Development and application of process-based simulation models for cotton

38 The development and application of cropping system simulation models for cotton production has 39 a long and rich history, beginning in the southeastern United States in the 1960's and now expanded to 40 major cotton production regions globally. This paper briefly reviews the history of cotton simulation 41 models, examines applications of the models since the turn of the century, and identifies opportunities for 42 improving models and their use in cotton research and decision support. Cotton models reviewed include 43 those specific to cotton (GOSSYM, Cotton2K, COTCO2, OZCOT, and CROPGRO-Cotton) and generic 44 crop models that have been applied to cotton production (EPIC, WOFOST, SUCROS, GRAMI, 45 CropSyst, and AquaCrop). Model application areas included crop water use and irrigation water 46 management, nitrogen dynamics and fertilizer management, genetics and crop improvement, climatology, 47 global climate change, precision agriculture, model integration with sensor data, economics, and 48 classroom instruction. Generally, the literature demonstrated increased emphasis on cotton model 49 development in the previous century and on cotton model application in the current century. Although 50 efforts to develop cotton models have a 40-year history, no comparisons among cotton models were 51 reported. Such efforts would be advisable as an initial step to evaluate current cotton simulation 52 strategies. Increasingly, cotton simulation models are being applied by non-traditional crop modelers, 53 who are not trained agronomists but wish to use the models for broad economic or life cycle analyses. 54 While this trend demonstrates the growing interest in the models and their potential utility for a variety of 55 applications, it necessitates the development of models with appropriate complexity and ease-of-use for a 56 given application, and improved documentation and teaching materials are needed to educate potential 57 model users. Spatial scaling issues are also increasingly prominent, as models originally developed for 58 use at the field scale are being implemented for regional simulations over large geographic areas. 59 Research steadily progresses toward the advanced goal of model integration with variable-rate control 60 systems, which use real-time crop status and environmental information to spatially and temporally 61 optimize applications of crop inputs, while also considering potential environmental impacts, resource 62 limitations, and climate forecasts. Overall, the review demonstrates a languished effort in cotton 63 simulation model development, but the application of existing models in a variety of research areas 64 remains strong and continues to grow. 65