Evaluation of Evaporative Stress Index Response in US Cropland and Grazingland Areas

The Evaporative Stress Index (ESI) is an estimation of vegetation water stress, representing the standardized anomaly of actual to reference evapotranspiration (ET), which is quantified by a surface energy balance model driven by infrared remote sensing data. Actual ET is estimated by the Atmosphere–Land Ex-change Inverse (ALEXI) model using thermal infrared observations of Earth's surface from the Geostationary Operational Environmental Satellites(GOES), meteorological data from the North American Regional Reanalysis (NARR), and leaf area index estimation from Terr MODIS. In this study, ALEXI is run on a4 km spatial resolution, and can be downscaled to 30 m using Landsat imagery. ESI is calculated over the continental United States (CONUS) for compositing intervals of two, four, eight, and twelve weeks. Previous research has found ESI to respond quickly to flash drought, as water stressed vegetation limits its transpiration, thus increasing the canopy temperature.

In this study, we investigate the time behavior of ESI derived over a range in spatial resolutions (30 m to 4 km) over the USDA Agricultural Research Service Long-Term Agroecosystem Research (LTAR) network and grassland sites involved in a regenerative grazing experiment to evaluate vegetation's response to climate and management drivers. Analysis of ESI also holds utility in monitoring yield/productivity. At all sites, micrometeorological flux towers and rigorous biophysical sampling protocols are employed.

The first goal is to evaluate ESI at multiple spatial and temporal scales in comparison with other vegetation health indicators, the US Drought Monitor (USDM), and National Agricultural Statistics Service (NASS) reports. The USDM is a weekly composite product that considers precipitation, soil moisture, land surface temperature, aridity, stream flow, and other environmental variables. Our validation of ESI to USDM and NASS reports is completed by the linear temporal correlation. ESI products to test are those based on just solar radiation, known as FSUN, or those based on estimated ET, known as FRET. We will compare performance of ESI products generated from actual ET normalized by reference ET and solar radiation, and will assess responsiveness as a function of compositing window length (2- to 12-weeks). As found by previous studies, we expect FRET to show more agreement to other products than FSUN during cross validation. Also we expect 2 week and 4 week products to show a stronger response to flash drought than 8 week and 12 week products, but with a noisier signal.

We also explore the impacts of temporally adjusting normalized ET timeseries used in anomaly computations to align with phonological stages of the crops at our study sites. This requires knowledge of development stage, either observed at field (in situ) or county scale (e.g., from NASS) or derived through remote sensing. Additionally, scaling within-season timescale using degree days vs. calendar days is explored. These adjustments are aimed at reducing false ESI anomalies due to year-to-year variability in phenological development rather than vegetation stress. We also identify periods of maximum sensitivity of different crops to moisture stress. We expect the greatest loss of production to occur with water stress around flowering / tasseling for most crops. This study could be up-scaled to cover the entire CONUS, with proper aggregation of USDM and NASS data.