SENSITIVITY ANALYSIS OF STATISTICAL CROP-WEATHER MODELS

A sensitivity analysis of statistical crop-weather models has been performed. The models considered in this study are of the type developed by Thompson, which predict yields from climatic variables using empirical relations derived from historical yields and weather. Ridge regression is used to perform the sensitivity analysis. The results of this sensitivity analysis indicate that the estimated coefficients for these models can be quite variable. These results have significant implications concerning: (1) the appropriate statistical methodology for developing yield models; and (2) the limitations inherent in using these models to assess the impact of climatic variability or change on food production. © 1979.