A new approach for flood risk estimation integrating remote sensing and in-situ data

A lot of different methods are used to estimate flood risk worldwide. The method that performs better depends on the catchment features and dimension, data time resolution and availability and uncertainty level required. Remote sensing approaches are more and more common, but because of the limited periods covered by time series derived by this new methodology, in-situ data integration is still required. A new methodology is proposed, based on a case-study of different reaches of Basento river, Basilicata (Southern Italy). Starting from hourly rainfall time series (covering not less than 20 years), for each pluviometric station taken into account into the catchment area, Intensity-Duration-Frequency (IDF) curves are computed (fitting a power law), in order to calculate the rainfall maximum at a certain percentile (typically 90° or 95° percentile are used) during the concentration time. Thiessen polygon method is used to divide the catchment area into smaller areas, each one corresponding to a pluviometric station, with the purpose of calculating weighted  rainfall values for each station area. A Digital Terrain Model is used to extract multiple cross sections of the river-bed, spanning different morphologies, from braided to meandering channels. For each cross section, starting from bankful level, it is possible to estimate diverse hydraulic parameters such as river stage, hydraulic radius, section’s surface area (using image analysis) and the mean velocity of the current, using the logarithmic law profile of the turbulent flow. Sediment size analysis is carried out as to estimate the river bed roughness for each cross section. The mean velocity value V can be used to estimate the concentration time t=L/V, where L is equal to the distance between the cross section and the hydraulically further point into the catchment area. The concentration time value t is used into the equation of the IDF curves, in order to link the corresponding rainfall height to the river stage reached at the cross section, eventually to estimate the rainfall value that, if exceeded, can cause flood. A FLO-2D model has been then used to run simulations with the aim to detect flood-prone areas, finding an overall good matching between the values of current mean velocity, discharge and river stage estimated in the cross sections.