Evaluation of csm-CROPGRO-cotton for simulating effects of management and climate change on cotton growth and evapotranspiration in an arid environment.

Abstract. Originally developed for simulating soybean growth and development, the CROPGRO model was recently re-parameterized for cotton. However, further efforts are necessary to evaluate the model’s performance against field measurements for new environments and management options. The objective of this study was to evaluate CSM-CROPGRO-Cotton using data from five cotton experiments conducted at the Maricopa Agricultural Center in Maricopa, Arizona. The field experiments tested ambient atmospheric carbon dioxide (CO 2 ) versus free-air CO 2 enrichment (FACE) over two growing seasons (1990 and 1991), two irrigation levels and two nitrogen fertilization levels for one growing season (1999), and three planting densities and two nitrogen fertilization levels with optimum irrigation for two growing seasons (2002 and 2003). The model was calibrated by adjusting cultivar and soil parameters for the most optimal or standard treatment of each field trial, and the model’s responses to suboptimal irrigation, suboptimal nitrogen fertilization, nonstandard planting density, and CO 2 enrichment were evaluated. Modifications to the model’s evapotranspiration (ET) routines were required for more realistic ET simulations in the arid conditions of central Arizona because default approaches underestimated seasonal ET up to 157 mm (15% of mean values). Data quality and availability among the field trials were highly variable, but the combination of data sets from multiple field investigations permitted a more thorough model evaluation. Simulations of leaf area index, canopy weight, canopy height, and canopy width responded appropriately compared to measurements from experimental treatments, although some experiments did not impose enough treatment variability to elicit substantial model responses. Simulation results for densely planted cotton were particularly deficient as compared to other experimental treatments. The model simulated seed cotton yield with root mean squared errors ranging from 105 to 1107 kg ha -1 (3% to 28% of mean values), and total seasonal ET was simulated with root mean squared errors ranging from 12 to 42 mm (1% to 5% of mean values). Seed cotton yield and ET variability due to the imposed experimental treatments were simulated appropriately (p 2 enrichment. Potential opportunities for further model improvement include the estimation of crop responses to high planting densities, the simulation of cotton maturity and defoliation events, and the calculation of canopy temperature as part of a complete energy balance algorithm.