Soil moisture deficiency is a major factor in determining crop yields in water-limited agricultural production regions. Evapotranspiration (ET), which consists of crop water use through transpiration and water loss through direct soil evaporation, is a good indicator of soil moisture availability and vegetation health. ET therefore has been an integral part of many yield estimation efforts. The Evaporative Stress Index (ESI) is an ET-based crop stress indicator that describes temporal anomalies in a normalized evapotranspiration metric as derived from satellite remote sensing. ESI has demonstrated the capacity to explain regional yield variability in water-limited regions. However, its performance in some regions where the vegetation cycle is intensively managed appears to be degraded due to interannual phenological variability. This investigation selected three study sites across the U.S. Corn Belt – Mead, NE, Ames, IA and Champaign, IL – to investigate the potential operational value of 30-m resolution, phenologically corrected ESI datasets for yield prediction. The analysis was conducted over an 8-year period from 2010 to 2017, which included both drought and pluvial conditions as well as a broad range in yield values. Detrended yield anomalies for corn and soybean were correlated with ESI computed using annual ET curves temporally aligned based on (1) calendar date, (2) crop emergence date, and (3) a growing degree day (GDD) scaled time axis. Results showed that ESI has good correlations with yield anomalies at the county scale and that phenological corrections to the annual temporal alignment of the ET timeseries improve the correlation, especially when the time axis is defined by GDD rather than the calendar date. Peak correlations occur in the silking stage for corn and the reproductive stage for soybean – phases when these crops are particularly sensitive to soil moisture deficiencies. Regression equations derived at the time of peak correlation were used to estimate yields at county scale using a leave-one-out cross-validation strategy. The ESI-based yield estimates agree well with the USDA National Agricultural Statistics Service (NASS) county-level crop yield data, with correlation coefficients ranging from 0.79 to 0.93 and percent root-mean-square errors of 5–8%. These results demonstrate that remotely sensed ET at high spatiotemporal resolution can convey valuable water stress information for forecasting crop yields across the Corn Belt if interannual phenological variability is considered.

在缺水的农业生产地区，土壤水分不足是决定作物产量的主要因素。蒸发蒸腾（ET）由蒸腾作用的作物用水和直接土壤蒸发的失水组成，是土壤水分供应和植被健康的良好指标。因此，ET已成为许多产量估算工作不可或缺的一部分。蒸发胁迫指数（ESI）是基于ET的作物胁迫指标，它描述了从卫星遥感推导的标准化蒸散量中的时间异常。ESI证明了有能力解释缺水地区的区域单产变化。然而，由于年际物候变化，其在某些密集管理植被周期的地区的性能似乎有所下降。这项调查选择了美国玉米带的三个研究地点（米德，东北，艾姆斯，爱荷华州和伊利诺伊州尚佩恩），以调查30毫米分辨率，物候校正的ESI数据集进行产量预测的潜在运营价值。该分析是在2010年至2017年的8年中进行的，其中既包括干旱条件，又包括干旱条件以及单产的广泛范围。玉米和大豆的减趋势异常与通过年度ET曲线计算的ESI相关，ESI是根据（1）日历日期，（2）作物出苗日期和（3）生长度日（GDD）缩放的时间轴按时间排列的。结果表明，ESI与县域范围内的产量异常具有良好的相关性，并且对ET时间序列的年度时间对齐进行物候校正可以改善相关性，特别是当时间轴是由GDD而非日历日期定义时。当玉米作物对土壤水分缺乏特别敏感时，玉米的丝绸化阶段和大豆的生殖阶段就出现峰值相关性。峰相关时导出的回归方程用于使用留一法交叉验证策略来估计县级产量。基于ESI的单产估算值与USDA国家农业统计局（NASS）县级农作物单产数据非常吻合，相关系数范围为0.79至0.93，均方根误差为5–8％。