Crop models range in complexity from simple ones with a few state variables to complex ones having a large number of model parameters and state variables. Determining and understanding how sensitive the output of a model is with respect to model parameters is a guiding tool for model developers. A new cotton model is being developed using the Cropping System Model (CSM)-CROPGRO crop template that allows the introduction of a new crop and its integration with other modules such as soil and weather without changing any code. The main goal of this study was to investigate whether global sensitivity analysis would provide better information on the importance of model parameters than the simpler and commonly used local sensitivity analysis method. Additionally, we were interested in determining the most important crop growth parameters in predicting development and yield and if the model sensitivity to these parameters would vary under irrigated and rainfed conditions. Sensitivity analyses were performed on dry matter yield and length of season model responses for a wet cropping season (year 2003) and a dry cropping season (year 2000) under irrigated and rainfed conditions. Results indicated that global sensitivity analysis improved our understanding of the importance of the model parameters on model output relative to local sensitivity analysis. Results from global sensitivity analysis indicated that the specific leaf area under standard growth conditions (SLAVR) was the most important model parameter influencing cotton yield under both irrigated and rainfed conditions when taking into account its range of uncertainty. Results from local sensitivity analysis indicated that the light extinction coefficient (KCAN) was the most influencing model parameter. In both global and local sensitivity analyses, the duration between first seed and physiological maturity (SD-PM) was the most important parameter for season length response. The differences obtained for global vs. local sensitivity analysis can be explained by the inability of local sensitivity analysis to take into consideration the interactions among parameters, their ranges of uncertainty, and nonlinear responses to parameters.