Multidecadal Analysis of an Engineered River System Reveals Challenges for Model-Based Design of Human Interventions

Hydraulic models were used in practice to predict the effect of human intervention during extreme conditions. However, the accuracy of such predictions remains untested. In this study, we compare a simulated trend in water levels covering a twenty-year period of large-scale human intervention with a thirty-year observational record. The results show that the observed water levels display a linearly decreasing trend attributed to channel bed erosion. A deviation from this trend, which would be an indication of the effect of human intervention, was not observed. We propose that the most likely explanation for this is that any effect observable at lower discharge is hidden in the uncertainty of the rating curve. Given the inherent uncertainties associated with making predictions about a changing system for conditions with a low period of return, we argue that model uncertainty should be considered in intervention design