Field‐scale analysis of miscanthus production indicates climate change may increase the opportunity for water quality improvement in a key Iowa watershed

The Raccoon River Basin is the primary source for drinking water in Iowa's largest city and plays a major role in the Mississippi River Basin's high nutrient exports. Future climate change may have major impacts on the biological, physiological, and agronomic processes imposing a threat to ecosystem services. Efforts to reduce nitrogen (N) loads within this basin have included local litigation and the implementation of the Iowa Nutrient Reduction Strategy, which suggest incorporating bioenergy crops (i.e., miscanthus) within the current corn–soybean landscape to reach a 41% reduction in nitrate loads. This study focuses on simulating N export for historical and future land use scenarios by using an agroecosystem model (Agro‐IBIS) and a hydrology model (THMB) at the 500‐m resolution, similar to the scale of agricultural fields. Model simulations are driven by CMIP5 climate data for historical, mid‐century, and late‐century under the RCP 4.5 and 8.5 warming projections. Using recent crop profit analyses for the state of Iowa, profitability maps were generated and nitrogen leaching thresholds were used to determine where miscanthus should replace corn–soybean area to maximize reductions in N pollution. Our results show that miscanthus inclusion on low profit and high N leaching areas can result in a 4% reduction of N loss under current climate conditions and may reduce N loss by 21%–26% under future climate conditions, implying that water quality has the potential continue to improve under future climate conditions when strategically implemented conservation practices are included in future farm management plans.


| INTRODUCTION
Nitrogen (N) loss from agricultural production in the US Corn Belt is one of the leading causes of impairment to water quality within the Mississippi River Basin . Some watersheds within the Mississippi River Basin, such as the Raccoon River Basin (RRB), have a particularly high fraction of land under agricultural production, and nutrient reduction has become a high priority (Iowa Agriculture Water Alliance [IAWA], 2020). The RRB, located in west-central Iowa, has contributed high N loads into the main drinking water source for the city of Des Moines as well as to the Gulf of Mexico Goolsby et al., 2000;Hatfield et al., 2009;Jones et al., 2016Jones et al., , 2018Schilling & Zhang, 2004). The cost of N removal from the Raccoon River to achieve nitrate concentrations below the 10 ppm EPA standard was greater than 1.5 million US$ in 2015 and prompted Des Moines Water Works (DSMWW) to pursue litigation in 2016 against three upstream drainage districts along the North Raccoon River (Stowe, 2015). Implementing changes to agricultural practices broadly at a watershed scale is thought to be untenable since land management is decided mostly at the scale of individual fields. This suggests reducing initial N inputs upstream may be a more sustainable solution to reducing N loss.
The amount of N lost from an ecosystem depends on the imbalance between N uptake by plants and N availability in soils which is dictated by numerous climate, management, and biogeochemical factors (Shrestha et al., 2023). In agroecosystems, N loss to the atmosphere via denitrification to N 2 O is one of the main greenhouse gas (GHG) contributing factors of agricultural production, however, nitrate dissolved in subsurface water is the largest N loss flux on a mass basis. Management decisions play a key role in dictating the imbalance of N (Gardner & Drinkwater, 2009). For example, warming soil temperatures associated with a changing climate along with ample soil moisture may increase net N mineralization rates thereby increasing the pool of nitrate in the soil (Cambardella et al., 1999;Fernandez et al., 2017;Melillo et al., 2002;Zaehle et al., 2010). Without a subsequent increase in plant uptake this may contribute to greater N loss especially during fallow periods (Tonitto et al., 2006). Systems with soybean in rotation with corn have shown to have slightly greater amounts of mineralizable soil N compared to continuous corn, even when no N fertilizer is applied (Carpenter-Boggs et al., 2000). Croplands managed under corn-soybean rotations in this region are often fertilized with synthetic N and manure and are located on land modified with subsurface tile drainage (Jones et al., 2019;Jones & Schilling, 2011;Schilling et al., 2008). While the practice of tile drainage allows for improved crop production by reducing excess soil moisture and lowering the soil water table, it also allows for more N loss into nearby waterways (Amado et al., 2017). Additionally, soybean and corn in rotation have the potential to leach similar amounts of N with and without the addition of fertilizers (Lawlor et al., 2008).
The Iowa Nutrient Reduction Strategy (INRS) has presented many management strategies to reach the goal of reducing nitrogen loss by 41% (INRS, 2013). Example strategies include adjusting the timing of fertilizer application and planting cover crops. The strategy that has potential for the largest reduction in N loss is planting perennial energy crops like miscanthus (Miscanthus × giganteus Greef et Deu.; Chae et al., 2014). While the inclusion of miscanthus on cropland has been shown to reduce nitrate loads in our current climate (Brandes et al., 2018;Daigh et al., 2015;Hernandez-Ramirez et al., 2011;Hussain et al., 2020;McIsaac et al., 2010;Smith et al., 2013;VanLoocke et al., 2017), few efforts have quantified how water quality will respond from land use change to perennials under future climates (Chen et al., 2017;Rajib et al., 2016;Teshager et al., 2016).
Water quality challenges that have previously affected the RRB occurred during relatively favorable climate conditions and stable corn and soybean production (Basso et al., 2019;Butler et al., 2018;Gutowski & Takle, 2020). Moving forward, global change in the form of increased atmospheric carbon dioxide (CO 2 ) concentration, temperature, precipitation, and vapor pressure deficit (VPD) could decrease crop productivity in future years to come (Gutowski & Takle, 2020;Jin, Ainsworth, et al., 2017;Jin, Zhuang, et al., 2017;Lobell et al., 2014;Ort & Long, 2014;Ruiz-Vera et al., 2013, 2015. While increases in CO 2 concentration alone have the potential to directly increase soybean yields from CO 2 fertilization and indirectly future climate conditions, implying that water quality has the potential continue to improve under future climate conditions when strategically implemented conservation practices are included in future farm management plans.

K E Y W O R D S
agroecosystem modeling, Agro-IBIS, climate change, miscanthus, nitrogen, water quality increase corn yields through improved water use efficiency (Leakey et al., 2004;Long et al., 2004Long et al., , 2006, incorporating the impacts of increased temperature, VPD, and extreme and variable precipitation events may have varying and confounding effects (Gutowski & Takle, 2020;Jin, Ainsworth, et al., 2017;Jin, Zhuang, et al., 2017;Ruiz-Vera et al., 2013, 2015. When crops become stressed due to these abiotic factors, photosynthesis and N uptake may be reduced (Basso et al., 2019;Hussain et al., 2019;Lamaoui et al., 2018), allowing more N to remain in the field. This N has the potential to be removed from a field through waterways via leaching, subsurface tile drainage, and surface runoff and in turn can worsen water quality. Drainage and streamflow rates could increase if the timing of increased temperatures and extreme precipitation events align with prime mineralization peaks within a field. This could result in a reduction of crop yield and ultimately lead to an increase in N loss prior to the establishment of a mature root system .
Stresses associated with climate change have impacts beyond a reduction in yield and profitability at the field scale and increased water quality issues at the watershed scale. Continuing to grow corn and soybean in specific regions while obtaining a profit and benefiting the environment may be challenging to achieve (Brandes et al., 2016(Brandes et al., , 2018Dixon & Segerson, 1999), especially in a changing climate. Land use and management changes will be needed to increase profitability and improve ecosystem services like increasing carbon sequestration and reducing runoff and N loss from fields. These changes could be driven from the "top down" by policy (i.e., Renewable Fuel Standard) or from the "bottom up" by profitability (i.e., economic returns to individual stakeholders).
Perennials have been identified as a potential option to mitigate profit loss and negative environmental impacts (Asbjornsen et al., 2014;Brandes et al., 2018). Coarse scale studies suggest that the inclusion of miscanthus could reduce N loss within small agricultural watersheds and HRUs (Hydrologic Response Unit; Housh et al., 2015;Rajib et al., 2016;Teshager et al., 2016) and across the entire Mississippi River Basin but only if policies couple economic and N loss targets . Cropping system processes like N leaching vary significantly at small spatial scales (Basso et al., 2019;Nowatzke et al., 2022). Targeted changes to land management, like perennial integration, could lead to disproportionate benefits (i.e., improvements that exceed unity with the proportion of area changed; Asbjornsen et al., 2014;Brandes et al., 2018) therefore signifying the importance of running simulations at the field scale. For example, Brandes et al. (2018) showed that one could reduce N loss by incorporating switchgrass into parts of fields with low profitability and high N leaching leading to 18% reduction in N loss with only a 12% conversion of cropland.
Previous studies have shown the importance of conducting a high-resolution field-scale analysis to account for high variability in yield and profit within a single field (Basso et al., 2019;Brandes et al., 2016). However, key limitations of these profitability studies include simple assumptions made about crop growth, lack of an agroecosystem crop model to resolve dynamics of N uptake and leaching with changes in yield, little to no representation of perennial grass yield, and lack of quantifying the impact of future climate and future crop production on land use change and water quality. Miscanthus yield has the potential to increase under future climate change due to greater biomass accumulation from longer growing seasons and earlier emergence, the effect of water savings from CO 2 fertilization at elevated CO 2 concentration, and a reduction in cold days that would trigger senescence due to increases in temperature (Chen et al., 2017;Purdy et al., 2015;Tejera & Heaton, 2017). Therefore, testing within mechanistic models is necessary to assess the potential of perennials like miscanthus to affect water and nutrient cycling under present and projected climate regimes.
This study combines the coarse scale biophysical approach of Ferin et al. (2021) with the field-scale approach presented in Brandes et al. (2018), in which perennials were placed on areas with low profitability and high N loss, with updated profit maps generated from model simulated corn and soybean yields, land rent, and input costs (Plastina, 2019) as well as an explicit consideration of climate change in the RRB. The objective of this study was to quantify changes in crop productivity (i.e., total production) and water quality (i.e., N leaching, dissolved inorganic N (DIN) export) and quantity (i.e., streamflow) under future land use and climate change in the RRB. Here we use an agroecosystem model (Agro-IBIS; Kucharik et al., 2000;VanLoocke et al., 2010) and a hydrology model (THMB; Donner et al., 2002), to simulate crop productivity and water quality and quantity for the RRB at the field scale (500 m resolution) for a combination of contemporary and future climate and land use scenarios (Table 1). These scenarios were used to address two hypotheses: (1) cropping system productivity and water quality will decrease more under future climate conditions for corn and soybean relative to miscanthus, and (2) strategic profit-based implementation of miscanthus production will have a greater impact on water quality under future climate relative to contemporary conditions. These hypotheses are evaluated on total crop production, the total mass of N leached within the RRB, and the total mass of N exported as DIN at the location of the Van Meter, Iowa U.S. Geological Survey (USGS) gauge.

| Model procedure
This study follows a similar modeling procedure from Ferin et al. (2021) which focused at the Crop Reporting District level (i.e., nine CRDs in Iowa), but here we have adapted the methodology to run at a field-scale resolution (i.e., 500 m; 0.25 km 2 ; 754,350 individual grid cells in Iowa). Our approach required the assembly of soil characteristics, climate, and land use/management data layers as inputs to physically based models which generated outputs of yield, water flow, and N leaching for each pixel in our domain and then routed flows from each pixel through waterways of the RRB. County-level N fertilizer and manure rates for corn and soybean from 2016 were applied within Agro-IBIS ( Figure S1; Lark et al., 2022). Scenarios were created to account for the combination of contemporary and future climate with and without the integration of miscanthus and Agro-IBIS was driven with contemporary and future (mid-and late-century) climate data generated from the Weather Research Forecast (WRF) model downscaled CMIP5 dataset (5th Coupled Model Intercomparison Project; Liess et al., 2022;Taylor et al., 2012) for two Representative Concentration Pathways (RCP) scenarios. Contemporary land use fractions were generated for continuous corn, corn-soybean rotation, and natural vegetation using Cropland Data Layer (CDL; Boryan et al., 2011) data from 2017 and 2018 ( Figure S2). These masks were used to create fractions of land use type per grid cell and used to weight each grid cell of Agro-IBIS output to create a mosaic landscape with heterogeneous land use .
We used the agroecosystem model, Integrated Biosphere Simulator-Agricultural Version (Agro-IBIS; Kucharik et al., 2000;VanLoocke et al., 2010), to simulate the growth of corn, soybean, natural vegetation, and miscanthus and associated N loss within the RRB located in central Iowa ( Figure 1). From there, N leaching, drainage, precipitation, and runoff from Agro-IBIS were simulated through the Terrestrial Hydrology Model with Biogeochemistry (THMB; Donner et al., 2002;VanLoocke et al., 2017) to quantify DIN export and streamflow. Profitability maps were generated using 2019 county average land rent price ( Figure S3), input costs, and grain prices (Plastina, 2019;Section S5). We used Agro-IBIS simulated yields (described in Section S5), which were evaluated against NASS yield data from 2010 to 2019 (see Section S4), to generate the profitability maps as described below. The average profit of corn and soybean along with N leaching rates from the Agro-IBIS simulations was used to determine the location of miscanthus placement. To assess the productivity of the landscape in future climates, we followed the same modeling workflow for Agro-IBIS and THMB as previously mentioned.
T A B L E 1 Descriptions detailing the propose, land use and management, and climate choices used for each modeling scenario.

| Scenario development
The Historical Evaluation determines model accuracy based on the Agro-IBIS simulated yields compared to USDA NASS yields as well as the THMB simulated N loads compared to the USGS observations (Table 1). This scenario used ZedX observational weather data (Kucharik et al., 2013), which was regridded to the 500 m resolution. County-based N and manure fertilizer rates were also applied to corn and soybean (Lark et al., 2022; Figure S1). Parameters were calibrated such that simulated corn and soybean yield fell within one standard deviation of the USDA NASS County average yields for 1980 to 2007. Historical CDL data ( Figure S2) was then used to weight the daily Agro-IBIS output using the method described in Ferin et al. (2021). Streamflow and DIN export observations from the Des Moines River Water Quality Network Van Meter USGS gauge were used for evaluating THMB output from 1980 to 2007 (DMRWQN, 2014;Lutz & Esser, 2002). Agro-IBIS modeled yield was also compared with the CMIP5 historical climate data for the same counties as the Historical Evaluation to ensure that downscaled CMIP5 simulated weather conditions provided reasonable growing conditions. The Baseline scenario used the same historical CDL land use ( Figure S2) and modeled yields grown under a CO 2 concentration of 417 ppm. Fertilizer and manure application rates were held constant at 2016 rates (Lark et al., 2022). Agro-IBIS simulated yields were compared to county average NASS yields for 2010-2019 (Section S4). Parameters were slightly adjusted to reach NASS yields in this timeframe and were then used for the CMIP5 simulations.
To address our hypotheses regarding the interacting effects of climate change on yield, profitability, and water quality, we created three additional scenarios; results from scenarios 1 and 2 addressed hypothesis 1; and results from scenario 3 addressed hypothesis 2 (Table 1). Scenario 1 used the same years of corn and soybean simulations that were used in the Baseline scenario. However, land use within Scenario 1 was modified to include miscanthus determined by 10-year average corn and soybean profitability and N leaching from the Baseline scenario. Scenario 2 consisted of corn and soybean yields grown under future climates. Future climate projections included the midcentury RCP 4.5 (Mid 4.5), late-century RCP 4.5 (Late 4.5), and late-century RCP 8.5 (Late 8.5) warming scenarios (van Vuuren et al., 2011). While all scenarios maintained a constant hybrid growing degree day (GDD) length to reach maturity in corn of 2000°C day, an additional Late 8.5 projection was included in which corn hybrid GDD for maturity was optimized to 2800°C day (hereby referred to as Late 8.5 Opt; see Section S5). The CDL land use fractions from the Baseline were also used in Scenario 2. Scenario 3 was created to include the effect of miscanthus integration on N loss under a future climate. This scenario uses the same climate conditions and corresponding projected corn and soybean yields as in Scenario 2, with the addition of miscanthus on grid cells using the same water quality and profitability thresholds from Scenario 1.

| Determination of land allocation
Land use masks created from 2017 and 2018 CDL data were used to weight modeled output to create a gridded heterogeneous landscape that reflects current RRB land use (e.g., Figure S2). Because corn-soybean rotation land was not an explicit layer type in the CDL dataset, we used 2017 and 2018 to determine where these rotations were F I G U R E 1 This figure represents the orientation of the North and South Raccoon River within Iowa. The blue lines represent the connection of rivers within this basin and the red star marks the outlet of the basin in Des Moines, IA. located (Lark et al., 2017). For each CDL polygon, if both years contained corn, this land was classified as continuous corn. When 1 year included corn and the following year had soybean, or vice versa, this land was classified as corn-soybean rotation. All remaining land was classified as natural vegetation (Foley et al., 1996). These CDL polygon masks were then regridded from 30 to 500 m resolution to match the resolution used for the Agro-IBIS and THMB simulations ( Figure S2).
The contemporary land use mask generated from 2017 and 2018 CDL data was then modified for each scenario to include miscanthus. Following the methodology of Brandes et al. (2018), if profitability for the average of corn and soybean was less than US$-100 ha −1 and N leached in the weighted scenario under contemporary land use was greater than 50 kg-N ha −1 , then the grid cell was replaced with 100% miscanthus. These updated land use masks were used to assess the impact of including miscanthus into the landscape under both contemporary and future climates.

| Soil, fertilizer, and climate data
The dominant soil texture classification was generated at 500 m resolution from the gSSURGO (Gridded USDA-NRCS Soil Survey geographic) database (Soil Survey Staff, 2018) for 11 soil layers (0-200 cm) and was used in Agro-IBIS for both historical and future scenarios. Spatial distributions of soil texture and fractions of plant-available water, field capacity, and wilting point can be found in Figures S4 and S5.
This study used county average fertilizer rates for 1980-2016 from Lark et al. (2022). These data included N fertilizer for corn and manure rates (constant across crop type) within Iowa. It was assumed that N fertilizer rates were held constant at 0 kg-N ha −1 for soybean, but manure was applied. These county-based data were regridded to match the 500 m resolution used for the other model inputs. The Historical Evaluation used nutrient application rates for 1980-2007 while all other scenarios depicted in Table 1 were held constant at rates from 2016 ( Figure S1).
Historical and future climate scenarios were developed using the 5th Coupled Model Intercomparison Project (CMIP5; Taylor et al., 2012). The historical simulation (1980-1999) used observed GHG concentrations while future scenarios included mid-century (2040-2059) forced by RCP 4.5 and two late-century (2080-2099) scenarios forced by RCP 4.5 and 8.5 warming projections. Atmospheric CO 2 concentration was assumed to be 417, 500, 540, and 825 ppm for the Baseline, Mid 4.5, Late 4.5, and Late 8.5, respectively. These values were determined from projected concentrations at the end of the respective period and held constant through the simulations (van Vuuren et al., 2011). Climate variables in this dataset included precipitation, incoming solar radiation, temperature, humidity, and wind speed. The CMIP5 model output at the coarse resolution (approximately 1° × 1°) was then dynamically downscaled to 10 km horizontal resolution for the Midwest US using the WRF model (Broadbendt et al., 2020;Jin, Zhuang, et al., 2017;Skamarock et al., 2008). Temperature and precipitation output were bias-corrected using a simple linear relationship (Teutschbein & Seibert, 2012) with the PRISM dataset (Daly et al., 2017).
The middle 10-year period of each model run was used (i.e., 1987-1996; 2047-2056; 2087-2096) to minimize the effects of initialization and model bias. The downscaled CMIP5 datasets were regridded for the state of Iowa at 500 m resolution to use in Agro-IBIS. This study focused on the BCC-CSM1-1 (Beijing Climate Center-Climate System Model) dataset which compared well to historical precipitation observations over Iowa ( Figure S6). Spatial maps of growing season average temperature and total precipitation are described in Section S3; Figures S7 and S8.

| Historical evaluation of THMB
Observed county average 1980-2007 NASS corn and soybean yields were evaluated against Agro-IBIS simulated yields for five counties within the RRB (Section S4; Figure S9). Averaged across the 28 years and five counties, observed and simulated corn yields were 8.72 Mg ha −1 (138.9 bu a −1 ) and 8.75 Mg ha −1 (139.4 bu a −1 ), respectively, while soybean yields were 2.90 Mg ha −1 (43.1 bu a −1 ) and 2.89 Mg ha −1 (42.9 bu a −1 ) for observed and simulated, respectively. Monthly modeled streamflow and DIN export at the Van Meter gauge was very similar to the observed data between 1980 and 2007 ( Figures S11 and S12). The timing of peaks in streamflow was captured in THMB with the occasional underestimation of magnitude during abrupt changes in streamflow rate (e.g., 1993 in Figures S11 and S13). There was slight overestimation in DIN export throughout the evaluation period, however, visually similar magnitudes and temporal patterns were observed in the THMB simulations relative to the USGS observation dataset. Additionally, the modeled yearly average streamflow and total DIN export were similar to the USGS observations with R 2 values of 0.70 and 0.60, respectively (Figures S13 and S14). Unless otherwise noted, results are reported as the mean ± 1 SD throughout the following sections. Overall, from 1987 to 1996, the simulated average of total N leached in the basin were 0.0456 (±0.0227) MMT-N year −1 but no observed total N leaching data for the basin was available for comparison. Simulated and observed average DIN export were 0.0255 (±0.0155) and 0.0183 (±0.0127) MMT-N year −1 and average streamflow rates were 51.63 (±27.11) and 62.02 (±39.54) m 3 s −1 , respectively.

| Baseline: Contemporary land use and climate
For the Baseline scenario, simulated total area of corn and soybean within the RRB was 0.391 and 0.291 Mha, respectively ( Table 2). The 10-year average simulated production of corn and soybean for the basin was 3.87 and 1.02 MMT, respectively, resulting in total production of 4.89 MMT ( Table 2). Average growing season temperatures in the Baseline were 19.9°C (Section S3; Figure S7). Similar to the yields in the Historical Evaluation scenario, the Baseline scenario yields were within 1 SD of the NASS yield means, however, corn yields were slightly greater than the Historical Evaluation (Section S5; Figures S9, S10, and S15-S17).
Net mineralization rates and average N leaching rates under the Baseline scenario were 65.02 (±3.27) kg-N ha −1 year −1 and 25.05 (±7.82) kg-N ha −1 year −1 , respectively ( Figure 2 and Figure S18, Table 3). Total average simulated N leached in the Baseline was 0.0234 (±0.0096) MMT-N for the entire RRB. The highest simulated N leaching rates were located along the North and South Raccoon basin boundaries and were correlated with regions of the highest fraction of continuous corn (Figure 2 and Figure S2). Total simulated DIN export for the Baseline scenario was 0.0129 (±0.0062) MMT-N at the Van Meter location near the RRB outlet. This scenario resulted in approximately 49% less DIN export than the Historical Evaluation. Simulated streamflow for this scenario was 67.67 (±24.82) m 3 s −1 (Table 3).

| Scenario 1: Conservation practices with contemporary climate
In Scenario 1, average corn and soybean profit maps show that the most profitable land is predicted to be in T A B L E 2 Total area and 10-year average simulated production of corn, soybean, and miscanthus (Mxg) inside the Raccoon River Basin for each scenario.
the southern portion of the basin due to higher simulated yields that were not offset by coinciding land rents and input costs in that area (Section S5; Figure 3 and Figures S3, S15, S16). Across the RRB there were minimal areas that exceed both of the replacement criteria for Scenario 1 resulting in minimal introduction of miscanthus and only a slight reduction of corn and soybean area relative to the Baseline. The total area of crop production included 0.383 Mha of corn, 0.289 Mha of soybean, and 0.011 Mha of miscanthus (Table 2; Figure 4). Miscanthus yields across the basin for the 10-year average were 20.8 Mg ha −1 ( Figure S19). The total production of cropland for Scenario 1 was 5.03 MMT and consisted of 3.79, 1.02, and 0.23 MMT for corn, soybean, and miscanthus, respectively (Table 2). With the addition of miscanthus, the total area of cropland did not change, but total production increased by 3% (Table S1). When miscanthus was added to the landscape, total N leaching in Scenario 1 was reduced by 4% relative to the Baseline (Table 3 and Table S2). Miscanthus grid cells in the Baseline had total N leaching rates between 0.5 and 7 kg-N ha −1 year −1 while the corn-soybean leaching rates were 25.05 kg-N ha −1 year −1 on average (Figures 2, 4, and 5). With miscanthus replacing grid cells in this scenario, the average N leaching rate was slightly reduced to 24.04 kg-N ha −1 year −1 ( Figure 5; Table 3). Similar to N leaching, DIN export in this scenario also decreased by 4% relative to the Baseline (Table 3 and Table S2). This scenario also had a very slight reduction of 0.1% for streamflow (Table S2).

| Scenario 2: Contemporary land use with future climate
In Scenario 2, both the Mid 4.5 and Late 4.5 projections had similar growing season average temperatures (21.5 and 22.1°C, respectively; Section S3) and had similar ranges of net mineralization rates across the RRB ( Figure S18). The future climate projection with the greatest warming was the Late 8.5 projections with a growing season average temperature of 25.7°C which is +5.8°C relative to the Baseline (Section S3).

T A B L E 3
Ten-year averages of simulated net mineralization rate, total N leaching rate, total N leached, DIN export, and streamflow rates in the Raccoon River Basin for each scenario. All values are shown as average (SD).

F I G U R E 3
Corn and soybean average profitability maps based on 2019 land rent ( Figure S3), cost values, and 10-year average simulated yields (Figures S15 and S16).
The Late 8.5 projection resulted in the highest simulated soybean production totals and the lowest corn production totals. Total soybean production was 1.10, 1.15, and 1.41 MMT, while total corn production was 2.95, 2.98, and 2.81 MMT for the Mid 4.5, Late 4.5, and Late 8.5 projections, respectively (Table 2). Under the Late 8.5 Opt projection, total corn production increased from 2.81 to 4.46 MMT. Relative to the Baseline, total crop production was reduced by 17%, 16%, and 14% for the Mid 4.5, Late 4.5, and Late 8.5 projections, respectively (Table S1). Under the Late 8.5 Opt, a 20% increase in simulated total crop production was obtained relative to the Baseline.
Net mineralization rates and average N leaching rates under the Mid 4.5 projection were 74.54 (±4.98) kg-N ha −1 year −1 and 46.91 (±9.27) kg-N ha −1 year −1 and the Late 4.5 projection values were 73.26 (±4.13) kg-N ha −1 year −1 and 51.19 (±15.56) kg-N ha −1 year −1 , respectively ( Figure 2 and Figure S18; Table 3). Net mineralization rates under the Late 8.5 and Late 8.5 Opt projections were 101.99 (±6.28) kg-N ha −1 year −1 and 109.26 (±5.75) kg-N ha −1 year −1 , respectively, making this RCP projection the scenario with the highest mineralization rates ( Figure S18). Average N leaching rates under the Late 8.5 and Late 8.5 Opt projections were 57.30 (±12.73) and 34.20 (±9.36) kg-N ha −1 year −1 , respectively (Figure 2). Total N leached in Scenario 2 was highest under the late-century projections relative to the mid-century projection, except for the Late 8.5 Opt. Total average N leached in these future climate projections for Scenario 2 over the 10-year period was 0.0438 MMT-N for the Mid 4.5, 0.0478 MMT-N for the Late 4.5, and 0.0535 MMT-N in the Late 8.5 scenarios (Table 3). Total N leached was reduced to 0.0319 MMT-N under the Late 8.5 Opt. The spatial distribution of the 10year average N leaching rates for each future RCP projection resulted in the greatest simulated increase in N leaching rates to be located in the northern portion of the South RRB ( Figure 2). As climate projections move from mid-to late-century, N leaching rates gradually increase. The Mid 4.5, Late 4.5, Late 8.5, and Late 8.5 Opt leached 87%, 104%, 129%, and 36% more N than the total N leached in the Baseline, respectively (Table S2).
Similar to total N leaching, the highest DIN export was obtained under the late-century projections relative to the mid-century projections, with the exception of the Late 8.5 Opt. These future climate projections in Scenario 2 resulted in DIN export of 0.0212, 0.0254, and 0.0270 MMT-N for the Mid 4.5, Late 4.5, and Late 8.5 scenarios, respectively (Table 3). Similar to what was found with the total N leaching, DIN export was reduced to 0.0162 MMT-N under the Late 8.5 Opt. There was an increase of 64%, 97%, 109%, and 26% for the Mid 4.5, Late 4.5, Late 8.5, and Late 8.5 Opt future projections, respectively, relative to the Baseline (Table S2).
The largest average streamflow rate was 72.72 m 3 s −1 and was found in the Late 4.5 projection coincided with this projection having the largest average precipitation totals across the basin (Table 3; Figure S8). The Mid 4.5 projection resulted in the lowest average precipitation total, resulting in the lowest average streamflow rate of 47.92 m 3 s −1 . The streamflow rate for the Late 8.5 projection was 63.69 m 3 s −1 and was reduced to 57.26 m 3 s −1 under the Late 8.5 Opt. Relative to the Baseline, streamflow was reduced by 29%, 6%, and 15% under the Mid 4.5, Late 8.5, and Late 8.5 Opt projections, respectively, but increased by 7% under the Late 4.5 projection (Table S2). F I G U R E 4 Spatial map of miscanthus placement (blue) in the Raccoon River Basin for each model scenario.

F I G U R E 5
Ten-year average simulated N leaching rate for each scenario with the inclusion of miscanthus as presented in Figure 4.

| Scenario 3: Conservation practices with future climate
Average corn and soybean profit maps show the most unprofitable land was predicted to be under the Mid 4.5 and Late 4.5 projections and isolated in the counties within the south-central portion of the RRB (Figure 3). The majority of low profitability and high N leaching areas that met the miscanthus placement criteria were along the boundaries between the North and South RRB (Figure 4). Miscanthus yields projected under the Mid 4.5, Late 4.5, and Late 8.5 were 20.3, 22.0, and 26.2 Mg ha −1 , respectively ( Figure S19). The Mid 4.5 projection incorporated 0.109 Mha of miscanthus while the Late 4.5 projection incorporated 0.116 Mha, resulting in a decrease in corn and soybean area (Table 2). Total crop production under the Mid 4.5 and Late 4.5 projections was 5.63 and 5.99 MMT, respectively (Table 2). Relative to the total production in Scenario 2, where land use was held constant, there was a 39% and 45% increase in production when miscanthus was included in the Mid 4.5 and Late 4.5 projections, respectively (Table S1). This resulted in a 12% and 19% increase respectively in production relative to the Baseline.
Under the Late 8.5 projection, simulated areas of unprofitable land were reduced relative to the Mid 4.5 and Late 4.5 (Figure 3). This coincided with greater simulated soybean yields at the higher CO 2 concentration and corresponded with an increase in profitability relative to the other future climate projections. Miscanthus was placed on 0.103 Mha of land within the RRB. The total production under the Late 8.5 was 6.33 MMT, which was slightly higher than the Mid 4.5 production due to the higher productivity in soybeans simulated under the Late 8.5 projection (Table 2). There was a 50% increase in total production when miscanthus was incorporated relative to the Late 8.5 projection in Scenario 2 (Table S1). Under the Late 8.5 Opt projection, miscanthus only replaced 0.036 Mha (see Figure 4 for placement comparison) and resulted in 6.48 MMT of total crop production since total corn production increased to 4.17 MMT compared to 2.35 MMT in the original Late 8.5 (Table 2). There was only a 10% increase in total crop productivity when miscanthus was included in the Late 8.5 Opt projection relative to the corresponding projection in Scenario 2 (Table S1). Relative to the Baseline, the Late 8.5 and Late 8.5 Opt projections with miscanthus had 26% and 29% more production, respectively.
Similar to Scenario 2, the late-century projections resulted in the highest N loss totals, even with the inclusion of miscanthus. The average N leaching rate for the Late 8.5 projection was 46.28 (±11.43) kg-N ha −1 year −1 ( Figure 5; Table 3). Total average N leached under these future climate projections was 0.0335 MMT-N for the Mid 4.5, 0.0372 MMT-N for the Late 4.5, and 0.0432 MMT-N in the Late 8.5 scenarios (Table 3). Under the Late 8.5 Opt projection, the rate of N leaching and total N leached was reduced to 30.98 (±8.49) kg-N ha −1 year −1 and 0.0289 (±0.0137) MMT-N, respectively (Table 3). The rate of N leaching for the Mid 4.5 and Late 4.5 projections was 36.11 (±7.94) and 39.79 (±13.63) kg-N ha −1 year −1 , respectively ( Figure 5; Table 3). Relative to the Baseline, the Mid 4.5, Late 4.5, Late 8.5, and Late 8.5 Opt resulted in 49%, 65%, 92%, and 28% more N leached, respectively, even when miscanthus was included (Table S2). However, the total N leached from the Mid 4.5, Late 4.5, Late 8.5, and Late 8.5 Opt projections with miscanthus was reduced by 24%, 22%, 19%, and 9% relative to the respective projection without miscanthus in Scenario 2 (Table S2). All future climate projections with the inclusion of miscanthus resulted in lower total N leaching relative to the Historical Evaluation totals (Table 3).
Future climate scenarios exported 0.0157, 0.0194, 0.0214, and 0.0146 MMT-N of DIN on average for the Mid 4.5, Late 4.5, Late 8.5, and Late 8.5 Opt projections, respectively (Table 3). Relative to the Baseline, simulated DIN export increased by 27%, 56%, 73%, and 18% for the Mid 4.5, Late 4.5, Late 8.5, and Late 8.5 Opt, respectively (Table S2). Future climate projections with miscanthus resulted in a reduction in simulated DIN export by 26% in the Mid 4.5, 24% in the Late 4.5, 21% in the Late 8.5, and 10% in the Late 8.5 Opt projections relative to the export in Scenario 2 (Table S2). As for streamflow rates with miscanthus inclusion, there was a 0.2%-5% reduction relative to streamflow values from Scenario 2 (Table S2).

| DISCUSSION
This study presents the first field-scale resolution analysis that includes the effect of future climate and strategically implemented land use change determined by profitability of crop production, leached N, DIN export, and streamflow for the RRB, a significant watershed for agriculture and the key source of water to Iowa's largest city. We used the agroecosystem model, Agro-IBIS, and hydrology model with biogeochemistry, THMB, to simulate crop production and N loss and transport for multiple land use and climate projection scenarios. Miscanthus incorporation was isolated to regions of low profitability and high N loss on the current landscape in the RRB under contemporary and future climate change projections to simulate the reduction of N loss using only unprofitable land. This study provides projections for potential changes in crop productivity (i.e., total production) and water quality (i.e., N leaching, DIN export), and quantity (i.e., streamflow) under future climates and with the inclusion of miscanthus. Our results indicate that strategically implemented miscanthus provides an opportunity to reduce N loss, especially under future climates. The following sections will describe how both cropping system productivity and water quality under future climates is likely to be reduced and that the inclusion of miscanthus in the landscape may alleviate N loss while maintaining agronomic value. Additionally, a more in-depth review of the literature presented below can be found in Ferin (2020).

| Reduced cropping system productivity and water quality under future climates
The majority of N loss in a field comes from land with high mineralization rates, subsurface drainage, and ample amounts of N fertilizer Broussard & Turner, 2009;Dinnes et al., 2002;Robertson & Saad, 2013), consistent with the CMIP5 Baseline scenario having a smaller N loss than the Historical Evaluation (Table 3). This difference in N loss was likely due to different weather inputs, a difference in N fertilizer and manure application rates, N mineralization rates, and increased corn productivity under the Baseline climate conditions (Tables 1 and 3; Figures S9 and S10).
Under future climate conditions, N loss may increase due to a reduction in plant demand for N when crop productivity is reduced and net mineralization rates increase due to elevated temperatures and variable precipitation patterns ( Figure S18; Section S3). Growing season temperatures between the Mid 4.5 and Late 4.5 projections were similar and was to be expected since the RCP 4.5 emissions stabilized near mid-century. This resulted in Agro-IBIS and THMB output to be fairly similar for these two projections.
Fertilizer and manure N application rates along with contemporary land use used in the Baseline and Scenario 2 were held constant and therefore was not a driving factor of N loss between contemporary and future climates ( Figures S1 and S2). For example, the region where N leaching was the highest under all climate projections was in the county with the highest manure application rates ( Figure S1). This county also consisted of the highest fraction of continuous corn and corn-soybean area relative to the rest of the basin ( Figure S2). While the assumption of county average manure rates (Lark et al., 2022) was used in this study, we realize that manure application is not uniform across an entire county. A combination of these factors coinciding with increased mineralization rates ( Figure S18) resulted in an increase in N leached and DIN export within the RRB under the future climate projections relative to the CMIP5 Baseline (Table S2). These are a specific example of our overall results that support a portion of hypothesis 1, suggesting that cropping system productivity and water quality would be reduced under future climates.
While the warmest projection scenario (Late 8.5) resulted in the highest soybean yields and total production, this scenario yielded the highest loss of N in the RRB, likely due to the largest reduction in corn yield and production and greatest increase in net mineralization relative to the Baseline (Tables 2 and 3). By optimizing the hybrid GDD (Minoli et al., 2022) of corn to 2800°C under the Late 8.5 projection (Late 8.5 Opt), there was an increase in corn yield and productivity relative to the Late 8.5 (Table 2), which resulted in a reduction of N leaching and DIN export (Table 3) even with high net mineralization rates. This implies that the increase in corn productivity allowed for an increase in N uptake relative to the benchmark Late 8.5 projection. While this scenario provides evidence that producers could potentially adapt by selecting hybrids more appropriate for projected climate change and may reduce water quality impacts, even this idealized scenario did not reduce N leaching and DIN export levels below that of the CMIP5 Baseline, which supports hypothesis 1.
Average streamflow rates under the Mid 4.5 scenarios (i.e., largest reduction in precipitation [18%] relative to the Baseline) were reduced by 28%, and increased by 8% in Late 4.5 (i.e., highest increase in precipitation [2%] relative to the Baseline; Table S1). The future climate projections with the largest range of total precipitation (Section S3) across the RRB for the 10-year period resulted in the largest standard deviation in streamflow (Table 3). Within a single global climate forcing projection, annual average precipitation totals varied between 349 and 1265 mm across the RRB for the late-century projection. While not the main focus of this study, streamflow rates were quite variable across the future climate projections which makes predicting future streamflow rates quite difficult (Chien et al., 2013;Stone et al., 2003).

| Incorporation of miscanthus in cropping systems can reduce N pollution while maintaining agronomic value
Average profitability of corn and soybean under future climate projections varied relative to the Baseline. For instance, the Mid 4.5 and Late 4.5 resulted in lower corn yields relative to the Baseline and lower soybean yields relative to the Late 8.5, which resulted in the lowest profitability outcomes (Figure 3). These projections resulted in the largest area of predicted miscanthus integration ( Table 2). As for the Late 8.5 projection, soybean yields were the highest, and corn yields were the lowest out of all the climate projections, coinciding with previous studies (Teshager et al., 2016;Wang et al., 2015). This resulted in regions of higher average profitability across the RRB ( Figure 3) and less miscanthus placement relative to the Baseline, Mid 4.5, and Late 4.5 projections (Table 2; Figure 4). Relative to all other future climate projections, the Late 8.5 Opt simulation resulted in the greatest profitability and smallest amount of miscanthus incorporation across the RRB. While the Late 8.5 Opt simulations provide some evidence that annual grain crops could be adapted to significant shifts in climate, current progress in crop breeding has been shown to be insufficient to offset rapid changes in growing season characteristics (Zhang et al., 2022). Another key caveat to consider is that like most crop models our simulations do not include reproductive heat stress which could significantly impact yields in high temperature conditions such as those in Late 8.5 (Ferin, 2020;Heinicke et al., 2022;Thomey et al., 2019;Webber et al., 2018).
In addition to profitability, N leaching across the RRB for each scenario was also used to determine the placement of miscanthus. Since both conditions of profitability less than US$ −100 ha −1 and N leaching greater than 50 kg-N ha −1 were required criteria to replace the current land use with miscanthus, not all areas of low profit and high leaching were mitigated. However, strategically implemented miscanthus still resulted in a reduction of N leaching and DIN export under all climate projections (i.e., Scenario 3 vs. Scenario 2; Table S2). Our results show miscanthus integration resulted in a reduction of N leaching and DIN export by 4% under the Baseline climate (i.e., Scenario 1 vs. Baseline; Table S2). The reduction in N leaching shown under our Baseline conditions was similar to that of Ha et al. (2020) and Ha and Wu (2022) when switchgrass was incorporated into the landscape within the RRB. Under the future climate scenarios, N leaching was reduced by 19%-24% when miscanthus was integrated into the landscape and DIN export was reduced between 21% and 26% (i.e., Scenario 3 vs. Scenario 2; Table S2). The N loss reduction achieved under a future climate when miscanthus was used in our study was very similar to the N reduction reported in Teshager et al. (2016). While the Baseline scenario in this study only incorporated miscanthus on 2% of the current landscape, future climate projections in Scenario 3 converted 15%-17% of the landscape to miscanthus, which resulted in a 19%-24% reduction in N loss at the basin scale. This magnitude of perennial grass incorporation and corresponding N reduction is consistent with previous studies (Brandes et al., 2018;Housh et al., 2015). Both results from the Baseline climate and future climate scenarios support hypothesis 2. Additionally, our findings support the previous findings that the inclusion of perennial grasses under future climate conditions and that strategically places perennial grasses will have a greater impact on improving water quality by reducing N loss (Rajib et al., 2016;Wu & Liu, 2012). While a reduction in N loss was one benefit of miscanthus integration into the landscape, this inclusion also led to an increase in total biomass production. When miscanthus replaced row crop production on low profit and high leaching regions in the RRB, total production increased between 3% and 50% depending on the climate projection (Table S1). A previous study that looked at a range of profitability thresholds for replacing corn with switchgrass for a county outside of the RRB in Iowa obtained a 48%-99% increase in biomass production under the higher range of the corn grain prices used in the study (Bonner et al., 2014). Our Baseline scenario only obtained a 3% increase in biomass production with miscanthus, but this is likely due to the additional N leaching threshold constraining the area of available land for conversion in our study. Total production in future climate also increased between 12% and 29% when miscanthus was included depending on the projection, relative to the Baseline. Our results showed an increase in miscanthus biomass under future climate conditions, coinciding with Chen et al. (2017), which may partially be the reason for this increase in total production when miscanthus is included. These results partially support hypothesis 1 in regard to the effects of future climate conditions for corn and soybean relative to miscanthus.

| Uncertainties and future consideration
Our analysis made many assumptions on the application of N fertilizer and manure, all of which may impact the exact magnitudes of our results but are not sufficiently large to change the overall findings of our study. First, from lack of data, we assumed a constant rate of fertilizer and manure applications across each county. Future work should develop approaches that allow for variations of fertilizer application rate within a particular county, especially when using an agroecosystem model at the field scale. We also assumed that the N fertilizer and manure rates remained fixed over 10 years simulated in both contemporary and future climates. If crop productivity and N uptake decreases farmers could potentially adjust the amount of N fertilizer and manure applied to the field. Additionally, we assumed the same rate of manure application across all crop types, including soybeans. As more data on manure applications becomes available, modeling efforts should be focused on creating crop-specific manure applications especially when modeling within the Corn Belt region of the Midwest, US. While both of these simplifications may have led to inaccuracies in N loss for individual fields, it is important to note that the Historical Evaluation simulation agreed well with observations at the basin outlet (Figures S11-S14) suggesting that the approach is still valid for estimating N transport at this key location of interest.
While this study only includes a single climate projection, it is important to include multiple future climate projections in future work (Chen et al., 2017;Rajib et al., 2016;Teshager et al., 2016). Future studies should include multiple global climate model forcings from the CMIP5 or CMIP6 database. This will improve our ability to better quantify the variability in projected temperatures and precipitation rates that influence crop productivity, mineralization, leaching, streamflow, and total N loss. Additionally, this study did not include reproductive heat stress, which may become more impactful, especially in higher emission scenarios (Ferin, 2020;Heinicke et al., 2022;Jin, Zhuang, et al., 2017;Peng et al., 2018;Thomey et al., 2019;Zhu et al., 2019). Future work should address other co-benefits of perennial grasses in addition to reductions to N loss (i.e., carbon sequestration, lower N 2 O loss, etc.; Anderson-Teixeira et al., 2009;Chen et al., 2021;Davis et al., 2010) to assess the full potential of these grasses to provide ecosystem services, especially under future climate conditions. While our relatively simple profit-based land use change and water quality analysis has provided strong evidence for economic and environmental co-benefits of perennial grasses under climate change, ultimately future efforts should include the multitude of biophysical factors and socioeconomic nuances mentioned above which could potentially be achieved with integrated biophysical and economic modeling at the field scale.