FACTORS AFFECTING MODEL SENSITIVITY AND UNCERTAINTY: APPLICATION TO AN IRRIGATION SCHEDULER

This work describes the global sensitivity and uncertainty analysis of the Mississippi Irrigation Scheduling Tool (MIST) using the Sobol' method. An often overlooked but driving factor in any sensitivity and uncertainty analysis is the selection of the prior probability distribution functions (PDFs) that are used to describe parameter and input uncertainty. These prior PDFs have a direct impact on the total model uncertainty as well as the ranking of importance of model inputs and parameters. Furthermore, an uncertainty and sensitivity analysis generally focuses on a single objective function for analysis, but model outputs are often analyzed and summarized using a variety of objective functions. Therefore, it is important to include this variety of objective functions in any sensitivity and uncertainty analysis. In this article, we show how the choice of prior PDFs and objective functions impacts the ranking of important parameters and inputs in the MIST model. For example, under the "first day to irrigate" objective function, precipitation was the most important input when using informed prior PDFs, but precipitation ranked as the tenth most important input when using uninformed prior PDFs. Similarly, when using the uninformed prior PDFs, the curve number was the second most important input for the water balance objective function but only the eighth most important when assessing the "first day to irrigate" objective function. Furthermore, in the MIST model, increasing model complexity through the addition of algorithms, inputs, and parameters increases model uncertainty. Finally, in this particular application using the data described, the crop coefficient and precipitation were the most important parameters or inputs, while the initial abstraction and minimum temperature were the least important parameters or inputs. These results provide theoretical insights into sensitivity and uncertainty analysis studies as well as context-specific implications for strategic enhancement of the MIST model.