Multi-Site Calibration and Validation of the Hydrological Component of SWAT in a Large Lowland Catchment

This study describes an application of a hydrological component of the catchment model, Soil and Water Assessment Tool (SWAT) in the Narew basin (ca. 28,000 km²) situated in the north-east of Poland. The main objective was to perform a multi-site (spatially distributed) calibration and validation of SWAT using daily observed flows from 23 gauging stations as well as to assess the model’s capability to perform reliable simulations at spatial scales that were smaller than those in the calibration phase. A detailed description of the model configuration for the Narew basin upstream from Zambski Kościelne gauge has been given. Building a SWAT project for a large-scale application appeared to be a demanding task, with the most critical part of preparing soil input data. Sensitivity analysis performed using a LH-OAT method indicated which parameters should be used in autocalibration. The ParaSol tool allowed to find the best parameter values from 8D parameter space in 11 calibration areas. The calibrated model generally performed well, with average Nash–Sutcliffe Efficiency for daily data equal to 0.68 for calibration period and 0.57 for validation period. SWAT correctly conserved the mass balance in different parts of the catchment as well as at the main outlet. The model results were significantly better in large basins than in small basins. Spatial validation performed at 12 independent catchments ranging in size from 355 to 1,657 km² revealed that adapted SWAT model should rather not be used in the Narew basin catchments smaller than ca. 600 km². It is believed that ensuring reliability of SWAT results at smaller spatial scales, which would be of interest to decision-makers, would require providing better input data and in particular using significantly more precipitation stations.