Developing an uncertainty-based auto-calibrated reservoir eutrophication model: a case study of Karkheh Dam, Iran

As crucial as it is to monitor the water quality in the reservoirs, constant measurement of different water quality indices can become a serious challenge. One way to alleviate this predicament is to create accurate numerical models and yet, the modeling process itself can have various problems that need to be addressed effectively. One major problem regarding water quality modeling is the high number of parameters and the uncertainty accompanying them. Eutrophication, being a significant water quality issue in the water reservoirs, is no exception and is related to a multitude of parameters. This paper presents an uncertainty-based auto-calibrated two-dimensional water quality and hydrodynamic (CE-QUAL-W2) model which was developed to simulate the eutrophication phenomenon in the Karkheh Reservoir in Iran. Automatic calibration of the CE-QUAL-W2 model using the Sequential Uncertainty Conformity Algorithm (SUFI-2) resulted in 11, 12.8, 10, 14.7, and 37.5 percent lower root mean square error (RMSE) for dissolved oxygen, total phosphorus, NH4, NO3-NO2, and Chlorophyll-a, respectively, compared to the results obtained by the particle swarm optimization (PSO). Furthermore, 76 percent of the measured data were in a 95 percent confidence interval as a result of calibration with the SUFI-2. In addition, the SUFI-2 algorithm outperformed the PSO algorithm within just 250 function evaluations, which was 20 times fewer than what was required by the PSO. Based on the results, it seems that the SUFI-2 algorithm can help the calibration process to achieve a more accurate and reliable model while decreasing the computational burden of the modeling process.