Quantitative global sensitivity analysis of the RZWQM to warrant a robust and effective calibration

Sensitivity analysis is a useful tool to identify key model parameters as well as to quantify simulation errors resulting from parameter uncertainty. The Root Zone Water Quality Model (RZWQM) has been subjected to various sensitivity analyses; however, in most of these efforts a local sensitivity analysis method was implemented, the nonlinear response was neglected, and the dependency among parameters was not examined. In this study we employed a comprehensive global sensitivity analysis to quantify the contribution of 70 model input parameters (including 35 hydrological parameters and 35 nitrogen cycle parameters) on the uncertainty of key RZWQM outputs relevant to raspberry row crops in Abbotsford, BC, Canada. Specifically, 9 model outputs that capture various vertical-spatial and temporal domains were investigated. A rank transformation method was used to account for the nonlinear behavior of the model. The variance of the model outputs was decomposed into correlated and uncorrelated partial variances to provide insight into parameter dependency and interaction. The results showed that, in general, the field capacity (soil water content at −33 kPa) in upper 30 cm of the soil horizon had the greatest contribution (>30%) to the estimate of the water flux and evapotranspiration uncertainty. The most influential parameters affecting the simulation of soil nitrate content, mineralization, denitrification, nitrate leaching and plant nitrogen uptake were the transient coefficient of fast to intermediate humus pool, the carbon to nitrogen ratio of the fast humus pool, the organic matter decay rate in fast humus pool, and field capacity. The correlated contribution to the model output uncertainty was <10% for the set of parameters investigated. The findings from this effort were utilized in two calibration case studies to demonstrate the utility of this global sensitivity analysis to reduce the risk of over-parameterization, and to identify the vertical location of observations that were the most effective to use as RZWQM calibration targets when water flux estimates are a key focus.