Akademik Veri Yönetim Sistemi
Tezin Türü: Doktora
Tezin Yürütüldüğü Kurum: Orta Doğu Teknik Üniversitesi, Fen Bilimleri Enstitüsü, Fen Bilimleri Enstitüsü, Türkiye
Tezin Onay Tarihi: 2016
Öğrenci: AYFER ÖZDEMİR
Asıl Danışman (Eş Danışmanlı Tezler İçin): UĞUR MURAT LELOĞLU
Özet:In recent years, water resources were negatively affected from uncontrolled agricultural, industrial activities and settlements on river basins. Hydrologists and water resource managers have widely used hydrologic models as tools for water resources development, water environment preservation, water resources allocation and understanding utilization. In order to apply hydrological models successfully in practical water resources investigations, careful calibration and uncertainty analysis are needed. Hydrological models are validated by comparing the outputs of the models to measurements. The deviations of the outputs from the ground truth, the error, can be the result of uncertainties of the inputs, uncertainties of the parameters of the model and the model itself. When a hydrologic model is calibrated, the parameters of the model are fine-tuned within a predefined range to minimize an error metric created from error terms. One such hydrological transport model is the Soil and Water Assessment Tool (SWAT) which is also integrated into a Geographic Information System (GIS) that supports the input of topography, land use, soil type, and other digital data. SWAT is a semi-distributed hydrological model that simulates hydrological processes at subbasin level by derivingHydrologic Response Units (HRU) by thresholding areas of soil type, land use and management combinations. Currently, there are automated calibration methods for SWATusing nonlinear optimization such as Levenberg-Macquart or global optimization methods like Genetic Algorithms. These optimization approaches that try to calibrate very complex models have some drawbacks: 1) Since the search space is large and the model is complicated, the convergence takes very long time, 2) for the same reasons, probability of finding a local optimum is large, 3) final result is too sensitive to the initial estimates of the parameters, 4) the sizes of the HRUs, which are the areas over which the parameters are assumed to be constant, are left to the user and their relation to the performance of calibration/validation is unknown. In this thesis, a hierarchical (coarse-to-fine) approach to HRU selection and calibration is investigated. The HRUs are generated automatically by a script and the number of HRUs are increased at each level of the hierarchy. The calibration results at each level are used as initial values for the next level. This way, we obtain not only an increase in the speed and accuracy of the calibration, but we also find out the optimum HRU sizes and HRU generation parameters. The algorithm developed in this thesis is tested on two basins with different properties and the results are promising.