Uncertainty analysis and parameter optimization of a water yield ecosystem service model: A case study of the Qilian...

Integrating accurate outcomes of ecosystem service assessments into decision analysis is crucial for an efficient environmental management. However, there are still gaps in our knowledge about estimating and reducing parameter uncertainty in ecosystem service assessments. These gaps may affect the reliability of assessment results and potentially lead to bias or even errors in decision-making. Our study conducted an uncertainty analysis and parameter optimization of the InVEST water yield model in the Qilian Mountains region. We identified sensitive parameters using a global sensitivity analysis and quantified the associated uncertainty using the Monte Carlo method. To optimize these parameters, we applied the Markov chain Monte Carlo method using runoff data from 2006 to 2018 in the trial subbasins. Additionally, we validated the robustness of the optimized parameters in other additional subbasins using runoff data from 2008 to 2018. The results revealed that parameters related to climatic factors (such as annual precipitation and annual reference evapotranspiration) were more sensitive than those influenced by both climate and human activities (such as the vegetation evapotranspiration coefficient). The uncertainty associated with the sensitive parameters was nearly equal to that associated with all the parameters combined, indicating that these sensitive parameters were the primary sources of overall uncertainty. Moreover, the estimated water yields obtained via the optimized parameters were generally closely related to the runoff data from the trial subbasins, especially the China meteorological forcing dataset/Penman-Monteith combination, which achieved an average Nash-Sutcliffe efficiency of 0.71. Additionally, validation in other subbasins confirmed the robustness and transferability of the optimized parameters. Nevertheless, the source and accuracy of these sensitive parameters are critical, and further validation in other regions is needed. This work underscores the importance of rigorous uncertainty analysis and parameter optimization in improving ecosystem service assessments for better decision-making.