Connections between the hydrological cycle and crop yield in the rainfed U.S. Corn Belt

Water stress is one of the major abiotic stresses and directly affects crop growth and influences crop yields. To better quantify the responses of crop yield to hydrological variability in the rainfed Corn Belt of the United States (U.S.), we analyzed the relationships between corn/soybean yield and hydrological cycle metrics, as well as their spatio-temporal dynamic at the agricultural district and interannual scale between 2003 and 2014. We used Partial Least Square Regression (PLSR) to optimally integrate different hydrological metrics and drought indices to define a crop-specific new drought index that uses crop yield as the target, and investigated the contributions of those hydrological cycle components to the new drought index. We used both observed and modeled hydrological cycle metrics, as well as several drought indices in this study, including evapotranspiration (ET) and potential ET (PET), terrestrial water storage change (ΔS), surface soil moisture (SSM), river discharge (Q), Standardized Precipitation-Evapotranspiration Index (SPEI), Palmer Drought Severity Index (PDSI), fET (the ratio of ET to PET), and vapor pressure deficit (VPD). Our results revealed that: (1) VPD, SSM, and fET showed the strongest correlations with crop yield, among the observation-based hydrological cycle metrics and drought indices considered here. Most of the hydrological cycle metrics and drought indices showed similar seasonal correlation patterns with crop yield, and this pattern revealed that the sensitivity of crop growth to water stress peaked in July for corn and in August for soybean in the rainfed U.S. Corn Belt. (2) The drought in 2012 started with higher water demand (reflected in abnormally high ET, PET, and VPD) and lower water supply (reflected in abnormally low P), followed by soil water depletion (as revealed in SSM and ΔS), leading to massive crop yield losses due to increased constraints on both water supply and demand. (3) The R² of the PLSR-based crop yield model reached 0.76 and 0.70 for corn and soybean, respectively. For both corn and soybean, the first PLSR component was mainly composed of information from VPD, fET, ΔS and SSM, indicating atmospheric water deficit and soil water storage both play critical roles in quantifying corn and soybean yield loss due to water stress.