Proceedings of the 2022 Mississippi Water Resources Conference

Year: 2022 Authors: Nepal D., Parajuli P.B.

For the management of watersheds, hydrological models play an important role. To improve the accuracy of a hydrological model, calibration is performed. Validation helps to evaluate the improved accuracy of the calibrated model. However, sometimes, calibration is challenged by limited data availability. Measurement of water quality samples is done less frequently than flow because of high cost and limited resources availability. Therefore, it is necessary to estimate their loads/ concentrations during the periods of no measurements. The present study aims to 1) test regression model performances in predicting water quality loads 2) Soil and Water Assessment Tool (SWAT) calibration and validation of flow, total suspended solids (TSS), total nitrogen (TN) and total phosphorous (TP) at three monitoring stations at Big Sunflower River Watershed (BSRW). In this study we used Load Estimator (LOADEST) which has different regression models to predict water quality loads (during the period when flow data are available) to increase the number of data availability for calibration and to convert concentrations of TSS, TN and TP into loads since SWAT outputs TN and TP loads. Model evaluation was performed using R², Nash-Sutcliff Efficiency (NSE) and Partial Load Ratio (PLR). The performance of LOADEST was found generally good in load prediction with a tendency towards overestimation in most of the cases (R²: 0.90-0.96, NSE: 0.50-0.95 and PLR: 0.84-1.17). The Sequential Uncertainty Fitting SUFI-2 algorithm inside SWAT-CUP was used for calibration and validation. The uncertainty analysis showed acceptable values of P and R factors for streamflow (p-factor: 0.72-0.87 and r-factor: 0.74-1.27) and TSS (p-factor: 0.56-0.89 and r-factor: 0.43-2.83). Model performance evaluation was performed using R2 and NSE. Model performed well for stream flow during both calibration and validation (R²: 0.60-0.86, NSE: 0.60-0.86). Similarly, performance evaluation for TSS indicated acceptable values (R²: 0.60-0.91, NSE: 0.38-0.91). The model evaluation statistics for TN and TP will be obtained further. The calibrated and validated model will then be used to simulate different Best Management Practice (BMP) scenarios. This study is believed to offer recommendations for successful multi-site and multi-variable calibration and validation to SWAT modelers under limited data availability.