Irrigation boosts crop productivity and stabilizes yields amid weather variability, making it a vital tool in global food systems. Irrigation extent can be dynamic, declining in some areas as water resources are depleted, while expanding in regions with natural water availability and economic incentives for supplemental water. In recent years, irrigation has been expanding in the traditionally rainfed and temperate US Corn Belt, a major global producer of maize and soybeans. Although still relatively rare on the landscape, it is important to quantify the drivers and effects of Corn Belt irrigation as farmers adapt to changing climate and economic forces. Understanding this evolving agricultural system will allow us to better predict future changes to production, farm economics, and regional water hydrology.

Here, we analyze remotely-sensed annual maps of field-level irrigation expansion across the US Corn Belt since 2000 to identify correlative factors in irrigation expansion. We then combine these with remotely-sensed maps of annual field-level crop yields to quantify yield impacts, using a machine learning causal inference approach to isolate heterogeneous treatment effects based on climate stresses. We find that historically, Corn Belt irrigation provides ~10% yield increase on average and nearly doubles yields in adverse weather conditions. We then analyze output from dynamically downscaled future climate projections and find substantial increases in yield-limiting vapor pressure deficit by mid-century, suggesting that the incentives for, and benefits of, expanded irrigation will increase over the next several decades. We conclude that substantial irrigation expansion is likely to continue in this region and that more study is needed to inform policies promoting sustainable growth in irrigation practices.