Considerations of Meteorological Time Series in Estimating Regional-Scale Crop Yield

Abstract

The sensitivity of simulated soybean yield to spatial averaging of meteorological data was analyzed for the central United States during a 23-year period. Regional yield was simulated using the physiological model, SOYGRO, in two sets of experiments. In the first set, yield was simulated using meteorological data at individual stations within grid cells ranging from 2° latitude ×2° longitude to 5° latitude ×5° longitude. In the second set, the daily meteorological time series were adjusted through spatial averaging over grid cells. Spatial averaging caused bias ranging from 18% in 2° latitude ×2° longitude grid cells to 28% in 5° latitude ×5° longitude grid cells when averaged over the study period. During individual years such averaging caused bias exceeding 80% of simulated yield. While spatial averaging provides a means of characterizing regional-scale climate, and has been used with empirical crop yield models, the sensitivity of physiological models to the timing of meteorological events requires more reliable daily input values. The precautions presently exercised in most impact studies using physiological models will be justified until the quality of climate change scenarios improves.