Measuring the uncertainty and sensitivity of the Mississippi Irrigation Scheduling Tool (MIST)

Author(s): Linhoss, A.; Tagert, M.; Bukah, H.

The Mississippi River Valley Alluvial Aquifer has seen dramatic declines due to pumping for irrigation in northwestern Mississippi. Irrigation scheduling saves water and energy without sacrificing yield through the optimal frequency and duration of water applications. Models, based on based on crop, soil, and climatic data, can be used for irrigation scheduling. The Mississippi Irrigation Scheduling Tool (MIST) is one such model. MIST uses the Penman-Montieth equation, along with crop coefficients, and the Soil Conservation Service curve number method to calculate runoff and evapotranspiration in a field. The resulting water balance can be used to schedule irrigation events. When using a model, such as MIST, for management purposes, it is important that the user be aware of the reliability or uncertainty of the model. Furthermore, model calibration can be optimized by identifying the most important driving parameters within a model. The objective of this research was to conduct a global sensitivity and uncertainty analysis of the MIST model to quantify model reliability and identify the most important model parameters. In order to provide a realistic representation of the model’s uncertainty and sensitivity, parameter probability distributions were based on measured values and compared to results that used generic percentages. Six global sensitivity analysis methods were employed to understand how those methods differ (Sobol, FAST, Morris, random, quasi-random, and Latin-hypercube sampling). The results show that the sensitivity analysis methods return similar results. However, the method by which the probability distributions were determined were important in determining the results. These results are useful in developing better modeling tools and can help farmers and managers successfully apply MIST model recommendations.
Download the presentation

Go back


Past Conference Archive