MODEL–BASED OPTIMIZATION OF CROP MANAGEMENT FOR CLIMATE FORECAST APPLICATIONS

Abstract

Recent improvements in climate forecast technology have led to new uses of crop models for exploring potentialbenefits of tailoring crop management to expected weather conditions. However, conventional use of crop models limits eachsimulation experiment to a small, predetermined subset of the possible combinations of variables. Unfortunately, much of thepotential contribution of dynamic models is untapped when seeking optimal management parameters under varyingenvironmental and/or economic scenarios. This research linked a widely used crop model, CERESMaize, to a simulatedannealing algorithm and a partial budget calculator, to permit optimizing economic results by varying crop management.Management was optimized by El NioSouthern Oscillation (ENSO) phase using a 67year series of daily weather data fromPergamino, Argentina. The perhectare value of the ENSOoptimized management was calculated. Nine managementvariables were included in the optimizations, at two levels of resolution (increments or step size) for each variable. Theoptimization algorithm, Adaptive Simulated Annealing (ASA), required tuning to achieve reasonable reliability andefficiency. Although the optimizer did not consistently find the precise optimal combination at either resolution, it didconsistently find the optimal region, with small differences in some management variables. A quenching variation ofsimulated annealing was found to be much faster but considerably less reliable. Optimization by ENSO phase leads tophasedifferentiated management: earlier planting date, higher N applications, and increased plant density lead to higheryields during El Nio, as compared to neutral and especially La Nia years. The CERESMaize to ASA linkage is useful forinvestigating optimal combinations of management practices using validated crop simulation models.