Evaluating spatial and temporal variations in sub-field level crop water demands

Abstract

Variable rate irrigation (VRI) requires accurate knowledge of crop water demands at the sub-field level. Existing VRI practices commonly use one or more variables like soil electrical conductivity, historical yields, and topographic maps to delineate variable rate zones. However, these data sets do not quantify within season variability in crop water demands. Crop coefficients are widely used to help estimate evapotranspiration (ET) at different stages of a crop’s growth cycle, and past research has shown how remotely sensed data can identify differences in crop coefficients at regional and field levels. However, the amount of spatial and temporal variation in crop coefficients at the sub-field level (i.e. within a single center pivot system) has not been widely researched. This study aims to compare sub-field ET estimates from two remote sensing platforms and quantify spatial and temporal variations in aggregated sub-field level ET. Vegetation indices and reference ET data were collected at Kansas State University’s Southwest Research Extension Center (SWREC) and two Water Technology Farms during the 2020 corn growing season. Weekly maps of the Normalized Difference Vegetation Index (NDVI) and the Soil Adjusted Vegetation Index (SAVI) from aerial imagery are combined with empirical equations from existing literature to estimate both basal and combined crop coefficients at a 1-meter resolution. These ET estimates are aggregated to a 30 m resolution and compared to the Landsat Provisional Actual ET dataset. Finally, actual ET estimates from aerial images were aggregated using k-means clustering and stationary variable speed zones to determine if there is enough variation in actual ET at the sub-field level to build variable rate irrigation schedules. An equivalence test demonstrated that the aerial imagery and Landsat data sources produce significantly different crop coefficient estimates. However, the two datasets were moderately correlated with Pearson’s product-moment correlation coefficients ranging from -0.95 to 0.86. Both the aerial imaging and Landsat datasets showed high variability in crop coefficients during the first 5-6 weeks after emergence, with these coefficients becoming more spatially uniform later in the growing season. These crop coefficients may help irrigators make more informed irrigation management decisions during the growing season. However, more research is needed to validate these remotely sensed ET estimates and integrate them into an irrigation decision support system.