Regional flood forecasting and modelling is an old yet largely outstanding problem in the hydrological sciences. The problem is even more pronounced in Bosnia & Herzegovina (BiH) due to inadequate hydrometeorological monitoring network. In the broad area of earth sciences, new remote-sensing systems, as well as ground-based sensor-network systems, are producing environmental big data with high spatial and temporal resolutions.

We brought together advances in remote sensing, and processed the data using machine learning and computer vision techniques to produce accurate yet cost efficient solution for flood prediction demonstrated on Bregava river catchment.