The Effects of Agricultural Conservation Practices on the Small Water Cycle: From the Farm- to the Management-Scale

: Reinforcing the small water cycle is considered to be a holistic approach to both water resource and landscape management. In an agricultural landscape, this can be accomplished by incorporating agricultural conservation practices; their incorporation can reduce surface runoff, increase inﬁltration, and increase the water holding capacity of a soil. Some typical agricultural conservation practices include: conservation tillage, contour farming, residue incorporation, and reducing ﬁeld sizes; these efforts aim to keep both water and soil in the landscape. The incorporation of such practices has been extensively studied over the last 40 years. The Soil and Water Assessment Tool (SWAT) was used to model two basins in the Czech Republic (one at the farm-scale and a second at the management-scale) to determine the effects of agriculture conservation practice adoption at each scale. We found that at the farm-scale, contour farming was the most effective practice at reinforcing the small water cycle, followed by residue incorporation. At the management-scale, we found that the widespread incorporation of agricultural conservation practices signiﬁcantly reinforced the small water cycle, but the relative scale and spatial distribution of their incorporation were not reﬂected in the SWAT scenario analysis. Individual farmers should be incentivized to adopt agricultural conservation practices, as these practices can have great effects at the farm-scale. At the management-scale, the spatial distribution of agricultural conservation practice adoption was not signiﬁcant in this study, implying that managers should incentivize any adoption of such practices and that the small water cycle would be reinforced regardless. cumulative probability that a value falls at or below the respective ratio.


Introduction
The small water cycle is the local cycling of water, wherein water should fall as rain in the same geographic area from which it evapo(transpi)rates. The small water cycle also greatly emphasizes a reduction in surface runoff generation in a landscape, and the cycle's reinforcement is considered to be a holistic approach to managing water resources at the catchment scale [1][2][3]. In an agricultural landscape, certain conservation techniques can greatly improve the water holding capacity of a soil and can, in turn, strongly reinforce the small water cycle, making an agricultural landscape more resilient in the face of climate change.
Agricultural conservation practices have been extensively studied over the last 40 years and have been shown to significantly improve a soil's infiltration capacity and, consequently, significantly decrease the surface runoff in a landscape [4][5][6][7][8]. The most common agricultural conservation practices in modern literature include reduced/no-tillage, mulch cover/crop residues and cover crops, and reduced application of herbicides. The goal of conservation agriculture is to make soils "self-sustainable" by: maintaining sources of organic matter above and below the soil's surface, recycling water and nutrients within the system, and ensuring that the infiltration rate of a soil is greater than the predicted rainfall affect their impacts on the small water cycle at the management scale? (iv) What do these results imply regarding catchment management and incentivizing farmers to adopt these practices?

Study Watersheds
Both study watersheds are located in the Central Bohemia region of the Czech Republic ( Figure 1). This region is characterized by a humid continental climate and receives approximately 600 mm of precipitation per year. The rainy season in this region occurs from May through August, and the driest month is usually February. These two basins were selected for this study because they are typical of an intensively agricultural Czech landscape. Nučice is a simply-shaped catchment and represents the farm-scale, containing three large fields, each with very similar crop rotations and management. Vrchlice represents the management-scale. It is much larger (~100 km 2 ), with a more diverse landscape, and its water resources are managed to meet municipal needs. It is valuable to land owners as well as basin managers to determine the effects of agricultural conservation practice adoption at each scale.
Land 2022, 11, x FOR PEER REVIEW 3 of 18 cycle at the farm-scale? (iii) Does the spatial distribution of agricultural conservation practices affect their impacts on the small water cycle at the management scale? (iv) What do these results imply regarding catchment management and incentivizing farmers to adopt these practices?

Study Watersheds
Both study watersheds are located in the Central Bohemia region of the Czech Republic ( Figure 1). This region is characterized by a humid continental climate and receives approximately 600 mm of precipitation per year. The rainy season in this region occurs from May through August, and the driest month is usually February. These two basins were selected for this study because they are typical of an intensively agricultural Czech landscape. Nučice is a simply-shaped catchment and represents the farm-scale, containing three large fields, each with very similar crop rotations and management. Vrchlice represents the management-scale. It is much larger (~100 km 2 ), with a more diverse landscape, and its water resources are managed to meet municipal needs. It is valuable to land owners as well as basin managers to determine the effects of agricultural conservation practice adoption at each scale. The Nučice experimental catchment ("Nučice") has been monitored since 2011 by the Landscape Water Conservation Department (in the faculty of Civil Engineering) of Czech Technical University in Prague. It is a small watershed (~0.52 km 2 ) consisting of three fields that are managed by two farmers and is appropriate for modeling at the "farmscale" in the Czech Republic. Its outlet is located at 49°57′49.230′′ N, 14°52′13.242″ E (Figure 1b). The soils in Nučice are classified mainly as Luvisols and Cambisols overlaying siltstone and sandstone. The average slope in Nučice is 3.9% but ranges from 1 to 12%. Nučice is primarily cropland, with a very narrow riparian/brush zone around the stream; the basin is bisected horizontally by a 2-lane road (Table 1). The Nučice experimental catchment ("Nučice") has been monitored since 2011 by the Landscape Water Conservation Department (in the faculty of Civil Engineering) of Czech Technical University in Prague. It is a small watershed (~0.52 km 2 ) consisting of three fields that are managed by two farmers and is appropriate for modeling at the "farm-scale" in the Czech Republic. Its outlet is located at 49 • 57 49.230 N, 14 • 52 13.242" E ( Figure 1b). The soils in Nučice are classified mainly as Luvisols and Cambisols overlaying siltstone and sandstone. The average slope in Nučice is 3.9% but ranges from 1 to 12%. Nučice is primarily cropland, with a very narrow riparian/brush zone around the stream; the basin is bisected horizontally by a 2-lane road (Table 1).  Impervious  2  3  Brush  2  4  Forest  -25  Grassland  -8  Cropland  95  54  Water  1  2  Gardens  -4 The Vrchlice Basin ("Vrchlice") is much larger than Nučice, at~97 km 2 (Figure 1c). Vrchlice also has a more diverse land use, with large areas of forested land as well as many townships (Table 1), but it is still primarily cropland. The Vrchlice Reservoir provides drinking water to the nearby town of Kutná Hora, serving approximately 40,000 inhabitants. Its outlet is located at 49 • 55 37.211" N, 15 • 13 37.07" E. The basin is covered in clayey soils classified as Cambisols overlaying a metamorphic bedrock [24]. Vrchlice contains a network of nearly 140 reservoirs, mostly small fish ponds, that serve cultural and hydrologic significance. The discharge at the outlet of the Vrchlice Reservoir has been monitored by the Elbe River Authority since 1979. The Vrchlice Basin is considered to be an appropriate size for modeling at the "management-scale" in the Czech Republic.

Soil and Water Assessment Tool (SWAT)
SWAT requires the following as its bare minimum regarding data requirements: soils, slopes, land uses, and daily weather data. The input data used for each of the models present in this study are outlined in Table 2. The daily meteorological data for the Nučice SWAT model was obtained from on-site gauges ( Table 2). The climate data for this model was obtained from the Climate Forecast System Reanalysis (CFSR) database; these data are used in case there are any gaps in the observed weather dataset. The digital elevation model (DEM) was obtained from the fifth generation of the digital relief model of the Czech Republic (DMR5G) and was point-cloud processed to obtain a 3 m spatial resolution. The SWAT model for Nučice was developed using the field boundary method [34]. In the field boundary method scheme, each field is defined as its own HRU by aggregating the primary soil type and elevation class for each field. This method was selected in order to incorporate reduced field sizes at the farm-scale and was accomplished through the use of soil dummy variables. The SWAT model for Nučice was run during the growing seasons (~April through October) from 2014 through 2019, using 2013 as a warmup period. Calibration and validation procedures followed those outlined in Noreika et al. 2020 [26]. The Vrchlice SWAT model was developed originally for Noreika et al. 2021 to study the effects of land use and management changes over time in the basin [27]. The model itself has not been edited further. This model was run at the monthly timestep from 2001 through 2019 with a 5-year warmup period (1996)(1997)(1998)(1999)(2000). The monthly timestep was chosen to minimize daily effects due to reservoir processes that are not publicly available and therefore unable to be represented in SWAT. The model was calibrated (2001-2012) and validated (2013-2016) at the monthly timestep with discharge data from the basin's outlet. The basin boundaries and stream network were largely DEM-based, but ground-truthed to existing data. Vrchlice was divided into 63 sub-basins, containing 1058 HRUs that were defined by their unique combinations of land use, slope class, and soil type. For further detail, parameterization, and intricacies of the model setup, please refer to Noreika et al. 2021 [27].

Literature Review
Contour farming results in a reduction of surface runoff by impounding water in small depressions, as well as a reduction of sheet and rill erosion by reducing the erosive power of surface runoff and preventing or minimizing the development of rills. This practice is represented by adjusting the Soil Conservation Service (SCS) curve number in SWAT. Residues are meant to slow down surface and peak runoff by increasing surface roughness. They also increase infiltration and reduce surface runoff by decreasing surface sealing and slowing down overland flow. Finally, residues reduce sheet and rill erosion by reducing surface flow volume. In SWAT, there is significant literature precedent to incorporate these practices; conservation tillage and residue management are typically represented by adjusting the curve number and Manning's roughness coefficient for overland flow, respectively. In order to incorporate these practices appropriately, a literature review was conducted using the following keywords: SWAT, best management practice, and conservation agriculture . A total of 33 articles were downloaded and narrowed down to 25 based on relevance. The 25 remaining papers addressed the incorporation of conservation tillage operations, contour farming, and residue management into SWAT (  [32,33], meaning that conservation practices were incorporated via the curve number (CN) method. Three publications introduced tillage operation changes and no CN edits (TO). Two introduced tillage operation changes along with the CN edits (CN + TO). Two modified the CN by a percent change, and two did not specify (NS) how the practices were incorporated into SWAT. We then conducted a scenario analysis at the Nučice basin to determine whether it is necessary to incorporate both CN shifts and tillage operation changes. We found no significant differences between water balance variable outputs (discharge at the basin's outlet, subsurface lateral flow, surface runoff, evapotranspiration, and soil water content, p > 0.05) when only the CN method was utilized versus shifting both the CN and the tillage practices. We concluded that the CN method is appropriate to incorporate agricultural conservation practices and is also more efficient in the modeling process. conducted a scenario analysis at the Nučice basin to determine whether it is necessary to incorporate both CN shifts and tillage operation changes. We found no significant differences between water balance variable outputs (discharge at the basin's outlet, subsurface lateral flow, surface runoff, evapotranspiration, and soil water content, p > 0.05) when only the CN method was utilized versus shifting both the CN and the tillage practices. We concluded that the CN method is appropriate to incorporate agricultural conservation practices and is also more efficient in the modeling process.

Scenarios Outlined
Five scenarios were run at the farm-scale for this study (Table 3). These scenarios incorporate contour farming, small residues (0.5-1 t/ha), large residues (1-9 t/ha), conservative tillage, and field size reductions at Nučice. To incorporate field size reductions, instead of three fields averaging 17 ha each, Nučice was divided into 52 fields averaging 1 ha each through the use of dummy soil variables. The other scenarios were incorporated as presented in Table 4.

Scenarios Outlined
Five scenarios were run at the farm-scale for this study (Table 3). These scenarios incorporate contour farming, small residues (0.5-1 t/ha), large residues (1-9 t/ha), conservative tillage, and field size reductions at Nučice. To incorporate field size reductions, instead of three fields averaging 17 ha each, Nučice was divided into 52 fields averaging 1 ha each through the use of dummy soil variables. The other scenarios were incorporated as presented in Table 4. In Vrchlice, only the "General Measures" agricultural conservation scenario was adopted (Table 4) at various scales across the basin (Table 3, Figure 3). The "General Measures" outlined in Table 4 are considered to be "best case scenarios" to represent conditions if the practices were incorporated properly and if the landscape responds as expected, but it is likely that any real-world result would fall somewhere between the calibrated model without any conservation practices and the "General Measures" scenarios. Vrchlice was divided into three regions based on location in the basin and percent area cropland (Figure 3  A requirement for the Random scenario is that no selected sub-basins should be adjacent. The Random scenario controls for the effects of connectivity of agricultural conservation practices to determine if individual farm adoption is "enough" or if regional adoption is necessary to more greatly reinforce the small water cycle. These scenarios were outlined so that the individual impacts of agricultural conservation practice continuity and spatial adoption within the basin could be evaluated.

Results
According to the global sensitivity analysis that was conducted, three parameters significantly influenced the modeled discharge flowing out of the Nučice experimental basin (Table 5). RCHRG_DP is the deep aquifer percolation fraction; this value should fall between 0 and 1 as it is the fraction of percolation past the root zone which recharges the deep aquifer. Since this value is very close to 0, this indicates that a very small fraction of water entering the Nučice Basin recharges the deep aquifer. Saturated hydraulic conductivity (SOL_K) and the available water capacity of the soil (SOL_AWC) govern how water is infiltrated and retained in a soil, respectively, were also significantly sensitive parame-

Results
According to the global sensitivity analysis that was conducted, three parameters significantly influenced the modeled discharge flowing out of the Nučice experimental basin (Table 5). RCHRG_DP is the deep aquifer percolation fraction; this value should fall between 0 and 1 as it is the fraction of percolation past the root zone which recharges the deep aquifer. Since this value is very close to 0, this indicates that a very small fraction of water entering the Nučice Basin recharges the deep aquifer. Saturated hydraulic conductivity (SOL_K) Land 2022, 11, 683 8 of 16 and the available water capacity of the soil (SOL_AWC) govern how water is infiltrated and retained in a soil, respectively, were also significantly sensitive parameters. Calibration (2016-2018) and validation (2019) for the Nučice basin were conducted with SWAT-Cup 2019, which is a semiautomatic calibration methodology [28]. Table 6 presents the selected model performance indicators during the calibration and validation periods for the Nučice SWAT model. Figure 4 presents a scatterplot, correlating the modeled discharge values with the observed discharge values at Nučice during the calibration and validation periods.  There were significant shifts across water balance parameters with the incorporation of agricultural conservation practices at the Nučice scale ( Figure 5). The incorporation of residues and contour farming reinforced all of the small water balance parameters when compared to the calibrated scenario, which included generalized conservation tillage. Resorting to conventional tillage from conservation tillage was consistently contradictory to the goal of reinforcing the small water cycle. Field size reductions resulted in the highest amount of streamflow contribution from subsurface lateral flow, but the model indicated that otherwise the adoption of smaller fields does not reinforce the small water cycle. All small water cycle parameters, except for discharge at Vrchlice's outlet, were significantly affected by the incorporation of agricultural conservation practices across the basin. Interestingly enough, neither the scale of adoption nor the spatial distribution of agricultural conservation practices significantly affected any small water cycle parameters at this scale; further figures presented compare only Vrchlice's conventional tillage (calibrated model) and the full adoption scenario. Both the available water content and evapotranspiration in the conventional tillage scenario are consistently lower than the full conservation adoption across the entire year (Figures 6 and 7). Both the surface runoff ratios and subsurface lateral flow were significantly higher throughout the year in the conventional tillage scenario when compared to the General Measures full adoption scenario (Figure 8 and 9). Generally surface runoff in the conventional tillage scenario is greater than 2× that of the conservation scenario ( Figure 9).  All small water cycle parameters, except for discharge at Vrchlice's outlet, were significantly affected by the incorporation of agricultural conservation practices across the basin. Interestingly enough, neither the scale of adoption nor the spatial distribution of agricultural conservation practices significantly affected any small water cycle parameters at this scale; further figures presented compare only Vrchlice's conventional tillage (calibrated model) and the full adoption scenario. Both the available water content and evapotranspiration in the conventional tillage scenario are consistently lower than the full conservation adoption across the entire year (Figures 6 and 7). Both the surface runoff ratios and subsurface lateral flow were significantly higher throughout the year in the conventional tillage scenario when compared to the General Measures full adoption scenario (Figures 8 and 9). Generally surface runoff in the conventional tillage scenario is greater than 2× that of the conservation scenario ( Figure 9).

Hydrological Modeling with SWAT
There are several possible sources of error in any hydrologic model; the first is input parameter uncertainty, which is the largest possible source of error and also influences uncertainties associated with output data. Model parameterization and model structural uncertainties are additional possible sources of error [62,68]. Furthermore, since SWAT is neither fully physically based nor fully distributed, some processes may not be properly represented, such as temporal changes in topsoil hydraulic properties, preferential flow, or the influences of the spatial distribution of fine-scale land management [28,69,70]. While there are some drawbacks to the SWAT model (as stated above), it is a very useful tool for hydrologic modeling, especially regarding scenario analysis. Currently, Nučice is equipped to model generalized processes rather than more spatially distributed processes (piezometer clusters and a cosmic-ray neutron sensor are currently being installed). The soil data at this scale is fairly coarse and is nearing the lower spatial range of SWAT's modeling capabilities, but SWAT was still able to model Nučice effectively with "good" or "very good" performance across the selected indicators [71,72,73]. The uncertainties associated with the Vrchlice model primarily include generalized reservoir processes and crop rotations [27]. Vrchlice was able to be effectively modeled at the monthly timescale, also with "good" and "very good" performance indicators [71,72,73]. While SWAT was able to model significant shifts in water balance parameters with the incorporation of agricultural conservation practices in Vrchlice, it was unable to represent significant differences at varying scales and distributions of incorporation across Vrchlice. This could be due to the fact that Vrchlice, while primarily cropland, contains significant areas of forested areas and riparian zones, which may disguise the effects of agricultural Figure 9. Surface runoff ratios between full agricultural conservation practice adoption and the default conventional tillage scenario. The histogram bars represent the cumulative probability that a value falls at or below the respective ratio.

Hydrological Modeling with SWAT
There are several possible sources of error in any hydrologic model; the first is input parameter uncertainty, which is the largest possible source of error and also influences uncertainties associated with output data. Model parameterization and model structural uncertainties are additional possible sources of error [62,68]. Furthermore, since SWAT is neither fully physically based nor fully distributed, some processes may not be properly represented, such as temporal changes in topsoil hydraulic properties, preferential flow, or the influences of the spatial distribution of fine-scale land management [28,69,70]. While there are some drawbacks to the SWAT model (as stated above), it is a very useful tool for hydrologic modeling, especially regarding scenario analysis. Currently, Nučice is equipped to model generalized processes rather than more spatially distributed processes (piezometer clusters and a cosmic-ray neutron sensor are currently being installed). The soil data at this scale is fairly coarse and is nearing the lower spatial range of SWAT's modeling capabilities, but SWAT was still able to model Nučice effectively with "good" or "very good" performance across the selected indicators [71][72][73]. The uncertainties associated with the Vrchlice model primarily include generalized reservoir processes and crop rotations [27]. Vrchlice was able to be effectively modeled at the monthly timescale, also with "good" and "very good" performance indicators [71][72][73]. While SWAT was able to model significant shifts in water balance parameters with the incorporation of agricultural conservation practices in Vrchlice, it was unable to represent significant differences at varying scales and distributions of incorporation across Vrchlice. This could be due to the fact that Vrchlice, while primarily cropland, contains significant areas of forested areas and riparian zones, which may disguise the effects of agricultural conservation adoption. Additionally, since Vrchlice is of significant size and SWAT is not fully distributed, the effects of the scale of agricultural conservation practice adoption may be aggregated across the basin, leading to insignificant changes across the agricultural conservation practice adoption scenarios.

The Small Water Cycle at the Farm-Scale
At the farm-scale, SWAT was able to model significant differences in water balance parameters across agricultural conservation practice scenarios. According to SWAT, residue incorporation and contour farming were the most effective at reinforcing the small water cycle and should be prioritized by farmers to aid the holistic management of their land in the face of future climate change [17]. Although it was not in the scope of this study to investigate the effects of crop changes in addition to the incorporation of agricultural conservation practices, the previous SWAT study of the Nučice basin indicated that crop changes also have significant impacts on the small water cycle [26]. For instance, winter wheat reinforces the small water cycle to a greater degree than rapeseed in the Czech landscape. The incorporation of contour farming and crop residues may be able to mitigate water balance issues that arise from less-sustainable crop choices, and the interaction should be studied further.
SWAT was not able to effectively model the impacts of incorporating smaller field sizes at Nučice. This may be due to several factors: SWAT is not fully distributed and cannot model the spatial effects influenced by smaller field incorporation, crop changes were not incorporated across the smaller fields, and SWAT does not model true border effects between fields. To replicate this in future studies, a trap efficiency would need to be applied to each HRU to simulate flow disruption between fields. The field boundary HRU method may be more useful to identify "hotspot" fields that may be susceptible to erosive events due to their slopes, crops, and soil types [28,30,34,69,70].

The Small Water Cycle at the Management-Scale
The adoption of agricultural conservation practices in at least 33% of the cropland across Vrchlice had significant effects on the small water cycle within the basin. Neither the distribution nor the scale of adoption (anything above 33%) significantly affected the small water cycle variables at Vrchlice any further. While Vrchlice is a very agricultural basin (>50% cropland), there are also very large forested and riparian areas that may mask the effects of various intensities of agricultural conservation practice adoption. It may also be due to SWAT's model structure, being semi-distributed and semi-physically based, that some effects at this scale may be lost due to HRU aggregation or generalizations due to using the curve number method [34,74,75]. While SWAT models significant impacts on the small water cycle due to the adoption of agricultural conservation practices, SWAT cannot represent realistic effects when additional spatial distribution and connectivity scenarios are introduced; a fully distributed model would be necessary for this purpose. However, SWAT was able to model general trends and could represent significant differences between conventional agricultural practices and full conservation adoption.
When compared to the effects of land use changes at Vrchlice [27], average soil water content and subsurface lateral flow shifts fell in similar ranges to that of agricultural conservation practice incorporation scenarios. However, the modeled adoption of agricultural conservation practices reduced the proportion of surface runoff at the management scale by up to 30×, which greatly outweighs the effects of the land use change scenarios previously modeled [27]. These findings indicate that, at the management-scale, the incorporation of agricultural conservation practices can have similar effects to land use changes on the small water cycle and can greatly reduce the overall proportion of surface runoff contribution to streamflow.

Implications for Agricultural Conservation Practice Incorporation in the Czech Republic
The incorporation of agricultural conservation practices tend to reinforce the small water cycle regardless of scale of incorporation. These effects are more obvious at the farm-scale than at the management-scale, which should motivate individual farmers to adopt such practices. At the management-scale, the effects of agricultural conservation practices were still significant but the scale and the spatial distribution of adoption were not. This implies that managers should incentivize any willing famers/conglomerates within their management area to adopt such practices. In addition to agricultural conservation practices, other land and crop management factors can also have significant effects on the small water cycle and their interactions should be studied further [26,27]. While soil erosion and sediment transport were not explored in this study, agricultural conservation practices have also been shown to have positive effects concerning these issues and can lead to increased soil conservation [14][15][16]76,77].

Conclusions
This study reinforces SWAT's applicability to the Czech landscape at both the farmand management-scales. SWAT is very effective in its ability to model various management, land use, and crop change scenarios. While likely exaggerated by the scale, agricultural conservation practice adoption at the farm-scale has significant effects on the small water cycle. The most effective practice modeled at this scale was the incorporation of contour farming. The effects of small field incorporation at the farm-scale tended to have significantly negative impacts on the small water cycle, but this result is likely an artifact due to the HRU processing in SWAT. At the management-scale in the Czech Republic, any degree of incorporation of agricultural conservation practices makes significant impacts on the small water cycle, according to the Vrchlice SWAT model. SWAT was able to model that the incorporation of agricultural conservation practices in a primarily agricultural landscape can have significant effects on the small water cycle, especially regarding surface runoff ratios. While SWAT is not fully distributed and real-world effects would likely vary, this study indicates that managers should encourage agricultural conservation practices, regardless of scale or spatial distribution. As this study only focuses on the effects of agricultural conservation practices on the small water cycle, further studies should be conducted to model their effects on erosion as well as the interactions between agricultural conservation practices and land use/management changes in the Czech landscape.

Funding:
The presented research has been performed within projects H2020 No. 773903 Shui, focused on water scarcity in European and Chinese cropping systems and the Grant Agency of Czech Technical University in Prague, No. SGS20/156/OHK1/3T/11.