Mapping annual irrigation extent at 30-m resolution across the United States, 1997-2017

Improved nationwide understanding of the spatiotemporal trends in irrigated croplands is urgently needed to improve assessment and monitoring of water resources and help navigate rising challenges related to regional and local water scarcity, increasing demands of agricultural production, and environmental sustainability. Despite the urgent need, substantial gaps in our understanding of the spatial and temporal distribution of irrigated agriculture and crop water use remain, hindering efforts to precisely estimate agricultural water use. We addressed this data gap by extending our recently developed methodology to generate annual 30-meter resolution datasets of irrigated agricultural extent across the conterminous U.S. for each year in 1997-2017. To map annual irrigation distribution, the entire period was divided into two eras: USDA NASS census years (i.e., 1997, 2002, 2007, 2012, and 2017) and other years. For the census years, we derived two sets of training data; the first (automatically generated based on vegetation greenness) was used in the western U.S., while the second dataset (manually collected via visual interpretation of very high-resolution imagery) emphasized coverage of the more humid eastern U.S. These census-year samples were then temporally extended to provide annual training data for the full study period. The input variables for irrigation vs. non-irrigation classification consisted of remote sensing features, climate variables, and soil texture. Our evaluation for the 2012 irrigation map had an overall accuracy of ~90% at the aquifer level, which provides a significant improvement over previous datasets. Initial results of mapping annual irrigation extent and its changes over time in the Ogallala Aquifer for the period of 2000-2017 also produced irrigation estimates consistent to USDA-NASS county-level census statistics, suggesting the approach holds promise for characterizing broad-scale trends in irrigation while also capturing critical fine-scale details in spatial and temporal dynamics. The preliminary results revealed overall increasing levels of irrigation over the Ogallala Aquifer for 2000-2017, primarily introduced for expanded production of corn and soybeans, although many more-localized trends and patterns exist throughout the region and nationwide.