Watershed Change Detection Using Hybrid Conceptual Model-Artificial Intelligence

This research investigated LULC changes and its effect on outlet runoff by detecting LULC changes location and severity via an inverse method for the Little River Watershed, USA. In this research, the artificial intelligence and soft computing capabilities such as wavelet-entropy were employed for this aim, as extraction of sub-watershed delineation and watershed information was done in geographic information system. Thereafter, a conceptual rainfall-runoff model (Clark method) was simulated via HEC-HMS. Different watershed outflow hydrographs were generated by variation of storage coefficient parameters of sub-watersheds in the conceptual Clark model. Then the relationship between storage coefficients (as the representative of land use/cover)-watershed outflow was also modeled by artificial intelligence models (artificial neural network and support vector machine). Whereas wavelet-entropy was utilized to avoid importing noise-full information and reducing huge volume of information to the model. Finally, the obtained model detected watershed land use/cover changes in the future years with dynamic watershed assumption. For the validation, the results were compared with recorded changes via normalized difference vegetation index extracted from landsat images. For instance, the comparison approved the ability of the proposed method for LULC change detection of the LRW in a way that deforestation and cropland increasing of the sub-watersheds from 1990 to 2013 were aligned with the SC reduction e.g., 26% decrease of SC for downstream sub-watershed versus 53% decrease and 21% increase of forest and crop lands, respectively.