Detecting Changes in Irrigation Technology Over the High Plains Aquifer, USA, Using More Than Three Decades of Satellite Data

Agricultural irrigation constitutes the largest consumptive water use across the United States. In the High Plains Aquifer (HPA), extensive groundwater withdrawals and decreased recharge rates have led to major water level declines in the aquifer. In response, more efficient irrigation technologies have been developed and promoted to increase the longevity of available aquifer storage. Widespread adoption of new irrigation technology can alter the regional water cycle and affect water use efficiency, economic input cost, and crop yield. However, with one notable exception, there is little publicly-available farm-level data on the adoption of new technologies across space and time. In the absence of such data, predicting adoptions is a viable strategy, though these complex adoption decisions are incentivized through policy and influenced by economic and social factors. Alternately, remote sensing may be capable of detecting technology changes, provided we understand the signal of such adoptions. Here we analyze satellite-derived annual irrigation maps (AIM) at 30-m resolution from 1984 to 2017, in combination with farm-level water use and irrigation technology data collected by the State of Kansas from all of its tens of thousands of water rights holders. We identify technology transitions from the farm level data and quantify characteristics of irrigated area in fields before and after the transition using our AIM data. With these characteristics, we then train a machine learning classifier to detect each transition across the neighboring states in the High Plains Aquifer (HPA). This method could be applied beyond the HPA, to identify technology changes across the US and globally, helping to provide a basis for improved policies focused on agricultural sustainability.