Rice (Oryza Sativa L.) is a major staple food for nearly half of the world's population. Rice is cultivated under diverse environments, from tropics to temperate regions, at latitude ranging from 35S to 53N. However, the recent and ongoing global warming poses a great risk to rice production. Dynamic crop growth models are frequently used to study the response of crops to variation in environmental conditions, including climate change. Crop phenology affects simulated crop yield; thus, accurate modeling of development rates is critical, since these models estimate dry matter production based on the developmental rates during a given period. Developmental rates affect yields in different ways. For example, rice generally develops faster under higher temperatures, which may reduce growth and lead to a lower yield potential. On the other hand, using a shorter duration crop may help avoid pest damage, escape drought, or optimize the cropping system. In the first chapter of this dissertation, we evaluated different optimization approaches in the Oryza2000 and CERES-Rice phenology sub-models to assess the importance of optimizing cardinal temperatures for model performance and systematic error (correlation between temperature and phenology prediction error). We used two optimization approaches -- single-stage (planting to heading) and three-stage (planting to panicle initiation (PI); PI to heading (HD); and HD to physiological maturity (MT)) -- for all model parameters. Our results indicate that three-stage optimization increased model accuracy, especially for maturity stage. We also show that optimization to minimize systematic error reduced bias when RMSE was constrained. However, relatively small systematic error was found for all phenological stages compared to previous studies. Finally, our results demonstrated that cardinal temperature optimization had no effect on systematic error reductionIn the second chapter of this dissertation, we developed a simple thermal time model to determine if optimal temperature parameter values differ among developmental stages in rice, and to quantify the effect of using stage-dependent temperature parameters on model performance and systematic error relative to using a constant optimized temperature parameters across all stages. Our results indicate that temperature dependence in rice phenology changes with development-stage. Furthermore, we show that rice is most responsive to temperature between planting to panicle initiation and optimum temperature threshold increases with plant development. Therefore, optimizing cardinal temperature parameters for each stage improves phenology model accuracy; stage-dependent temperature parameters reduced bias in phenology models. In the final chapter of this dissertation, we compared the use of water (T[subscript w]) and air (T[subscript a]) temperatures in a rice phenology model. Specifically, we evaluate whether T[subscript w] has the stronger influence on development rate when the growing point is under water, while T[subscript a] is more important when the growing point is above the water. We found that T[subscript w] and T[subscript a] influence rice development but at different times. During the first part of the season when the growing apex is under water T[subscript w] determines developmental rates, while later in the season it is T[subscript a]. Incorporating both T[subscript w] and T[subscript a] into crop development models increased the prediction accuracy. Our study demonstrates that it was maximum temperature differences between T[subscript w] and T[subscript a] that affected thermal time accumulation and consequently developmental rates.