Methodology

The approach adopted and demonstrated on Bregava river catchment was to extract information relevant to rainfall–runoff behaviours from remotely sensed observable catchment attributes. Application of deep learning helped to map meteorological time series into streamflow that itself depends on a set of static catchment characteristics and dynamic meteorological properties and results into an accurate flood forecast model. The proposed solution highly relies on EO data inputs and derived products as Land Cover or vegetation characteristics and EO modelled meteorological variables. Compared to other modelling approaches, the CV has many advantages with regard to fusing spatiotemporal data. This project considered precipitation (P), temperature (T), and LAI data fields as the inputs of the data-driven models resulting in flood forescast and runoff estimate.