Impacts of hydrometeorological factors on discharge simulation in the North West Himalayas: a SUFI-2 algorithm-driven investigation using the SWAT model.

The Soil and Water Assessment Tool (SWAT) is a computational hydrological model extensively utilised for developing sustainable strategies and viable approaches for prudent management of water resources. The central emphasis of this study is on the utilisation of SWAT model along with SWAT-CUP (SWAT calibration toolbox) to simulate streamflow in the upper Jhelum basin, the North West Himalayas, for a period of 20 years from 2000 to 2019. The global sensitivity analysis algorithm, Sequential Uncertainty Fitting 2 (SUFI-2) of SWAT-CUP, is used for sensitivity and uncertainty analysis. The optimised parameter set estimated by SUFI-2 constitutes 11 parameters that are found to be sensitive with soil conservation service (SCS) curve number (CN) being the most influential parameter followed by snowmelt base temperature. Autocorrelation analysis using the autocorrelation function was conducted on the temperature and precipitation time series data, followed by a pre-whitening procedure to remove any autocorrelation effects. Subsequently, the modified Mann-Kendall (MMK) test was applied to examine trends in the annual temperature and precipitation data. The results indicated statistically significant positive trends in both datasets on an annual scale. The results for the calibration period (2003-2014) for monthly simulation displayed good model performance at three gauging stations, Rambiara, Sangam and Ram Munshi Bagh with R2 values of 0.83, 0.847, 0.829, P factor values of 0.73, 0.76, 0.75 and R factor values of 0.61, 0.58, 0.63, respectively. The validation results for monthly simulation for the 2015-2019 period showed good model agreement with R2 values of 0.817, 0.853, and 0.836, P factor values of 0.76, 0.8, and 0.75 and R factor values of 0.62, 0.53, and 0.65, respectively. The study concludes that the SWAT hydrological model can perform satisfactorily in high mountainous catchments and can be employed to analyse the impact of land use-land cover changes and the effect of climate variation on streamflow dynamics.

Employing sensitivity analysis to catchments having scanty data.

Sensitivity analysis (SA) is generally desirable for parameter optimization, mapping, and calibration in hydrological models, yet the implementation of SA in data-sparse regions is usually avoided due to a lack of continuous data. The present study proposes the novel concept of "minimum continuous data period" to overcome this constraint. It analyses the sensitivity profile of two data-suffice sub-catchments of a data-sparse watershed using data at various timescales to determine the minimum data period required for the SA. The results suggest that the SA employing a minimum data period (2 years in this study) replicated the actual sensitivity profile by an average of 77.5% while replicating the most sensitive and insensitive parameters by 100%. The study encourages the use of the data from sub-catchments to determine the sensitivity profile of the data-sparse catchment. It would benefit in improving the use of SA for rainfall-runoff modelling in data-scanty regions.

Investigating the potential of Morris algorithm for improving the computational constraints of global sensitivity analysis.

Sensitivity analysis (SA) is widely acknowledged as advantageous and worthwhile in recognizing parameters for model calibration and optimization, especially in complex hydrological models. Although Sobol global SA is an efficient way to evaluate the sensitivity indices, the computational cost is a constraint. This study analyzes the potential of Morris global SA to achieve results tantamount to Sobol SA, at a much cheaper computational expense, using a new approach of increasing the number of replications for the Morris algorithm. SA for two catchments is performed on a coupled hydrological model using Morris and Sobol algorithms. Two target functions are used for each of the algorithms. Sobol SA required 660000 model simulations accounting for about 400 computing hours, whereas increasing the replications from 1000 to 3000, the Morris method called for 63000 runs and 06 computing hours to produce significantly similar results. The Morris parameter ranking improved 50% with respect to Sobol SA by a three-fold increase in replications with a small 5-h increase in the computational expense. The results also suggest that target functions and catchments influence parameter sensitivity. The new approach to employ the Morris method of SA shows promising results for highly parameterized hydrological models without compromising the quality of SA, specifically if there are time constraints. The study encourages the use of SA, which is mainly skipped because of higher computational demands.