Calibration of a Physically-Based Semi-Distributed Hydrologic Model: The Importance of Internal Justification

Abstract:

Calibration of physically-based distributed hydrologic models has always been a challenging task and subject of controversy in the literature. This study is aimed to investigate how different physiographic characteristics of watersheds call for adaption of the methods used in order to have more robust and internally justifiable simulations. Haw Watershed (1300 sq. mi.) is located in the piedmont region of North Carolina draining into B. Everett Jordan Lake located in west of Raleigh. Major land covers in this watershed are forest (50%), urban/suburban (21%) and agriculture (25%) of which a large portion is pasture. Different hydrologic behaviors are observed in this watershed based on the land use composition and size of the sub-watersheds. Highly urbanized sub-watersheds show flashier hydrographs and near instantaneous hydrologic responses. This is also the case with smaller sub-watersheds with relatively lower percentage of urban areas. The Soil and Water Assessment Tool (SWAT) has been widely used in the literature for hydrologic simulation on daily basis using Soil Conservation Service Curve Number method (SCS CN). However, it has not been used as frequently using the sub-daily routines. In this regard there are a number of studies in the literature which have used coarse time scale (daily) precipitation with methods like SCS CN to calibrate SWAT for watersheds containing different types of land uses and soils reporting satisfying results at the outlet of the watershed. This is while for physically-based distributed models, the more important concern should be to check and analyze the internal processes leading to those results. In this study, the watershed is divided into several sub-watersheds to compare the performance of SCS CN and Green & Ampt (GA) methods on different land uses at different spatial scales. The results suggest better performance of GA compared to SCS CN for smaller and highly urbanized sub-watersheds although GA predominance is not very significant for the latter. Also, the better performance of GA in simulating the peak flows and flashy behavior of the hydrographs is notable. GA did not show a significant improvement over SCS CN in simulating the excess rainfall for larger sub-watersheds.