This paper investigates how natural conditions affect agricultural unmanned aerial vehicle (UAV) purchase intensity and model choice in China. Using administrative subsidy transaction records, county-level weather and terrain data, city-level economic controls, and county socioeconomic characteristics, the paper combines two empirical approaches. First, a zero-truncated Poisson model with correlated random effects estimates county-year purchase counts. Results show that terrain and precipitation exhibit nonlinear associations with purchase intensity, low temperatures are negatively associated, and net price is negatively associated. Second, a conditional logit model examines top-20 model choice among family farms and individual buyers. Subsidies significantly influence model choice, but this effect varies with local conditions: precipitation weakens the subsidy effect, whereas wind and high temperatures strengthen it. Policy simulations predict changes in model shares under no-subsidy, uniform subsidy increase, and condition-targeted scenarios relative to baseline. The findings imply that UAV subsidy design should account for geographic heterogeneity. Furthermore, product-choice models that exclude non-purchasers should be interpreted as capturing reallocations among existing models rather than changes in total adoption.
