Irrigation benefits outweigh costs in more US croplands by mid-century

The study investigates whether and where expanding irrigation for maize and soybean across the United States will be economically viable and hydrologically sustainable under climate change through mid- and end-of-century. With U.S. temperatures projected to rise ~1.4 °C and global food demand increasing 35–56%, agriculture faces heightened climate risks and yield declines already attributed to warming. Irrigation is a prominent adaptation option, yet it is the largest consumer of freshwater, has reduced environmental flows in many regions, depleted aquifers, and emits greenhouse gases. Uncertainties in future precipitation, irrigation requirements, marginal yield responses, and water availability complicate planning. Prior assessments offer mixed conclusions for the U.S. Midwest, ranging from little added irrigation need to substantial increases, depending on climate projections and crop model assumptions. This study aims to quantify, at high spatial resolution, the future economic return of irrigation (benefit-to-cost) and its sustainability given groundwater recharge, thereby informing where irrigation expansion could be both profitable and feasible.

Existing work has: (1) documented climate-induced yield reductions and slowed productivity growth; (2) identified irrigation as a potential adaptation to meet rising food demand; (3) shown that irrigation has stressed environmental flows and depleted groundwater; (4) projected regional declines in irrigated water applied due to resource limits or demand changes; and (5) highlighted large uncertainties around future precipitation, irrigation needs, and marginal yield responses across models and scenarios. Early studies suggested Midwestern productivity could persist with minimal irrigation through mid-century, while more recent research indicates substantial added irrigation may be required absent anomalous historical cooling patterns. Advances in CMIP6 projections and process-based crop models (e.g., growth stage deficit irrigation) can reduce uncertainty. Global analyses have mapped present marginal irrigation yield gains and gross economic value of irrigation water and delineated where water supply may meet future crop demand, but none have projected the future value of irrigation water at high resolution for the U.S. Combining economic valuation with water availability estimates is critical to identify areas that are both profitable and sustainable for irrigation.

Study design: The authors simulate maize (CERES-Maize) and soybean (CROPGRO-Soybean) with pDSSAT (parallelized DSSAT via pSIMS) on a 5-arcminute grid over all historically cultivated U.S. lands for three 30-year periods: historical (1981–2010), mid-century (2036–2065), and end-of-century (2071–2100). They drive pDSSAT with a 5-member CMIP6 ensemble representing the range of midwestern futures (cool/wet, cool/dry, hot/wet, hot/dry, and inter-model median) under SSP245 and SSP585. GCMs are selected by an envelope-based approach using projected Midwest changes in temperature and precipitation and a historical skill score in temperature and precipitation, then statistically downscaled by a delta change method (additive for temperature; multiplicative for precipitation and solar radiation) to the Livneh 1/16° grid. Crop model setup and calibration: Inputs include daily TMAX/TMIN/precip/solar radiation (Livneh), detailed soils, and management. Historic cultivar parameters were estimated from USDA NASS crop progress dates; maize phenology parameters p1 and p5 were derived from accumulated GDDs between phenological stages, with other parameters (e.g., g2, g3, PHINT) following published values. Soybean maturity groups were mapped by accumulated photothermal time. Planting density (NASS) and select physiological parameters were tuned to match county yields (NASS) via iterative calibration. Soil Level Productivity Factor (SLPF) was calibrated per county (1991–2000) to minimize RMSE between observed and simulated yields; evaluation used 2001–2010. Adaptive management in futures: Planting dates for maize and soybean were predicted by a random forest trained on historical planting dates and early-season climate metrics (GDDs, SPEI, temp, precip) plus state location; it reproduces maize planting dates well (r²=0.88). Maize cultivar length (p1, p5) was scaled with future GDDs; soybean cultivars were held constant due to wide maturity bands. Irrigation representation: Irrigated simulations apply 10 mm irrigation automatically on days when plant available water falls below 40%, from an unlimited source; fallow periods between harvest and next planting prevent soil re-initialization biases. Economic metric (IrB/C): For each grid cell-year-crop, the irrigation benefit-to-cost ratio IrB/C is computed as (marginal yield gain × constant 2000–2020 average world market price: maize $160 t−1, soybean $365 t−1) divided by pumping costs plus annualized ownership cost, minus 1. Ownership is based on a central pivot over a ¼-mile field with a 20-year life; values synthesized from extension sources. Pumping cost uses regional electricity prices (EIA) and energy required to lift seasonal pumped volume from total dynamic head: static water table depth plus drawdown from cone of depression and within-well effects, and pressure head; efficiencies follow literature. Contemporary depth-to-water and transmissivity (Zell & Sanford) inform lift; seasonal drawdown is computed from pumping rate and aquifer properties. Future starting water table depths are held at contemporary levels due to uncertainty in withdrawals. Sustainability metric (groundwater deficit): Annual irrigation groundwater deficit = simulated recharge (from pDSSAT rainfed runs’ subsurface drainage) minus “other uses” (county-level consumptive industrial and residential use, held constant), minus irrigation water applied. Positive values indicate surplus; negative values indicate deficit. Analyses highlight where IrB/C is frequently positive and where surplus water could support sustainable irrigation. Validation and checks: Historical simulated yields correlate with county averages (maize r²=0.59; soybean r²=0.57). Simulated applied water aligns with USDA 2013 irrigation survey (maize r²=0.88; soybean r²=0.81). Estimated pumping costs approximate survey values (RMSE $68 ha−1). Sensitivity tests explore cultivar assumptions and ownership cost ranges.

Climate projections: Midwestern growing season temperatures rise by ~1.1–4.6 °C by mid-century and ~1.9–8.8 °C by end-century (across SSPs and GCMs). Precipitation changes are uncertain: mid-century changes range from +10.2% (NESM3, SSP245) to −7.9% (HadGEM3, SSP585); by end-century from +12.3% to −16.9%. Yields: Under SSP245, mid-century rainfed and irrigated maize and soybean yields increase across most of the Corn Belt and eastern U.S., with decreases in the southern High Plains. By end-century, maize (and to a lesser extent soybean) declines in the southern High Plains, while Midwest yields remain above present on average under SSP245. Under SSP585, end-century maize suffers widespread reductions; irrigated soybean yields remain above present in most areas except the southern High Plains, Mississippi, and parts of GA/SC. Marginal yield: Soybean marginal yield increases substantially by mid-century; maize marginal yield rises in parts of the western and north-central U.S., including Michigan, but slightly declines in some currently irrigated areas of KS/OK. Economic returns (IrB/C): Historically, High Plains exhibit the highest average annual IrB/C. Historical mean IrB/C values: maize—Corn Belt 0.83, High Plains 0.76, Eastern irrigated 0.89; soybean—Corn Belt 0.37, High Plains 0.15, Eastern irrigated 0.54 (Table 1). By mid-century, IrB/C increases throughout the north-central U.S. and upper Midwest, especially for soybean; in many currently rainfed midwestern areas, average annual soybean IrB/C becomes positive by mid-century (SSP585) or by end-century (SSP245). Under SSP585, end-century maize IrB/C is negative over most of the U.S. Risk mitigation: The frequency of positive annual irrigation returns increases. In the Corn Belt (SSP245 mid-century), maize positive return frequency rises from 14% to 31%, and soybean from 14% to 41%. During the driest years, IrB/C exceeds ~1.5 over nearly all currently irrigated cropland. Water sustainability (groundwater deficit): Large average annual irrigation water deficits persist in the southern High Plains and California’s Central Valley. Historically there is limited overlap between water surplus areas and areas with routine positive returns, except for soybean in the Mississippi Valley and parts of the Upper Midwest. By mid-century, overlaps of positive IrB/C and water surplus emerge in northern Michigan and Wisconsin (maize and soybean), and for soybean in parts of the Corn Belt, Mississippi Valley, and Mid-Atlantic. Closing 30-year average local water deficits in the southern High Plains could require up to a 100% reduction in applied irrigation. Despite higher pumping costs, marginal profits could sustain pumping from depths likely exceeding many aquifers’ saturated thicknesses (indicative of potential unsustainability). Regional implications: Economic pressure for irrigation expansion intensifies in MT/ND/SD by mid-century (high IrB/C for both crops) despite limited groundwater to support sustainable expansion. Soybean’s higher productivity and lower water demand make expanded, sustainable irrigation more feasible across much of the eastern U.S. than maize.

The findings show that while average rainfed yields in the Corn Belt may remain stable or improve modestly through mid-century, climate change increases the value of having irrigation capacity as a hedge against more frequent droughts. Irrigation profitability expands into parts of the north-central and eastern U.S., particularly for soybean, but sustainability is constrained by local groundwater recharge. In the High Plains, continued positive irrigation returns coupled with water deficits imply economic incentives may perpetuate unsustainable groundwater extraction unless policy intervenes. Conversely, some regions in the northern Midwest and parts of the East show both positive returns and water surpluses, suggesting feasible and sustainable irrigation expansion—more so for soybean than for maize due to soybean’s lower water demand and temperature sensitivity. These dynamics could influence future crop rotations, market structures, and insurance strategies. The results support nuanced, region-specific irrigation policies that integrate local water availability, economic incentives, and risk management objectives.

By mid-century, marginal yield gains from irrigation and resulting IrB/C values increase across the Dakotas, northern Michigan, Wisconsin, and parts of the eastern Texas region, potentially incentivizing new irrigation installations and adding pressure on groundwater. In the Corn Belt, average annual rainfed maize and soybean yields are likely to remain at or above present levels through mid-century, but more frequent droughts will make access to irrigation increasingly valuable in dry years. Sustainable irrigation expansion is primarily limited by water availability: soybean irrigation could be both feasible and profitable across multiple regions of the eastern U.S., while viable, sustainable expansion of maize irrigation is more constrained, notably to northern Michigan and Wisconsin. Policymakers should account for local water demand and availability when guiding irrigation installation and use, as economic incentives alone may drive unsustainable groundwater use in deficit regions. Overall, irrigation remains an essential component of U.S. agriculture under climate change, with soybean offering a more favorable balance of productivity, water use, and profitability in many prospective expansion areas.

Key limitations include: (1) Static groundwater depth at the start of each growing season; the framework does not simulate feedbacks from changing withdrawals on water tables, so results isolate climatological impacts given present availability. (2) Simplified irrigation management (fixed 10 mm applications when plant available water <40%) and no exploration of deficit irrigation strategies that could conserve water with limited yield penalties. (3) Limited representation of future adaptations (e.g., cultivar changes, management, technological innovation, policy incentives), with soybean cultivars held constant and maize cultivar length scaled only by GDD changes. (4) No coupling of irrigation with regional climate feedbacks, which can affect local temperature and precipitation. (5) Single-crop-model approach (pDSSAT) with inherent biases; multi-model crop ensembles were not used. Additional simplifying assumptions include constant crop and energy prices, constant industrial/municipal consumptive use in the future, ownership cost generalizations for irrigation systems, and the use of a delta-change downscaling method.