|Hoogenboom, Gerrit - UNIV OF GEORGIA|
|Herndl, Markus - UNIV OF HOHENHEIM GERMANY|
Submitted to: Biological Systems Simulation Group Proceedings
Publication Type: Abstract Only
Publication Acceptance Date: March 15, 2007
Publication Date: April 15, 2007
Citation: White, J.W., Hoogenboom, G., Herndl, M., Mcmaster, G.S. 2007. Regression-Based Analysis of Modeled Responses to Water Deficits in Wheat and Common Bean. Biological Systems Simulation Group Proceedings, pp. 29-30. Technical Abstract: A key issue in modeling crop response to water deficits or other environmental factors is how best to assess model performance. Simple comparisons of observed vs. simulated data often provide little information beyond the trivial conclusion that growth increases with available soil water. Multiple regressions have been used to extend conventional comparisons of observed vs. simulated responses by including the effects of locations, cultivars, environmental conditions, and management. They can also be applied to comparisons of modeling approaches and appear to be readily extended to evaluating modeled responses to soil water deficits. Issues arise, however, relating to the non-linear responses of crops to soil water availability and to problems in comparing water regimes across experiments. This study used multiple regression for analyzing responses of crop growth and development to varying soil water levels using the Cropping System Model (CSM v4.0.2) as provided in DSSAT 4.02 for bread wheat (Triticum aestivum) and common bean (Phaseolus vulgaris) experiments that comprised different locations, cultivars, and irrigation regimes. Results for two bean irrigation studies in Gainesville, FL showed that the CSM model successfully predicted yield better than simply assuming that yield varied with irrigation level or evapotranspiration. The effectiveness of the model appeared due to accurate simulation of total crop dry weight rather than variation in partitioning. Further examples are provided for tests of effects of year, location and time series data within years. Multiple regression is a powerful technique that merits wider use in the crop modeling community.