Global Sensitivity Analysis of Variably Saturated Flow and Transport Parameters and Its Implication for Crop Yield and Root Zone Hydrosalinity

Abstract

Modeling of crop yield and root zone hydrosalinity usually requires a large number of parameters that are often expensive to obtain and have associated measurement errors. Consequently, identifying the most relevant parameters, and their contributions to the uncertainty of the output, might be utilized as a basis to focus research resources efficiently. Global sensitivity analysis (GSA) is a powerful tool that can be utilized to achieve this goal. However, some GSA methods perform better than others, which introduces the risk of mistakenly prioritizing a secondary parameter while neglecting a primary one. This paper evaluates four GSA methods utilized to rank the importance of a total of 18 input parameters with respect to five performance indices using a variable saturated flow and transport model. Some of the major findings of this research are (1) the crop-yield variance is controlled largely by only four of the 18 parameters that were considered in this study; (2) the van Genuchten pore-size parameter was found to be very important to the relative crop-yield prediction and the water-availability index; and (3) the variance-decomposing method, the screening method, and the Monte Carlo–filtering method are generally consistent in their performance, whereas the partial-correlation coefficient method is significantly different.