The acquisition of river-level data is a critical task for the understanding of flood events but is often complicated by the difficulty to install and maintain gauges able to provide such information. This study proposes applying deep learning techniques on river-camera images in order to automatically extract the corresponding water levels. This approach could allow for a new flexible way to observe flood events, especially at ungauged locations.

We estimate the contribution of extreme precipitation, antecedent soil moisture and snowmelt to changes in small and large floods across Europe. In northwestern and eastern Europe, changes in small and large floods are driven mainly by one single driver (i.e. extreme precipitation and snowmelt, respectively). In southern Europe both antecedent soil moisture and extreme precipitation significantly contribute to flood changes, and their relative importance depends on flood magnitude.

We investigate changes that occurred in small vs. big flood events and in small vs. large catchments across Europe over 5 decades. Annual maximum discharge series between 1960 and 2010 from 2370 gauges in Europe are analysed. Distinctive patterns of flood regime change are identified for large regions across Europe, which depend on flood magnitude and catchment size.

We investigate the seasonal memory properties of a large sample of European rivers in terms of high and low flows. We compute seasonal correlations between peak and low flows and average flows in the previous seasons and explore the links with various physiographic and hydro-climatic catchment descriptors. Our findings suggest that there is a traceable physical basis for river memory which in turn can be employed to reduce uncertainty and improve probabilistic predictions of floods and droughts.

The knowledge of discharge hydrographs is useful for flood modelling purposes, water resource management, and the design of hydraulic structures. This paper presents a novel methodology to estimate the unknown discharge hydrograph in an ungauged river section using only water level information recorded downstream. A Bayesian procedure is coupled with a 2-D hydraulic model parallelized for GPUs. Finally, the proposed procedure has been applied to estimate inflow hydrographs in real river reaches.