Uncertainty analysis for integrated water system simulations using GLUE with different acceptability thresholds

Integrated water quantity and quality simulations have become a popular tool in investigations on global water crisis. For integrated and complex models, conventional uncertainty estimations focus on the uncertainties of individual modules, e.g., module parameters and structures, and do not consider the uncertainties propagated from interconnected modules. Therefore, this study investigated all the uncertainties of integrated water system simulations using the GLUE (i.e., generalized likelihood uncertainty estimation) method, including uncertainties associated with individual modules, propagated uncertainties associated with interconnected modules, and their combinations. The changes in both acceptability thresholds of GLUE and the uncertainty estimation results were also investigated for different fixed percentages of total number of iterations (100000). Water quantity and quality variables (i.e., runoff and ammonium nitrogen) were selected for the case study. The results showed that module uncertainty did not affect the runoff simulation performance, but remarkably weakened the water quality responses as the fixed percentage increased during calibration and validation periods. The propagated uncertainty from hydrological modules could not be ignored for water quality simulations, particularly during validation. The combination of module and propagated uncertainties further weakened the water quality simulation performance. The uncertainty intervals became wider owing to an increase in the fixed percentages and introduction of more uncertainty sources. Moreover, the acceptability threshold had a negative nonlinear relationship with the fixed percentage. The fixed percentages (similar to 20.0%-30.0%) were proposed as the acceptability thresholds owing to the satisfactory simulation performance and noticeably reduced uncertainty intervals they produced. This study provided methodological foundations for estimating multiple uncertainty sources of integrated water system models.

Automated summary

The study investigated uncertainties in integrated water system simulations using the GLUE method, including uncertainties from individual modules, propagated uncertainties from interconnected modules, and their combinations. Results showed that both module and propagated uncertainties affected water quality simulation performance, and the acceptability threshold had a negative nonlinear relationship with the fixed percentage, with 20.0%-30.0% proposed as satisfactory thresholds.