Estimating changes in center pivot irrigation in the High Plains Aquifer using a hybrid GIS-remote sensing deep...

Since the introduction of center pivot irrigation (CPI) technologies in the 1950s, there has been a precipitous increase in irrigated cropland in the U.S. High Plains. While CPI has helped "drought proof" agriculture operations, thereby enhancing food security and permitting the expansion of agriculture operations in this semi-arid, drought-prone region, it has also contributed to groundwater depletion, which now threatens the very livelihoods of those who have become dependent on these technologies. Further compounding the issue, groundwater is a shared resource often underlying numerous jurisdictional boundaries and governed by a patchwork of regulations. It is therefore incumbent to develop spatially explicit models of CPI changes over time. However, given both the geographic extent and spatial and spectral variations of the phenomena, traditional photogrammetric and remote sensing methods to detect CPI dynamics have been inadequate or difficult to replicate. To address this, we utilized a hybrid GIS-remote sensing approach that combines multi-year Landsat imagery extracted from Google Earth Engine with Esri's deep learning framework for ArcGIS to rapidly estimate CPI extent with the High Plains Aquifer of the U.S. Great Plains for every other year from 2001 to 2023. Our results show an overall trend of increasing CPI throughout the region, with a net increase from ~ 3,779,777 ha under CPI in 2001 to ~ 5,293,245 ha in 2023 (+ 40% increase), with similar patterns emerging at both the state and county levels. Our method is practical, replicable, and extensible with readily available imagery datasets and deep learning tools, and accurately estimated CPI (~ 89.79% overall accuracy on average) in the region, providing a framework for continued monitoring of CPI changes.