Toward a global scale runoff estimation through satellite observations: the STREAM model

River discharge monitoring is critical for many activities ranging from water resource management to flood risk reduction. Due to limitations of in situ stations (e.g. low station density, incomplete temporal coverage and delays in data access), river discharge is not always continuously monitored in time and space. This has led researchers and space agencies, among others, to develop new methods based on satellite observations for estimating river discharge.

In recent years, ESA has funded the SaTellite-based Runoff Evaluation And Mapping (STREAM) and STREAM-NEXT projects, which propose to use satellite observations of precipitation, soil moisture and terrestrial water storage within a simple and conceptually parsimonious model, STREAM, to estimate runoff.

The model, applied to five large basins in the world (Mississippi-Missouri basin, Amazon basin, Danube basin, Murray-Darling basin and Niger basin)  has demonstrated a high ability to estimate runoff and river discharge in both natural and non-natural basins with a high anthropogenic impact (i.e. in basins where flow is regulated by the presence of dams, reservoirs or floodplains along the river; or in heavily irrigated areas). In particular, the good results obtained paved the way for the application of the STREAM approach on a global scale. For this purpose, the STREAM-NEXT project will generalise the STREAM model to make it applicable to more than forty basins worldwide. Depending on the availability of in situ discharge data, the selected basins shall be grouped into calibration and validation clusters. The purpose is to use the basins into the calibration cluster to tune the parameters of the regionalized STREAM model and apply the regionalised model parameters to the validation cluster basins to estimate the accuracy of the STREAM model.  Additional satellite observations, such as altimetric water levels, will be used to estimate the water stored in the reservoirs; gravimetric data with different spatial/temporal resolutions will be explored to investigate the impact of these data on the model results.

Finally, a calibration procedure and a regionalisation approach will be developed to make the STREAM model applicable to non-calibrated basins.

Here we present the STREAM-NEXT project and some preliminary results related to the generalization of the STREAM model framework. Different basins with different climate, topography and level of anthropisation will be selected to demonstrate the suitability of the approach for a global scale application.