Optimizing Crop Management Practices with DSSAT

Abstract

Rising world population and climate change are making the ability to predict crop yields and optimal management parameters under stressed environments increasingly important, but current models only provide recommendations for a limited set of parameters and stresses. This paper presents a model that utilizes the Genoud evolutionary algorithm with derivatives for predicting crop yields as well as optimal management practices that maximize harvestable yield or profit in any environmental condition and location, while simultaneously minimizing nitrogen pollution. Unlike previous models, the one presented in this paper can simultaneously handle multiple environmental stresses and provide recommendations for optimal irrigation schedules, fertilization schedules, the timing of planting and harvest, and the choice of cultivar and soil type to maximize either profit or harvestable yield, making it much more generalizable and accessible for a larger audience. The results in this thesis showcase the model’s ability in providing optimal recommendations under a variety of stresses and for various demographics of farmers that have differing accessibility to resources, proving its use for a wide range of applications.