An Iterative Regression Model for Estimating Soybean Yields from Environmental Data

Abstract

A model was developed for using weather data, to estimate the yields of soybeans for varieties adapted to the central United States. The model utilized an iterative regression analysis for relating soybean yields to environmental variables. This technique evaluated the simple and interacting contributions to soybean yield of environmental variables in terms of a time scale related to soybean development (biometeorological time). The environmental variables tested were daily climatological data (rainfall and maximum and minimum air temperatures), derived agrometeorological variables (actual and potential evapotranspiration) and a soil moisture index. The maximum air temperature, potential evapotranspiration and soil moisture index accounted for more of the variability in soybean yields (coefficient of determination of 0.75) than other combinations of the tested variables. For verification of the model, a sample of 20 yields were withheld from the iterative regression analysis and comparisons were made between the yields simulated from the regression equations and the observed yields. The mean difference of 0.98 q ha?1 between observed and estimated yields for the twenty cases did not differ from zero by a statistically significant amount. The standard error of estimates was 4.79 q ha?1. Although this precision provides estimates of field yields which may be used for many practical purposes, the low correlation between the observed and estimated yields for the test cases indicates the need for caution in using this type of analysis.