U.S. farmers have reallocated land across crops on an extraordinary scale over the past thirty years, and how much of this reflects climate adaptation matters for the welfare cost of warming. A farmer who switches from cotton to soybean may be chasing higher expected returns or fleeing rising downside risk. The two motives carry very different welfare and policy implications but reduced-form acreage-share methods cannot tell them apart. We build a structural land-use framework in which risk-averse farmers choose among crops by maximizing CARA expected utility over believed non-Gaussian per-acre profit distributions. The believed distributions are constructed from a thirty-year rolling weather belief and a Just–Pope yield model. The acreageshare system estimates a behaviorally disciplined distribution of farmer absolute risk aversion from observed land shares, rather than calibrating it from surveys. Federal crop insurance enters explicitly through its indemnity-and-premium structure. The framework lets us decompose climate-driven between-crop reallocation into a productivity channel and a risk channel, evaluate the welfare value of letting farmers re-optimize across crops, and assess subsidized federal crop insurance as a climate-adaptation policy. We apply the model to U.S. county-level land-allocation data 1987–2022 and project a CMIP6 SSP2-4.5 climate forward to mid-century. Two findings emerge. Between-crop reallocation absorbs most of the welfare cost of climate change the model assigns. About one-third of the projected reallocation reflects substitution away from worsening downside risk rather than toward higher expected returns — a channel reduced-form acreage methods cannot identify — and the same one-third risk share appears retrospectively in the actually-observed 1990–2020 reallocation.
