectiveness analysis.

The effectiveness of scenarios was quantified by extracting the yearly average values for the SWAT model output variables NO3L (nitrate–nitrogen leaching—kg N/ha) and YLD (crop yield—tonnes of dry mass/ha). The first was used to evaluate the environmental, and the second was used to evaluate the agricultural effectiveness of scenarios. We compared the results for each alternative scenario with its business-as-usual counterpart to calculate a percentage difference between the two. Depending on the magnitude of difference, each scenario-soil-type combination was labeled either effective, very effective, without effect, counter-effective, or very counter-effective. In order to deal with model uncertainty and to avoid misinterpretation of the model results, these categories were attributed based on the magnitude of each variable. This approach, where rather large ranges were considered, was adapted from previous research [32]. Since the effectiveness results of the different mitigation measures usually vary, authors proposed to use such ranges to minimize the uncertainty and ensure that, even if the results carry a degree of error, this method prevents the error to influence final effectiveness classification as much as possible. Consequently, all differences from −19 to +19% were labeled without effect to prevent measures with minor differences from being labeled effective. Effective scenarios showed either an increase in yield or a decrease in nitrate–nitrogen leaching from 20 to 49% of the difference compared to the base scenario. Very effective measures showed a difference of over 50%, while the counter-effective measures showed a difference of 20–49% and very counter-effective ones a difference of over 50%, respectively.

Based on these two indicators, we were later able to propose the most suitable scenarios (mitigation measures) for a particular combination of soil type and type of farming activities.

3. Results and Discussion

3.1. Calibrating and Validating the SWAT Models for Both CS Areas

Calibration and validation results for Dravska kotlina (DK) CS are presented in Figure 5.

A comparison of the observed and simulated discharge rates for the calibration (PBIAS = 2.8, R2 = 0.55, NSE = 0.52) and validation (PBIAS = 7.9, R2 = 0.55, NSE = 0.4) periods shows that the model performance is sufficient based on the guidelines in the literature [33,34]. Total nitrogen calibration results were also acceptable but barely within limits, (PBIAS = 44.1, R2 = 0.54, NSE = 0.39), mostly because of the lack of total nitrogen data (only quarterly measurements throughout a couple of years). The lack of data did not enable us to perform the validation step for total nitrogen. Crop yields were not calibrated, but only validated (against the average yearly reported yield values for each CS region [35]). Comparisons of measured and simulated crop yields for some of the important crops are presented in

Table 4. Comparison of measured and simulated crop yields for some of the important crops for DK CS.

Crop	Dry Matter Content (%) *	Average Yearly Yield for the Year 2010 (t/ha)
		SWAT Model Simulation	Crop Yield Statistical Data for the Podravska Region
Winter barley	88	5.4	4.6
Permanent grassland	20	2.3	6.8
Silage corn	35	51.4	45
Apples	15	15.6	15

Notes: * The SWAT model outputs yields in T/dry matter per ha, therefore, for easier comparison; the simulated values were calculated based on dry matter contents in the second column.

.

The Krška kotlina (KK)–CS calibration and validation results also show that the model performs satisfactorily. The results were published in a previous study [28].

3.2. Identifying the Anticipated Locations of the Nitrate–Nitrogen Leaching Hot Spots Based on the Business-as-Usual Scenario

After the sound performance of both CS area models was confirmed, the base scenario’s model results were analyzed to obtain information on average yearly nitrate–nitrogen leaching intensities in each CS area. Average yearly nitrate–nitrogen leaching values were extracted for each HRU and traced back to each area’s spatial HRU map to identify areas with higher nitrate–nitrogen leaching values. The effectiveness of measures tested in this study depends strongly on the agricultural management (fertilization) and soil type and not nearly as much on weather conditions, as these were intentionally made uniform across each CS area via the model input data. This ensured that precipitation patterns and localized storm events did not influence the results and potentially alter the effectiveness of different measures. The crop rotations (fertilization) were different between arable fields of the different farm types. This way, a relatively accurate spatial representation of crop rotations was achieved, certainly more accurate than the approaches usually followed in SWAT model application, where all arable fields feature one crop rotation, with rare exceptions [36,37]. Despite the effort to improve the crop-rotation representation, the low resolution of the soil map remains an important source of uncertainty.

Although the soil map was prepared by experts based on extensive field sampling and surveys, its current resolution does not allow for a field-scale representation of nitrate–nitrogen leaching. The soils in nature are very heterogeneous, but the soil-type map used for modeling assumes the whole area under a specific soil type has uniform characteristics, simplifying reality. In modeling, it is impossible to avoid such simplifications, commonly known as uncertainties of the model, but we can be cautious of them and adequately interpret data to avoid misinterpretation. Without field observation, our results only simulate the nitrate–nitrogen leaching magnitude in a specific HRU, not the actual field parcel in that geolocation. The issue of soil-map resolutions in SWAT modeling was previously discussed in greater detail [38]. The authors of that paper concluded that models with different input data resolutions might all perform satisfactorily for streamflow, but a representation of more complex water-quality processes might be questionable if the resolutions are too coarse. In our case, we cannot say that nitrate–nitrogen leaching reduction in a specific field parcel will be of some magnitude if we change the management in a certain way, but we can assume that in soil with specific properties, reduction in nitrate–nitrogen leaching can be expected to be of some other magnitude, as a result of the particular management change. Maps can be helpful in the overall case study overview as they show us the general areas (hot spots) of possible vulnerabilities or improvements. For this purpose, we did not further specify the geolocation of these hot-spot areas. Further research should be performed, possibly by field work to determine the texture of each field parcel in the area, as texture is an important indicator of the hydraulic conductivity of soils. Maps with simulated average yearly nitrate–nitrogen leaching intensities are presented in Figure 6.

The spatial map of nitrate–nitrogen leaching shows that soil-type and agricultural-management combinations of each HRU are very influential on the nitrate–nitrogen leaching intensity. This is expected based on the results of other similar studies and wider scientific knowledge [39]. The soil type explains the differences in soil properties, especially the hydraulic ones, which determine the way water with soluble nutrients moves in the soil profile. Even in fields with uniform management, differences in plant growth (crop yields) and nutrient leaching can be observed because of this variability in soil properties [40]. Shallow, sandy soils are unable to store water for long, usually allowing large quantities to quickly leach through the profile and beyond the reach of plant roots, taking water soluble nutrients with it. Deeper soils with higher clay contents, on the other hand, can hold the water for longer, decreasing the risk of nitrate leaching. In nature, soil hydraulic properties can vary even in very small land parcels, while in modeling, depending on soil-map resolution, areas with uniform properties can be quite large. Agricultural management varies a lot as well, and the different methods of soil cultivation or different plant species further influence the soil hydraulic properties. In addition fertilization adds additional amounts of nutrients into the system to possibly leach through the soil profile [41]. These factors make in-situ research challenging because it is difficult to control all the different variables. In modeling, the simplified resolutions of input data allow for easier control of variables and, therefore, easier evaluation of results, but such simplifications obstruct us from resembling nature’s complexity.

In DK CS, the North–South axis in the center of the area with higher leaching intensities corresponds with the shallow Dystric Cambisol soil type. Agricultural land use is also prevalent on the right bank of the Polskava river south of the hot spot area, where the nitrate–nitrogen leaching intensity, despite similar management, seems to be lower due to a less vulnerable soil type, mainly the Gleysols. In KK CS, the pattern is less visible, but the more vulnerable areas also correspond to the area of the Eutric Cambisol soil type. Nitrate–nitrogen leaching intensities, in some cases, exceed the EU Drinking Water Directive threshold of 50 mg NO3/L [42], though this is not the concentration of nitrate–nitrogen in groundwater but merely the concentration in soil water that is yet to percolate towards groundwater. Because of the dilution and possible denitrification processes, it is impossible to predict groundwater concentrations only by using the SWAT model. We were solely interested in knowing the intensity of nitrate–nitrogen leaching under different agri-environmental conditions.

3.3. Evaluating the Environmental and Agricultural Effectiveness of Agricultural Scenarios

Results presented in

Table 5. Comparison of average yearly nitrate–nitrogen leaching and crop-yield results for the base and alternative scenarios as percent change for the Dravska kotlina CS. Effective and very effective scenarios are highlighted in green and bold green for environmental, and blue and bold blue for agricultural effectiveness, respectively.

Scenarios		Percent Difference (10-Year Average Values)
		Nitrate–Nitrogen Leaching					Crop Yields
		Dystric Cambisol Deep	Dystric Cambisol Med. Depth	Dystric Cambisol Shallow	Eutric Gleysol	Dystric Gleysol
CTL-	Base	0.0%	0.0%	0.0%	0.0%	0.0%	0.0%
	70	−35.1%	−33.6%	−29.1%	−49.2%	−51.9%	−5.1%
	80	−21.8%	−22.2%	−21.0%	−74.8%	−70.4%	−2.5%
	90	−30.1%	−24.9%	−18.7%	−59.3%	−57.1%	−1.4%
	Min	20.3%	8.0%	7.5%	33.9%	47.2%	0.3%
	Soy	−66.6%	−59.9%	−70.0%	−46.5%	−60.0%	0.4%
	Strict	−35.3%	−33.2%	−30.7%	−47.4%	−53.1%	−4.8%
	Broad	−46.6%	−38.9%	−33.0%	−81.5%	−80.1%	−3.6%
CRP-	Base	0.0%	0.0%	0.0%	0.0%	0.0%	0.0%
	70	−6.6%	−12.6%	−27.1%	−49.3%	−43.9%	−21.8%
	80	−31.2%	−27.7%	−28.2%	−28.5%	−32.9%	−12.6%
	90	27.0%	18.6%	4.8%	60.9%	61.1%	−6.8%
	Strict	−21.4%	−21.0%	−25.7%	−20.0%	−26.0%	−7.7%
	Broad	0.0%	0.0%	0.0%	0.0%	0.0%	0.0%
PGS-	Base	0.0%	0.0%	0.0%	0.0%	0.0%	0.0%
	70	−50.3%	−40.3%	−37.5%	−51.6%	−52.6%	−6.4%
	80	−28.2%	−19.8%	−22.0%	−21.4%	−23.0%	−3.1%
	90	−33.4%	−24.3%	−17.8%	−38.4%	−42.2%	−0.7%
	Min	−22.8%	−19.5%	−20.7%	−48.5%	−41.3%	0.4%
	Soy	−35.1%	−32.6%	−48.5%	−32.9%	−36.8%	0.4%
	Strict	−67.4%	−55.8%	−54.6%	−69.8%	−70.9%	−6.2%
	Broad	−9.3%	0.0%	−0.8%	0.1%	−7.9%	−4.6%
GRS-	4C-org	0.0%	0.0%	0.0%	0.0%	0.0%	0.0%
	4C-min	2.9%	15.5%	20.1%	15.0%	4.3%	−9.3%
	3C-org	−41.9%	−29.0%	−30.7%	−29.1%	−36.9%	−26.7%
	3C-min	−69.4%	−56.0%	−44.1%	−67.6%	−68.4%	−36.1%
	2C-org	−84.5%	−73.8%	−81.7%	−76.3%	−72.7%	−66.7%
	2C-min	−62.7%	−51.6%	−60.0%	−48.8%	−40.6%	−78.7%
	1C	−83.0%	−62.9%	−86.0%	−50.3%	−32.8%	−100.0%
	Broad	531.1%	321.1%	177.8%	260.1%	218.0%	38.5%
	Strict	7.0%	5.4%	11.5%	4.1%	5.1%	−25.5%
	4C-int	845.3%	571.0%	314.4%	1219.1%	950.5%	80.0%

and

Table 6. Comparison of average yearly nitrate–nitrogen leaching and crop-yield results for the base and alternative scenarios as a percent change for the Krška kotlina CS. Effective and very effective scenarios are highlighted in green and bold green for environmental, and blue and bold blue for agricultural effectiveness, respectively.

Scenarios		Percent Difference (10-Year Average Values)
		Nitrate–Nitrogen Leaching			Crop Yields
		Fluvisol Deep	Fluvisol Med. Depth	Eutric Cambisol
CTL-	Base	0.0%	0.0%	0.0%	0.0%
	70	−37.1%	−36.1%	−8.7%	−5.1%
	80	−48.9%	−51.9%	−49.9%	−2.5%
	90	−3.1%	−7.5%	19.8%	−1.4%
	Min	−28.2%	−2.9%	−10.0%	0.3%
	Soy	−51.3%	−18.9%	−72.6%	0.4%
	Strict	−68.2%	−61.2%	−63.0%	−4.8%
	Broad	−35.1%	−40.7%	−19.9%	−3.6%
CRP-	Base	0.0%	0.0%	0.0%	0.0%
	70	−24.9%	−26.7%	−54.4%	−21.8%
	80	−54.3%	−45.6%	−67.3%	−12.6%
	90	−8.7%	−11.3%	−36.6%	−6.8%
	Strict	−51.0%	−47.9%	−53.8%	−7.7%
	Broad	0.0%	0.0%	0.0%	0.0%
PGS-	Base	0.0%	0.0%	0.0%	0.0%
	70	−56.5%	−49.9%	−56.0%	−6.4%
	80	−32.9%	−28.6%	−33.4%	−3.1%
	90	−46.2%	−36.4%	−39.1%	−0.7%
	Min	−42.2%	−38.7%	−40.2%	0.4%
	Soy	−29.4%	−22.9%	−29.7%	0.4%
	Strict	−72.1%	−70.7%	−72.8%	−6.2%
	Broad	0.0%	0.0%	0.0%	−4.6%
GRS-	4C-org	0.0%	0.0%	0.0%	0.0%
	4C-min	63.6%	46.6%	305.2%	−9.3%
	3C-org	3.7%	−1.1%	−10.2%	−26.7%
	3C-min	15.7%	3.7%	74.7%	−36.1%
	2C-org	4.7%	−3.8%	−49.9%	−66.7%
	2C-min	−54.9%	−51.5%	−62.9%	−78.7%
	1C	132.3%	124.6%	−49.5%	−100.0%
	Broad	345.1%	367.8%	860.3%	38.5%
	Strict	24.6%	11.0%	102.0%	−25.5%
	4C-int	793.1%	679.5%	1262.5%	80.0%

show that, compared to base scenarios, at least one environmentally effective scenario is available for each farm type. Specific alternative scenarios decreased nitrate–nitrogen leaching by over 80% compared to business-as-usual ones. It is important to note that scenario effectiveness differed between the CS areas and different soil types. CRP–70 scenario in DK CS is a prime example of this, and its significance is particularly evident as the environmental outcome in deep Dystric Cambisol had no effect, while in Eutric Gleysol, it was in the upper boundary of the effective interval. In the GRS-2C-org scenario in KK CS, there was no effect on deep Fluvisol, while it was also in the upper boundary of the effective interval on Eutric Cambisol. This is well-aligned with past publications, where researchers determined that one-size-fits-all solutions do not work in vulnerable areas and that a focused, localized approach is necessary in order to successfully mitigate nitrate–nitrogen leaching in a specific area [9,11,43]. The results of this study further strengthen claims that a localized approach is a method to be used in further research.

A nitrate–nitrogen leaching increase (counter-effective scenarios) was observed in many grassland scenarios, particularly in KK CS. Only one of them (GRS-2C-min) effectively decreased the leaching, while many others fell into the counter-effective or even very counter-effective interval. This effect is less pronounced in DK CS, but the intensive practice scenarios (GRS-broad, GRS-strict, and GRS-4C-int) resulted in a large increase of nitrate–nitrogen leaching, similar to KK CS. It is essential to state that the seemingly large increases in nitrate–nitrogen leaching (up to 1200%) in absolute terms are not as significant as it seems. They are a result of really low business-as-usual values, where a nitrate–nitrogen concentration increase of 10 mg/L already meant a relative difference of 200%. According to simulation results, average nitrate–nitrogen leaching for the intensive scenarios does not exceed an average arable rotation leaching intensity, so grassland management should not be considered worse than that of the arable fields. In addition, the intensive practice scenarios were the most problematic, while the others, based on effective agricultural practice guidelines, were not. Considering this, it is important to note that, as much as grassland land use is usually considered an excellent option for nitrate-vulnerable areas [3,44,45] when it is not managed following effective agricultural practice, the nitrate footprint is not any better than that of the typical arable land use.

Regarding effectiveness, it would be hard to determine how much a certain measure evaluated in this study would change the nitrate–nitrogen leaching in the field. However, since a large margin of 20% was considered, we can be confident that the measures labeled effective, and especially very effective (50% margin), would help mitigate leaching in the field with similar environmental and agricultural management conditions. If nothing else, the model results could help weed out the worst scenarios and only pick the more promising ones to test further in a field trial. Nonetheless, this efficiency scale was found unsuitable for evaluating agricultural effectiveness due to the scenarios not being designed to increase crop yields. Consequently, crop yields for most scenarios did not increase but instead decreased. When we compare the magnitude of nitrate–nitrogen leaching and crop-yield decrease, we can observe that while crop yields sometimes decreased slightly, while the nitrate–nitrogen leaching decreased significantly. The GRS and CRP scenarios exhibited the most prominent crop-yield decrease, while the CTL and PGS scenarios (soy, –min, and −90) experienced a crop-yield decrease of less than 2%, with –min and –soy even indicating a slight increase in crop yield. For the CRP group, the base scenario appeared to be well-optimized, as all alternatives, except –broad, which had a crop rotation nearly identical to the –base and showed a crop-yield decrease of over 5%. This would still fall inside the “no effect” class in our initial classification, so all of these could still be considered sustainable by the standards of environmental effectiveness. Based on this, we assumed that the crop yield-decrease is minimal, and such scenarios were deemed suitable for further evaluation. When multiple environmental scenarios were similarly effective, the one with the smallest crop-yield decrease was selected for spatial optimization, allowing for the best alternatives to be chosen for each modeled soil type.

Agricultural productivity is an important aspect of this study, as food production is fundamental to society’s functions. Even the European Union Farm-to-Fork Strategy [46], while addressing environmental issues, calls for improvement of the economic state of the agriculture sector. Crop-yield reductions were evaluated as a metric of measures’ impact on productivity, and the results helped us select only those environmentally effective measures that proved to sustain (as much as possible) the agricultural productivity.

3.4. Allocating the Most Effective Agricultural Scenarios to the Nitrate–Nitrogen Leaching Hot Spots with the Goal of a Win-Win Solution for the Environment and Agriculture

The last step in the process was to allocate the promising agricultural management scenarios to each of the HRUs so that nitrate–nitrogen leaching was decreased while crop yields remained stable as much as possible. The results for all scenarios are plotted on a graph where combined effectiveness can be evaluated (Figure 7 and Figure 8). Each dot is a representation of results for a combination of agricultural management and soil type.

The most optimal management scenarios are located In the top-right quadrant of each graph, with environmentally effective scenarios located in the bottom left, and agriculturally effective scenarios in the top right. Conversely, those in the bottom right are less effective than the base scenario in both aspects. Scenarios with the highest reduction in nitrate–nitrogen leaching and lowest reduction in crop yield were allocated to each CS area according to the soil and farm types. Since the whole of Slovenia is designated as a nitrate-vulnerable zone, according to the national law [14], the allocation of effective measures was done universally on agricultural land across entire CS areas. In order to carry out this task, the base management scenario for each HRU was replaced with the most effective one in its farm-type category. This way, the total CS area management was optimized, and nitrate–nitrogen leaching intensity decreased everywhere. The resulting map (Figure 9) represents the potential reduction in nitrate–nitrogen leaching intensity in each CS area.

The presented maps of CS display the distribution of possible improvements in nitrate–nitrogen leaching based on the selected effective measures. Results demonstrate that the improvements are promising, as in most areas, to the magnitude of over 20, or even over 50%, which is in line with findings of many past publications [47,48,49]. Most areas with seemingly no effect are located where no optimization occurred (outside the selected soil types or not an agricultural area). Considering this, we can state that the presented approach would also most certainly affect the nitrate–nitrogen concentration in groundwater, most likely decreasing it. This is an important finding, as it shows that the current extent of water-protection zones is not the most appropriate, as other research from Slovenia has revealed in the past [50]. Instead of imposing strict restrictions in a small area around each pumping well, it would be more prudent to impose targeted measures based on soil type and management to a wider extent of the groundwater body, especially considering that the WFD calls for all water bodies to reach “good” status, not just the parts around extraction wells [51]. In order to translate the findings of this study into practice, the uncertainty of soil-map resolution could be circumvented by testing the arable fields for soil texture and, based on pedo-transfer functions, determine a rough estimate of hydraulic conductivity. Based on this, the farmers could choose from the management practices that fared well on the more or less permeable soil types without the need to know all the details.

Finally, based on the model results showing large reductions in nitrate–nitrogen leaching and relatively well-sustained crop yields for different alternative scenarios, we conclude that the use of such an approach would certainly have positive effects on balancing both environmental and agricultural sectors in vulnerable areas.

4. Conclusions

The demonstrated approach is an important step forward in finding balanced solutions to environmental protection and socioeconomic issues present in many nitrate-vulnerable areas around the globe. It enables nitrate–nitrogen leaching mitigation with site-specific measures that do not significantly affect agricultural productivity.

The aim of the study was to question whether it is possible to reduce nitrate–nitrogen leaching in a vulnerable area without imposing measures with negative effects to agricultural production. Based on the model results showing large reductions in nitrate–nitrogen leaching under different alternative scenarios, we can state that the first part of this assumption is true. The second part, concerning the sustainability of agricultural production, is mostly true, but not entirely. The grassland and arable crop scenarios especially demonstrated a more severe decrease in crop yields, while for the other groups, at least one of the alternatives retained the crop yields relatively well. Ultimately, the aim was reached as the results show that improvements in the environmental state do not necessarily deteriorate the state of agricultural sector equally.

As with any modeling study, there are plenty of modeling-related uncertainties to consider when interpreting the results. Unconventional model setups, where the watershed boundary does not match the model boundary exactly, carry more uncertainty than the conventional ones, but in such specific cases where modeling the whole large watershed is not possible due to different restraints, they are a valuable tool. Still, maybe the most significant uncertainty of this study is the soil-map resolution. Despite the efforts in obtaining the best possible input data, we were unable to spatially locate the hot spots in the actual CS areas. Rather, the information we obtained from the results was used to illustrate how the nitrate leaching and crop yields behave under different sets of conditions.

An important finding that is relevant for both the scientific community and broader public is that different measures or scenarios could prove effective at decreasing nitrate–nitrogen leaching in one soil type, while in another, the nitrate–nitrogen leaching under the same management could even increase, similar to the effectiveness across CS areas. This is well-aligned with past publications, where researchers found that one-size-fits-all solutions do not work in vulnerable areas and that a focused, localized approach must be utilized in order to successfully mitigate nitrate–nitrogen leaching in a specific area. The results of this study further strengthen the claim that a localized approach is a method that should be utilized by policymakers when designating groundwater-protection measures; this study presents a working method of how to approach this issue.

Despite the omnipresent calls for cleaner environment, the farming sector still needs to function in the economy, despite the subsidy system that alleviates some of the burden of environmental protection from its shoulders. Crop yield is one of the metrics for its effectiveness, but a more wholesome metric would be the economic outcome of the different farm enterprises. It would be interesting to study this aspect in future studies as the crop yield might decrease somewhat. However, operational costs would also decrease because of a decrease in the amount of necessary fertilizer.

The main improvements to the approach would be automatizing the allocation of certain measures to vulnerable areas and assessing the measures’ economic and social effects. The first would allow for the approach to be used on larger areas with more complexity, both in soil types and agricultural management. In contrast, the second would ensure policymakers are more informed on the effects of environmentally effective measures on the agricultural sector’s economic and social state. With this approach, researchers and policymakers have a tool to effectively balance agricultural and environmental needs.