DBPapers
DOI: 10.5593/sgem2017H/33/S12.014

COMPARISON OF THE HYDROLOGIC AND DATA DRIVEN RIVER FLOW GENERATORS

M. Cisty, V. Soldanova
Thursday 23 November 2017 by Libadmin2017

References: 17th International Multidisciplinary Scientific GeoConference SGEM 2017, www.sgemviennagreen.org, SGEM2017 Vienna GREEN Conference Proceedings, ISBN 978-619-7408-27-0 / ISSN 1314-2704, 27 - 29 November, 2017, Vol. 17, Issue 33, 111-118 pp; DOI: 10.5593/sgem2017H/33/S12.014

ABSTRACT

This work discusses the creation of a river flow generator, which uses as an inputs temperature and precipitation data. These input data could be an artificial series of mentioned climatic variables (especially in context of climate change studies), but when assessing the present-day state, measured data may be used too. For this purpose hydrological model and machine learning model is used. In order to increase the accuracy of the flow generation, the authors have designed and tested various variants of data driven models, which are described in the paper. As is demonstrated in the case study, better modelling results are more affected by hydrology-inspired feature engineering then by searching for better data driven method. Proposed approach allows for a significant increase in the accuracy of the flow generation in comparison to hydrological and machine learning models in their standard application and using standard inputs. The proposed model is compared with the hydrological model and other variants of machine learning models. Comparison is made on stream flows in Slovakia.

Keywords: streamflow forecasting, hydrologic model, data driven modelling, gradient boosting machines