Open AccessUsing support vector regression to predict direct runoff, base flow and total flow in a mountainous watershed with limited data in Uttaranchal, India
Author(s) -
Jan Adamowski
Publication year - 2013
Publication title -
annals of warsaw university of life sciences-sggw. land reclamation
Language(s) - English
Resource type - Journals
eISSN - 2081-9617
pISSN - 1898-8857
DOI - 10.2478/sggw-2013-0007
Subject(s) - surface runoff , base flow , watershed , support vector machine , environmental science , hydrology (agriculture) , water resources , flow (mathematics) , geography , mathematics , computer science , drainage basin , engineering , ecology , machine learning , cartography , geotechnical engineering , biology , geometry
Using support vector regression to predict direct runoff, base flow and total flow in a mountainous watershed with limited data in Uttaranchal, India. In the ecologically sensitive Himalayan region, land transformations and utilization of natural resources have modified water flow patterns. To ascertain future sustainable water supply it is necessary to predict water flow from the watersheds as affected by rainfall and morphological parameters. Although such predictions may be made using available process- -based models, in mountainous and hilly areas it is extremely difficult to determine the numerous parameters needed to run such models, thus limiting their applicability. Artificial intelligence (AI) based models are a possible alternative in such circumstances. In this study an AI technique, support vector machines (SVM), was used for modeling the rainfall-runoff relationship from three hilly watersheds in the state of Uttaranchal, India. Different SVM models were developed to predict direct runoff, base flow, and total flow based on the daily rainfall, runoff, and morphological parameters collected from each watershed. The results confirm the potential of SVM models in the prediction of runoff, base flow, and total flow in hilly areas.