Sustaining crop yield and water quality under climate change in intensively managed agricultural watersheds—the need for both adaptive and conservation measures

The projected near-future climate (2031–2059) of wetter springs and drier summers may negatively affect agricultural production in the US Midwest, mostly through reduced aeration of the root zone due to excess soil water and frequent loss of nutrients such as nitrate (NO3-N) and total phosphorus. Several agricultural adaptations—such as adding tile drains and increasing fertilizer rates—may be deployed to mitigate potential reductions in crop yield. However, these adaptations (generally driven by economic benefits) may have a severe impact on water quality, which is already under stress due to excess nutrient runoff from agricultural fields causing hypoxia in inland and coastal waters. Here, we evaluate the crop yield and water quality consequences of such adaptations under future climate with the Soil and Water Assessment Tool in a testbed watershed located in central Illinois. We show that additional tile drains and increased fertilizers can help achieve baseline (2003–2018) corn yields but with a nearly two-fold increase in riverine NO3-N yield affecting a major drinking water supply source. However, a shift to spring-only fertilizer application may not require additional fertilizer and reduces the increase in NO3-N loss to 1.25 times above the baseline. We also show that water quality may improve (better than baseline) with conservation measures such as cover crops and switchgrass. Our findings highlight the need to develop efficient climate change adaptation and conservation strategies for sustainable agriculture and water quality.


Introduction
Both agriculture and water quality are expected to be affected by climate change.In particular, the US Midwest-which produces about one-third of global corn and soybean (Wang et al 2020)-may experience significant yield reduction with climate change (Hatfield et al 2018), challenging the 2050 global food security goals (FAO 2009, Ray et al 2013).On the other hand, water quality-which is already under stress due to excess nitrate and phosphorus runoff from agricultural fields (Michalak et al 2013)may further degrade in the future, leading to more frequent algal blooms and greater hypoxic zone formation (Sinha et al 2017, Laurent et al 2018, Rabalais and Turner 2019) in inland and coastal waters.Such an intricate causal nexus among climate, agriculture and water quality requires efficient adaptive measures to ensure the sustainability of agriculture under a changing climate.
Farm-scale adaptations are gaining momentum in the United States, with the impacts of climate change becoming more apparent (Ishtiaque 2023).For instance, Midwest farmers have been adopting several measures, such as protecting crops from frost, purchasing additional crop insurance, using new technologies (Mase et al 2017) and improving field drainage conditions (Skevas et al 2022), to deal with increasing variability of weather and climate.These (and several other) adaptive measures are more likely to increase in the future with increasing weather variability (Travis and Huisenga 2013).However, when agricultural adaptations are driven by private economic benefits (Davidson et al 2019, Skevas et al 2022) they may bring about unintended adverse environmental consequences, such as degradation of water quality.
As shown by Fezzi et al (2015), historically colder and wetter regions, which are currently less suitable for crop production, may become more favorable for agriculture in the future due to increased temperatures.Agricultural expansion in these areasgenerally supported by fertilizers-may further contribute to degradation of water quality (Tilman et al 2011).Studies on climate change are generally siloed, often ignoring such feedback between agricultural adaptations and water quality (Fezzi et al 2015).For example, in the Midwest, such adaptations may require dealing with increased spring wetness, which is likely under projected future climate (Reidmiller et al 2017, Wuebbles et al 2021).Extra spring precipitation can lead to increased soil water, which may result in unsuitable conditions for crop growth due to reduced root zone aeration, an increased risk of seedling diseases, limited operability of farm equipment and delayed spring planting that shortens the growing cycle; all these can ultimately lead to lower crop yields (Urban et al 2015).In addition, a higher water yield (WY) may frequently flush nutrients from the fields, causing nutrient stress for crops (Hatfield et al 2011).Such challenges may trigger adaptations such as adding tile drains (Arbuckle et al 2015, Skevas et al 2022) and adding more fertilizers to maintain or increase crop yields, but may further degrade riverine water quality.
Policy interventions to promote conservation measures, such as adopting cover crops and perennials, can help mitigate the water quality consequences of profit-driven agricultural adaptations.Expanding existing national and state conservation programs to enable greater cost-and technologysharing opportunities with farmers and creating new markets for crops such as perennial wheat and switchgrass (Roesch-McNally et al 2018) may be vital for improving future riverine water quality.Moreover, the economic and environmental benefits provided by switchgrass, such as high yields on marginal lands, resilience to varying climate, high nutrient use efficiency and reduction in nitrate leaching (Tulbure et al 2012, Smith et al 2013, Brandes et al 2018), may incentivize more farmers to adopt the energy grasses and improve water quality or mitigate water quality degradation caused by climate change (Yoder et al 2021).
In this study, we simulate the joint effects of potential adaptations of agricultural practices and conservation measures on riverine water quality for near-future climate (2031-2059) in a US Midwest (testbed) agricultural watershed.Specifically, we consider the addition of subsurface tile drains and increases in fertilizer rates as agricultural adaptations to changing climate to maintain crop yields and profits.In addition, we test the effectiveness of cover crops and switchgrass as potential conservation measures to mitigate the impacts on water quality.These latter practices may result in reduced production for the crop grower but may be adopted voluntarily with or without compensation from government programs.We use the Soil and Water Assessment Tool (SWAT; Arnold et al 1998) to conduct our analysis at Lake Decatur watershed, located in central Illinois, USA.

Study area: Lake Decatur watershed
About 80% of the watershed area has been used for corn and soy production between 1980 and 2022 (figure 1).The soils are rich in organic content but are generally poorly drained (Demissie and Keefer 1996)-typical of the Corn Belt region (see figure 1 of Lawrence et al 2021).In addition, with topography being relatively flat (80% of the watershed is in the slope class 0%-2%; see figure S1), extensive networks of tile drains, used to reduce waterlogging and improve soil aeration, have enhanced agricultural productivity in the region (refer to Niroula et al (2023) for more details on the watershed).However, water quality of Lake Decatur is affected by upstream agricultural activities, mostly through sediment and nutrient runoff leading to lake sedimentation, high nitrate concentrations and phosphorus deposition.Lake nitrate concentration often exceeds the drinking water quality threshold of 10 mg NO 3 -N l -1 , threatening the water security of 79 000 individuals whose domestic water comes from the lake (figure 2(b) in Keefer et al 2010).The City of Decatur drinking water treatment facility reduces excess nitrate concentration when needed, but the number of days in a year when treatrment is required has been high for several years in the past, probably due to the impact of extreme events.For instance, the concentration threshold exceeded 55 days in 2008 (a wet year) and 113 days in 2013 (following the drought of 2012; see figure S2).This pattern could occur more frequently under future climate conditions, leading to an increase in energy demand and cost for the water treatment facility (Li et al 2021a).Such water quality issues are common in tile-drained corn belt watersheds (Loecke et al 2017), whose topography, soil and hydrology are favorable for corn-soy production.As increased spring wetness is projected across the entire corn belt region, analysis based on our testbed can provide a general understanding of the impacts on water quality in the region.

Model and data
We used SWAT, a semi-distributed hydrological model, to simulate WY, nitrate (NO 3 -N), total phosphorus (TP) and crop yield (corn and soy) in the watershed.At user-defined sub-watersheds, the model simulates the hydrological response under varying weather forcing and agricultural management decisions, including crop planting/harvesting, fertilizer schedule and tile drains.The sub-watersheds are further discretized into hydrological response units (HRUs), where each unit represents a unique combination of slope, soil type and land use (Neitsch et al 2011).The hydrological and water quality components obtained at the HRU scale are aggregated at the sub-watershed scale and are routed along the stream network.
The SWAT model has been established for the Upper Sangamon River watershed (USRW), which includes the Lake Decatur watershed.The model for the USRW is described in Niroula et al

SWAT scenarios 2.3.1. Baseline scenario
In the baseline, about 98% of the annual row crop area was simulated as a two-year corn-soy rotation while the rest was simulated for corn without rotation following the general patterns of cropland in the USRW.Commonly adopted agricultural practices such as tillage and fertilizer application were included in the baseline.While the planting and harvesting dates may vary regionally, a typical growing season in the region is from late spring to mid fall (figure 2(a)).About half of the nitrogen fertilizers in the region are believed to be applied after harvest in the fall-inferred through the regional fertilizer sales data (Gentry et al 2014).Following this, we choose fall-spring split application as a representative fertilizer schedule for the watershed, where 55% of the total agronomic nitrogen requirements (224 kg N ha −1 ) were applied in the fall as diammonium phosphate (DAP) and anhydrous ammonia (AA).In spring, the remaining 30% and 15% of the total requirements were applied as preplant and sidedress, respectively.No fertilizers were applied during soy-growing years (figure 2(a)).About 65% of the watershed area was tile drained.Tile drains were adopted in the flat agricultural HRUs (with a slope of <4%) which have poor natural drainage (i.e.'somewhat poorly' and 'poorly' drainage classes).Details on baseline tile adoption are provided in table S2.In addition, to represent the existing buffer strips on 36% of Illinois stream miles (IEPA and IDOA 2015), 10 m filter strips were simulated on 36% of the randomly selected agricultural HRUs.

Future climate, addition of tile drains and increase in fertilizer rates
The watershed hydrology and nutrient loads were simulated in SWAT using future climate data obtained from all 12 GCMs considering an average CO 2 of 485 ppm for 2031-2059 (van Vuuren et al 2011, Chen et al 2019).The simulated annual, non-growing and growing seasonal average flow and load estimates across the study period were obtained from each GCM.With potential increase in future WY during spring, we assumed that current soil types with a 'moderately well-drained' drainage class may require tile drains, especially in flat regions, as an adaptation measure to reduce spring wetness and prepare for better growing conditions.Hence, tile drains were added on HRUs that had a slope of 0%-2% and a 'moderately well-drained' soil type (see table S2).This led to an increase in tile-drained HRUs to 314 from the baseline of 236 (the total watershed HRUs being 534).In terms of area, this led to a 21% increase in tile-drained areas from the baseline.Likewise, we assumed that AA fertilizer application rates are likely to increase in the future so that crop yields can be maintained at least to the baseline levels, despite increased N losses due to leaching and denitrification.Hence, AA was increased from 0% to 100% under a fall-spring split scenario to simulate the crop yield response under future climate.In addition, the spring-only fertilizer scenario was tested, where all fertilizers (224 kg N ha −1 ) were applied in the spring season (figure 2(b)) together with an increase in AA (table 1).

Adoption of conservation practices
We simulate the effectiveness of conservation practices, namely cover crops and switchgrass, in improving future riverine water quality.We chose winter wheat as a representative cover crop in our analysis.
With the increase in future winter temperatures and elevated CO 2 levels, these crops may have less freeze damage (EDF 2022), higher biomass growth and possibly a similar nitrate reduction potential to other crops such as cereal rye (Lee et al 2017).Winter wheat was planted a week after the corn harvest and terminated 5 weeks prior to the planting of soy.
Likewise, switchgrass was simulated on steeper agricultural HRUs where tile drains were not present.Switchgrass was adopted under three slope conditions, >4%, 2%-4% and >2%, resulting in 2%, 5.5% and 7.5%, respectively, of the agricultural land cover being converted to perennials.This is in reference to the recent findings (Li et al 2023) that about 4% of land cover changed to perennials under improved wastewater treatment technology could help achieve watershed system-level benefits in the USRW.The fertilizer application rate for switchgrass under a 10-year growing period was obtained from Li et al (2023).

Scenarios in combination
After the baseline (B), different scenarios combining future climate (F), tile drain addition (T), springonly fertilizer application (S), cover crops (C) and switchgrass adoption (P), were simulated in SWAT.
All scenarios used in this study are summarized in table 1. Adaptations and practices outlined in table 1 were assumed to be applied instantaneously in areas deemed suitable.Moreover, all the analysis follows the water year (October to September), with October-March representing the non-growing season and April-September representing the growing season.Also, WY, NO 3 -N and TP yield discussed here represent the estimates entering Lake Decatur, while crop yield represents the average yield across the watershed.

WY and water quality
Projected future annual average precipitation is likely to remain stable, with a slight decrease of 3% from the baseline.However, significant variability in seasonal precipitation may be realized (figures 3(a) and (b)).For instance, average summer and fall precipitation may decrease by 11% and 5%, respectively, possibly leading to a drier growing season.On the other hand, changes in winter (decrease of 8%) and spring precipitation (increase of 11%) show potential to alter the hydrology and riverine nutrient export in the non-growing season.Increases in both monthly maximum and minimum temperature (figures 3(c) and (g)) lead to an increased in annual average warming of 2.15 • C by 2059.Annual maximum temperatures may increase by 2.3 • C from baseline, mostly driven by an increase in summer temperatures, thus leading to a hotter growing season.WY in the future generally reflects precipitation, with a slight decrease in annual estimates (by 2%), and shows significant seasonal variability (figures 3(e) and (f)).WY increases with the end of winter and onset of spring, particularly during February to May (43% from baseline).However, WY decreases by 59% and 29% during summer and fall, respectively.The impact of seasonally variable WY is reflected in nutrient yields: both NO 3 -N and TP loads entering Lake Decatur increase in spring by 79% and 55%, respectively (figures 3(i) and (k)).The higher nutrient yield observed in the early months of the growing season (i.e.April and May) indicates potential flushing of nutrients during the planting period.The annual NO 3 -N yield increases by 37%, while annual TP yield decreases by 8% from the baseline (table 2).The reduction in TP, also found in Kalcic et al (2019), is mostly due to reduced WY in all seasons other than spring.

Crop yield
Under future climate, the annual average corn yield is simulated to decrease by 8% from the baseline (scenario F in figure 4(a)).This could be due to several reasons.First, under wetter springs, fertilizers applied in the previous fall could be frequently flushed to the riverine systems before substantial crop uptake, leading to nutrient-deficient soils and reduced crop yields.Second, due to excess soil water under a higher WY, the reduced oxygen in the root zone causes aeration stress and thus reduces corn yields (Boles 2013, Wang et al 2016).Lastly, the increased temperature might offset some benefits obtained from elevated atmospheric CO 2 (e.g.reduced stomatal conductance and increased photosynthesis) in C 4 species such as corn (Bassu et al 2014).The rise in temperature accelerates crop phenology, shortening its growth cycle (e.However, soy yield is simulated to increase by 9% (scenario F in figure 4(b)).As no fertilizers were

Additional tile drains
With a 21% increase in tile-drained areas from baseline, average corn yield in the watershed in the FT scenario increases by 3% from F (figure 4(a)), mostly due to a reduction in water saturation of the root zone, thereby reducing aeration stress.Under FT, tile flow increases in the non-growing and growing seasons by about 20% but flow from non-tile sources is reduced by a similar amount so that seasonal and annual WYs remain consistent across these scenarios (figure 4(c)).Soy yield under FT remains consistent with F and seems to be unaffected by additional tiles.
Under FT, tile NO 3 -N in both non-growing and growing seasons increases by 15%, leading to an overall increase in annual NO 3 -N yield by 12% relative to F (figure 4(d)).Compared to the baseline annual average, this increase would be 54% (see table 2).This is probably because the solubility of NO 3 -N and tile drains reduce saturation-induced denitrification losses while providing a rapid pathway from the crop root zone to streams (McIsaac andHu 2004, Ma et al 2023).As shown, the addition of tile drains as a possible adaptation measure may have severe environmental consequences for marginal improvement in corn yield.
However, TP yield in the FT scenario decreases by 10% from F (figure 4(e)).This is potentially due to reduced surface runoff (or increased tile flow), which reduces the overall phosphorus export.In addition, with reduced surface runoff, surface erosion decreases, leading to lower availability of sedimentadsorbed phosphorus for riverine export.However, the standard version of SWAT lacks the quantification of TP from tiles, and may underestimate the contribution of dissolved reactive phosphorus (Lu et al 2016, Bauwe et al 2019).

Increased fertilizer rates
Simulated near-future corn yields increased with increasing AA under fall-spring split application (figure 5).The historical corn yield of 10 Mg ha −1 yr −1 could be reached by at least a 50% increase in AA from the baseline.
However, increased fertilizer rates also mean more nitrate loss to streams.Increased spring wetness leads to more frequent flushing of the nutrients, which have been available in the fields since fall.For instance, with a 50% increase in AA, the NO 3 -N yield entering Lake Decatur may increase by 97% (about two-fold) from the baseline (see table 2).Such a high NO 3 -N yield could greatly elevate the lake's nitrate concentration, affecting Decatur's drinking water supply.Historically, most farmers prefer fall fertilization to avoid wetter field conditions in the spring (Gentry et al 2014), which also helps to reduce their spring workload.Fall-spring application may be more common in future, especially with increased spring precipitation leading to more inconvenient field conditions.As shown, such a fertilizer strategy could lead to severe degradation of water quality.On the other hand, a mass shift to spring-only fertilizer application could substantially improve water quality.As shown in figure 5, under spring-only application, the baseline corn yield could be obtained without any added fertilizers.This results in an increase in NO 3 -N yield by only 25% from the baseline (compared with 97% under fall-spring application).Such a shift to spring-only application could be more suitable for local water quality (Randall et al 2003) as solutes would have a shorter residence time for riverine export.However, the challenge of wetter spring field conditions remains.Improvement in farm equipment for better workability in wetter fields and the use of fertilizer stabilizers may be vital in adapting to springonly application under changing climate.

Conservation practices
Water quality may further improve with the adoption of conservation practices.Figure 6 shows the simulated response of incorporating winter wheat as a cover crop in agricultural lands.With cover crops adopted in the non-growing season (FTSC), the NO 3 -N and TP yields decrease by 47% and 27%, respectively, in the non-growing season compared with FTS.Under FTSC, the annual NO 3 -N and TP yields decline by 37% and 24%, respectively, resulting in belowbaseline estimates (figures 6(a) and (b)).The future climate seems conducive to cover crops, primarily due to rising winter temperatures and elevated CO 2 concentrations, which result in greater biomass and consequently higher nutrient uptake.Thus, adoption of cover crops may substantially improve future water quality.Our results are generally consistent with studies evaluating the effects of cover crops on water quality in agricultural watersheds (e.In addition to cover crops, switchgrass could provide more benefits for riverine water quality.Simulated responses indicate that under FTSCP1, FTSCP2 and FTSCP3 annual TP yield could be reduced by 8%,12% and 20%,respectively,from FTSC (figure 6(b)).This is potentially because switchgrass, by reducing soil erosion, reduces the availability of sediment-adsorbed phosphorus.However, improvement in nitrate-N is minimal under FTSCP1, FTSCP2 and FTSCP3, with a 0.7%, 1.6% and 2.3% decrease in nitrate-N yield, respectively, from FTSC (see tables 2 and S3 for the changes from baseline).
Switchgrass also requires annual application of nitrogen fertilizer, which may offset some of its nutrient reduction potential.Moreover, tile nitrate-N constitutes 93% of annual nitrate-N yield under future climate (see figure 4(d)), and as switchgrass was adopted in non-tile-drained areas over a smaller fraction of the watershed, its effect on reducing nitrate-N yield is minimal.However, as a greater area is converted to switchgrass, incremental improvement in riverine NO 3 -N load reductions illustrate the potential for switchgrass to reduce NO 3 -N loads (Li et al 2023).The conversion of corn and soy cropland to switchgrass would lead to lower production of corn and soy (figures 6(d) and (f)).However, it is interesting to note that the future average corn yield increases slightly by 0.5%, 0.9% and 1.2% under FTSCP1, FTSCP2 and FTSCP3 scenarios, respectively, with large variability across GCMs (figure 6(c)).The increased corn yield in 8 out of 12 GCM scenarios (table S4) suggests that the land converted from corn to switchgrass was of relatively poorer quality for corn production.Thus, converting these areas to switchgrass may seem more suitable for growers and the environment.However, such an effect was not observed for soy, which suggests that soy yields were less impacted by the quality of land and converting soy to switchgrass only affected the total soy production and not average yields for the watershed.

Limitations and outlook
In this work, we have attempted to understand the impacts of a few key adaptations to climate change on water quality and crop yields that would involve relatively minor changes to crop rotation and land use.We acknowledge the possibility of more numerous and extensive agricultural adaptations, but exploring all options is beyond the scope of this study.For instance, if corn yields stagnate or decrease, and soybean yields increase, the crop mix may shift toward soybeans, which would result in reduced application of N fertilizer.Other adaptations, such as shifting planting and harvest dates, are already occurring (especially for soybeans in the last 6 years in Illinois; see figure S3), which depends on year-to-year weather and can thus change the fertilization schedule and rates.Moreover, crop yields might change with improved crop varieties in the future (we assumed constant crop technology after 2018) and could affect water quality (Ren et al 2022).Additionally, market changes such as lower demand for transportation fuels (e.g.corn-based ethanol) due to increasing transition towards hybrid and electric vehicles may cause changes in the crops grown.Examining the water quality and economic consequences of these and other adaptations should be the subject of future research.
Although SWAT is well suited to represent a wide variety of anthropogenic activities and their impacts on watershed hydrology and water quality, there are several limitations in the model processes.For instance, the effect of increased pest and disease stresses on crops that might increase with climate change is not simulated in SWAT.In addition, improvements in water quality by conservation measures may take time to materialize due to legacy effects (Muenich et al 2016, Van Meter et al 2016); such effects are not currently simulated in the standard SWAT model.Recent advances in modeling legacy nutrients-although developed in isolation for nitrate-N (e.g.Ilampooranan et al 2019) and TP (e.g.Wallington and Cai 2023)-may help to explore the effects of legacy nutrients on water quality improvements.Future research efforts coupling these relevant tools could include a more comprehensive set of potential adaptive and conservation measures to conduct a holistic water quality assessment.

Conclusion
Near-future climate projections (2031-2059) in the Lake Decatur watershed show an 11% increase in spring precipitation and an 11% decrease in summer precipitation, leading to wetter springs and drier summers.Such seasonal variability along with increased temperatures results in a simulated 8% reduction in corn yield under static technology.Several forms of agricultural adaptation may be deployed to mitigate these potential reductions in yield.In tile-drained agricultural watersheds, such as the USRW, these adaptations may entail adding more tile drains and increasing fertilizer rates to deal with excess soil water and subsequent nutrient flushing.SWAT simulations suggest that the projected decrease in corn yield may be recovered through additional tile drains (in 21% of the watershed area) and an increase in N fertilizer (by 50% above baseline AA rates) under fall-spring application.Such adaptations, however, may have severe environmental consequences because nitrate entering the downstream lake may rise two-fold above the baseline estimates, affecting a major drinking water supply for the region.Moreover, our analysis shows that shifting to springonly fertilizer application and adopting conservation measures such as cover crops and switchgrass may significantly improve future riverine water quality; nitrate and TP yield may be reduced by nearly onefifth and two-fifths, respectively, from the baseline without affecting crop yield.These measures also exhibit high potential to mitigate the degradation in water quality caused by climate change in agricultural watersheds.Van Vuuren D P et al 2011 RCP2.6: exploring
(2023)   andLi et al (2021b).The model adequately simulated flow, sediments, nitrate-N or NO 3 -N and TP under the baseline(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018) across several monitoring points including Monticello and Decatur.With improvement in crop technology represented through increase in the crop leaf area index (Neitsch et al 2011), the model satisfactorily simulated corn and soy yield for Macon County.Details on model performance, parameter range and input dataset, including weather and agricultural management practices, can be found in Niroula et al (2023).Near-future (2031-2059) climate projections including precipitation and maximum and minimum temperature were obtained from the Coupled Model Intercomparison Project Phase 5 (Brekke et al 2013) for 12 general circulation models (GCMs; see table S1) that are widely used in climate change studies involving the United States (e.g.Cousino et al 2015, Demaria et al 2016, Chen et al 2019).The daily downscaled and bias-corrected GCM datasets were suitable for SWAT simulations.For our analysis, we chose the moderate emission scenario of Representative Concentration Pathways 4.5 (RCP4.5).RCP4.5 is deemed appropriate for the near term (Hausfather and Peters 2020, EDF 2022), which would elevate the global average temperature by 2.4 • C by mid-century and slowly decline afterward.

Figure 2 .
Figure 2. (a) Baseline management operation and schedule with fall-spring split fertilizer application for a typical corn-soy rotation (Niroula et al 2023) (AA, anhydrous ammonia; DAP, diammonium phosphate).(b) Similar to (a) but with fertilizer application only in the spring.

Figure 3 .
Figure 3. Mean monthly (a) precipitation, (c) maximum temperature, (e) water yield, (g) minimum temperature, (i) NO3-N yield and (k) TP yield shown for the baseline (water years 2004-2018) and future (water years 2031-2059) scenario, with the average computed over the simulated years.For future scenarios, the mean of 12 GCMs is shown in red, and light red indicates the variability (maximum and minimum) in mean monthly estimates across the GCMs.Annual average (b) precipitation, (d) maximum temperature, (f) water yield, (h) minimum temperature, (j) NO3-N yield and (l) TP yield shown for the baseline (water years 2004-2018) and future (water years 2031-2059) scenarios, with the average computed over the simulated years.Vertical lines on top of each bar show the variability (maximum to minimum) in annual average estimates across 12 GCMs.
g. reduced time to anthesis and crop maturity), thus depriving corn of the opportunity to adequately capture radiation and assimilate CO 2 (Bassu et al 2014, Chen et al 2019) during the growing season.Such effects of increased temperature and higher WY are represented in the SWAT model and the results are consistent with previous studies (Boles 2013, Butcher et al 2014, Chen et al 2019).

Figure 4 .
Figure 4. (a) Simulated annual average corn yield across each GCM (shown by dots) and variability across GCM estimates shown by the box plot for two scenarios: future (F) and future + added tile drains (FT).The dashed horizontal line shows the annual average baseline (B) corn yield (average of yield from 2004 to 2018).(b) Similar to (a) but for soy yield.(c) Annual average water yield (averaged across years) for B, F and FT scenarios, shown in the box plots with each dot in the F and FT scenarios indicating the GCM-wise estimate.The bar plot shows the seasonal contribution (averaged across years and further averaged across GCMs) to the total annual estimates, which is further partitioned into tile and non-tile sources.The vertical line on top of each bar shows the variability (maximum to minimum) across GCM estimates.(d) Similar to (c) but for NO3-N yield.(e) Similar to (c) but for total phosphorus (TP) yield without partitioning the contribution from tile drains.

Figure 5 .
Figure 5. Simulated response of corn and NO3-N yield to increasing anhydrous ammonia (AA) fertilizer for future climate + added tile drains (FT) scenario.For the fall-spring strategy, the simulation response is shown for 0% to 100% at 10% interval increases in AA.For the spring-only strategy, the simulation response is shown for a 0%, 6%, 12% and 18% increase in AA.The vertical lines indicate the variability in annual average corn yield (maximum to minimum) across 12 GCMs.The horizontal lines indicate the variability in annual average NO3-N yield (maximum to minimum) across 12 GCMs.

Figure 6 .
Figure 6.(a) Annual average nitrate-N yield (averaged across years) shown in box plots with each dot representing the GCM-wise estimate across several scenarios: FTS (= future (F) + added tile drains (T) + spring fertilizer application (S)), FTSC (= FTS + cover crops (C)) and perennials (FTSCP1, FTSCP2, FTSCP3).The dashed horizontal line shows the annual average baseline nitrate-N yield (average of yield from water years 2004-2018).The bar plot shows the seasonal contribution (averaged across years and further averaged across GCMs) to the total annual estimates.The vertical line on top of each bar shows the variability (maximum to minimum) across GCM estimates.(b) Similar to (a) but for total phosphorus (TP).(c) Mean of annual average corn yield (averaged across years and further averaged across GCMs) shown by symbols, with vertical lines indicating the variability (maximum to minimum) in annual average yield across GCM estimates.(d) Corresponds to (c) and shows the percentage change in annual corn production (mass) from baseline.(e) Similar to (c) but for soy.(f) Similar to (d) but for soy.

Table 1 .
Scenario simulations considered in the study.

Table 2 .
Percentage change in annual average water, nitrate-N yield, TP yield and corn yield crossing multiple GCMs from the baseline (see tableS3for variability in estimates and table 1 for scenario definitions).FTSi with 0% increase in AA fertilizers is similar to the FTS scenario. b the possibility to keep global mean temperature increase below 2 • C Clim.Change 109 95 Wallington K and Cai X 2023 Updating SWAT+ to clarify understanding of in-stream phosphorus retention and remobilization: SWAT+ PR &R Water Resour.Res.59 e2022WR033283 Wang R, Bowling L C and Cherkauer K A 2016 Estimation of the effects of climate variability on crop yield in the Midwest USA Agric.For.Meteorol.216 141-56 Wang S, Di Tommaso S, Deines J M and Lobell D B 2020 Mapping twenty years of corn and soybean across the US Midwest using the Landsat archive Sci.Data 7 307 Wang Z, Qi Z, Xue L, Bukovsky M and Helmers M J 2015 Modeling the impacts of climate change on nitrogen losses and crop yield in a subsurface drained field Clim.Change 129 323-35 Wuebbles D, Angel J, Petersen K and Lemke A M (eds) 2021 An assessment of the impacts of climate change in Illinois The Nat. Conserv. 10 B2IDB-126094 Xu G, Singh S K, Reddy V R, Barnaby J Y, Sicher R C and Li T 2016 Soybean grown under elevated CO2 benefits more under low temperature than high temperature stress: varying response of photosynthetic limitations, leaf metabolites, growth, and seed yield J. Plant Physiol.205 20-32 Yoder L, Houser M, Bruce A, Sullivan A and Farmer J 2021 Are climate risks encouraging cover crop adoption among farmers in the southern Wabash River Basin?Land Use Policy 102 105268 Yuan Y and Koropeckyj-Cox L 2022 SWAT model application for evaluating agricultural conservation practice effectiveness reducing phosphorous loss from the Western Lake Erie Basin J. Environ.Manage.302 114000