Parameter estimation for process-oriented crop growth models

The accuracy of process-oriented crop growth models depends on the soundness of conceptual representation of physiological processes and the proper parameter values used in their mathematical representations. These parameters are often difficult to measure, and the data needed to estimate them are often not readily available. This constitutes a major limitation to the applicability of process-oriented models. In this article, a procedure for estimating plant growth parameters for a single species is described. GRASIM, a mechanistic grazing simulation model, is used for this work. GRASIM's plant growth module integrates major processes including light interception, photosynthesis, respiration, tissue recycling, and senescence using ten primary crop growth parameters. GRASIM was first converted into Microsoft Excel spreadsheets. The methodology then uses Excel Solver's Generalized Reduced Gradient (GRG2) algorithm to estimate crop growth parameters by iteratively minimizing the difference between simulated and field-observed data. Physiologically sound ranges for these parameters were defined from the literature and incorporated into the optimization procedure, guiding the initialization of parameters and serving as boundary constraints to ensure the feasibility of the optimized parameter set. The methodology was evaluated using observed barley (Hordeum vulgare L.) growth and soil water data from a two-year (1999-2000) experiment conducted at the American University of Beirut Agricultural Research Center in the Bekaa Valley, Lebanon. Optimized parameters using 1999 data were within the specified physiological ranges, and they gave a good fit between simulated and measured crop biomass in both seasons. Although the simulated soil water contents in the top 30 cm and bottom 70 cm soil layers followed the observed general trends, they lacked the observed fluctuations. Due to the fact that GRASIM modifies daily potential crop growth based on soil water availability, it is important to continue model development to achieve more accurate estimate of soil water contents. This will allow the parameter estimation procedure to find crop growth parameters closer to those defined by the combined effects of plant physiology and field physical conditions. This study will benefit the use of mechanistic crop models and help to extend the applicability of such models to species with little available growth data. © 2004 American Society of Agricultural Engineers.