DEVELOPMENT OF AMODEL TO PREDICT SOYBEAN POD COLOR DISTRIBUTION

The beanleaf beetle (BLB) is one of the most damaging soybean insects in the midwestern US. Zeiss (1994) found that the second generation BLB population prefers to feed on pods, and developed information relating BLB feeding preference to pods having different colors, or maturity levels. The objective of this work was to develop a model to compute pod color distribution, which could be used to determine food source availability and economic treatment thresholds for the second generation BLB. A model was developed to compute the color distribution of soybean pods based on the hypothesis that pod color is related to pod physiological age. The model consisted of a set of coefficients that divided pods into different color classes based on the physiological age of individual pods. The CROPGRO-Soybean growth model was used to compute the distribution of pods in different age classes. The color model then computed the pod color distribution based on the physiological age distribution computed by the soybean model. The model was calibrated using six experiments conducted from 1991 to 1993 near Ames, Iowa. The model was first calibrated to predict percent of pods in three color classes, which corresponded to pod colors representing three maturity levels which had different preferences by BLB. The root mean square error (RMSE) between predicted and measured percent of pods in each color class ranged from 10.2% to 15.3% for the three color classes. The model was then calibrated to predict percent of pods in seven color classes. In this case, the RMSE between predicted and measured percent of pods in these color classes ranged from 8.1% to 26.5%. The model generally gave poorer predictions for young pods than for older pods. The shape of the predicted pod color distribution curves were similar to measured pod color distributions in all cases and displayed a distinct transition between color classes. However, relatively large errors resulted in some experiments because the timing of the predicted pod color distributions did not agree with the measured distributions.