An integrated land use-hydrological model for the Bartın spring watershed

Abstract

The Bartın spring watershed located in northwestern Turkey supplies the water needs of the Bartın, Amasra and nkumu city centers. Although the water is presently adequate for the spring, autumn and winter seasons, the city centers suffer water scarcity in summer seasons. It is widely recognized that land use changes have a significant impact on the water budget of a watershed. The purpose of this study is to model the land use and hydrological processes within the Bartın spring watershed and simulate the water yield under different future land use scenarios. For this purpose, a coupled land use-hydrodynamics model was developed. The land use dynamic simulation model was built using STELLA dynamic simulation platform while the MIKE SHE computer program was used to simulate the hydrodynamics of the watershed. The link between the two models is through the Leaf Area Index (LAI) and Root Depth (RD) parameters which are generated in the land use model and supplied to the hydrodynamics model. The dynamic land use model represents several forest stand groups and land use categories with their respective acreages and their conversions. The model was structurally validated and analyzed through a series of sensitivity tests. It was calibrated with respect to the historical geographical data and the calibration results are quite satisfactory. The calibration target for the hydrodynamics model was the river discharge at the downstream end of the watershed, while the calibration parameters were the saturated hydraulic conductivity of the deeper soil, the threshold melting temperature values and the RD value. The optimal simulation produced correlation coefficients, R=0.72 and R2=0.52 with a mean error of 0.01 m3/s. Sensitivity analyses of the hydrodynamics model indicate that it is quite sensitive to the land use type; complete agricultural cover would yield 25- 33% higher discharge compared to a completely forested watershed. The hydrodynamics model is also sensitive to the LAI up to a value of 3 and becomes insensitive for higher values. The model however, is not strongly sensitive to hydraulic conductivity of the saturated zone.