A combined simulation-optimization approach for predicting crop yields

Abstract

The goal of predicting crop yields is shared by a broad range of interest groups ranging from farmers to crop processing companies, investors and governmental policy makers. Through many recent research efforts, models of crop growth have been developed for simulating the yield of crops under various environment and management conditions. The purpose of this paper is to investigate the possibility of using one such crop model to improve yield predictions as the season progresses and the crop realizes a specific time sequence of conditions. Because the model does not include every site-specific condition, a combined simulation-optimization procedure was developed to determine if yield predictions could be improved through on-line adjustment of parameters based on monitored crop variables at the site. A Fibonacci search algorithm was coupled with a soybean crop growth model for adjusting several important parameters to minimize error between field sample and simulated data. Results from two locations indicate that improvements in yield forecasts during a season can be significant with as few as four sample dates.