Assessing model parameters sensitivity and uncertainty of streamflow, sediment, and nutrient transport using SWAT

Abstract

The objectives of this study were to (1) simulate streamflow, total sediment (TS), total phosphorus (TP), and total nitrogen (TN) loads; and (2) use the sequential uncertainty fitting (SUFI-2) algorithm to quantify the model parameter sensitivity and uncertainty in simulating streamflow, TS, TP, and TN loads using Soil and Water Assessment (SWAT) model in Big Sunflower River watershed (BSRW). The model was calibrated from 1996 to 2003 and validated from 2004 to 2010 for daily streamflow, TS load, TP load, and TN load. The model performed well simulating daily streamflow (R2 = 0.58–0.75, NSE = 0.47–0.75), TS load (R2 = 0.50–0.72, NSE = 0.47–0.66), and TP load (R2 = 0.79–0.82, NSE = 0.73–0.77), and the model performance was slightly low for TN load (R2 = 0.13–0.31, NSE = −0.09 to 0.07). This study determined that parameter uncertainty was greatest for simulating TN load (p-factor = 0.48, r-factor = 1.25) and that parameter uncertainty was lowest for simulating streamflow (p-factor = 0.70–0.78, r-factor = 1.18–1.19). Output uncertainty was much greater during peak streamflow and peak pollutant loads compared to periods of low streamflow and low pollutant loads. The sensitivity analyses found that streamflow was most sensitive to Manning’s roughness coefficient for the main channel (CH_N2), TS load was most sensitive the peak rate adjustment factor for sediment routing in the tributary channels (ADJ_PKR), TP load was most sensitive to the Phosphorus enrichment ratio for loading with sediments (ERORGP), and TN load was most sensitive to the denitrification exponential rate coefficient (CDN). Uncertainty was found to be much greater during peak streamflow and peak pollutant loads compared to periods of low streamflow and low pollutant loads. Keywords: Model parameters, Sensitivity, Uncertainty, SWAT, Pollutant loads