Reducing uncertainty in a semi distributed hydrological modeling within the GLUE framework

The calibration of hydrologic models is a worldwide challenge due to the uncertainty involved in the large number of parameters and the inability to reliably measure the distributed physical characteristics of a catchment results in significant uncertainty in the parameterization of physically based, semi-distributed models. The difficulty even increases in a region with high seasonal variation of precipitation. Therefore, a successful application of a hydrologic model in applied water research strongly depends on calibration and uncertainty analysis of model output (). Shen and et al. () quantify the parameter uncertainty of the stream flow and sediment simulation by SWAT model in the part of China. The research indicated that only a few parameters affected the final simulation output significantly. Shape and et al. () assessed the capability of the Soil and Water Assessment Tool (SWAT) model to capture event-based and long-term monsoonal rainfall–runoff processes in complex mountainous terrain and found that high elevation steep sloping regions were generally base flow dominated while lower elevation locations were predominately influenced by surface runoff. In this paper, the Generalized Likelihood Uncertainty Estimation (GLUE) method was combined with the Soil and Water Assessment Tool (SWAT) to quantify the monthly stream flow in the eastern Gorganrood river basin.
The Golestan Regional Water Company (GRWC) monitoring site at Gazaghly was chosen as the outlet for the entire watershed since it is the lowest monitoring station on the river not subject to dam influence. Annual precipitation decreases in the west to east direction, low (200 mm) to high (880 mm) and from south to north direction. Model has been calibrated and validated using monthly runoff flow data of ten years 1984 and 1993. Data pertaining to year 1984-1990 has been used for calibration and 1991-1993 for validation.
In semi-distributed models such as SWAT, it is necessary to identify the most sensitive parameters to obtain a better understanding of the overall hydrologic processes before calibration. Based on this study, only a few parameters affected the final simulation output significantly. The parameters such as CN2 (curve number), GWQMN (threshold in the shallow aquifer), RCHRG_ DP (deep aquifer percolation fraction), ALPHA_ BNK (base flow alpha factor for bank storage), ESCO (soil evaporation compensation factor) and SOL_ K were found to be the most sensitive parameters. According to results, the parameter CN2, was the most effective parameter on the output discharge of the studied area and CN2, was identified as a main source of uncertainty in results. Statistical model performance measures, coefficient of determination (R2) of 0.80, the Nash-Sutcliffe simulation efficiency (ENS) of 0.72, for calibration and 0.83, and 0.80, respectively for validation, indicated good performance for runoff estimating on monthly time step in the outlets of the Gazaghly gauging station. 69–74% of the observed runoff data fall inside the 95% simulation confidence intervals in the calibration and validation periods. The evaluation statistics for the daily runoff simulation showed that the model results were acceptable, but the model underestimated the runoff for high-flow events.
SWAT was applied to simulate monthly runoff in part of the Gorganrood river basin. Results of uncertainty analysis indicated that SWAT model had large uncertainties for calibration period, although the simulation of monthly runoff for the Gazaghly station was satisfactory during the calibration period and in the model calibration stage 69 of runoff observations were within the corresponding 95% confidence interval. This study would provide useful information for hydrology modeling related to policy development in the Gorganrood river basin and other similar areas.