Computer Vision Flood Forecasting using Remotely Sensed Data
Regional flood forecasting and modelling is a big problem in the hydrological sciences. The issue is even more pronounced in Bosnia and Herzegovina due to inadequate hydrometeorological monitoring network, where only 100 out of 500 rainfall stations remained operational after the war.
The project solicits the latest technology from different fields: Remote Sensing to acquire Earth Observations (EO) data, Computer Vision to process these data and Machine Learning (ML) to develop models in order to relate meteorological and other EO data with flooding. The approach is deployed to address one of the most pressing problems B&H currently faces – increasingly frequent occurrence of floods, and shall result in accurate models of rainfall–runoff behaviours, which would help Water Agencies and the Federal Hydro Meteorological Institute with flood prediction and management.
The project is implemented by GISAT, a SME located in the Czech Republic providing a complete portfolio of value-added, high quality and state-of-the-art geoinformation services based on the EO technology and a growing portfolio of services related to innovative spatial data integration, visualization, exploration and analysis using web and mobile platforms.