Identifying Spatiotemporal Changes In Irrigated Area Across Southwestern Michigan, USA, Using Remote Sensing and Climate Data

Irrigation, which has become more common in humid regions, is the largest consumptive water use across the US and the globe. In southwestern Michigan, there has been a dramatic expansion in irrigation water use for row crops (primarily corn and soybean) in the past decade, mostly from groundwater pumping. The rapid expansion of irrigated row crops has potentially profound implications for terrestrial water balances, food production, and local to regional climate. Detailed maps of spatio-temporal changes in irrigation are essential to better understand irrigation impacts. However, accurate monitoring of irrigation area can be difficult in humid regions using remotely sensed methods due to the similarity in greenness between non-irrigated and irrigated areas in most years. Here, we use remote sensing to create annual, 30m-resolution maps of irrigated cropland by integrating Landsat and MODIS satellite products along with the PRISM climate dataset. From these data we developed spatial time series of vegetation and extreme weather indices, including novel indices we developed specifically to maximize detection of irrigation. Using these input data, machine learning classification was then performed over the region to identify irrigated crop area for each year. The resulting annual irrigation maps suggest that total irrigated area in southwestern Michigan increased by 160% from 2000 to 2017. The accuracy of the maps is assessed relative to maps created for an arid region using the same method. The maps can be integrated into hydrologic models to quantify irrigation impacts and support water resources management.