River Discharge estimation from optical satellite data: latest advances using NIR sensors

River discharge is worldwide recognized as one of the variables whose knowledge is most needed in order to understand the evolution of climate, to assess the related risks and to develop mitigation and adaptation strategies. Notwithstanding this, river monitoring is still an open issue. The ground monitoring network is declining, with many rivers that are ungauged due to the difficulties of installing instruments on remote areas and the high costs of installation and maintenance of instruments. Furthermore, the absence of strategies for data sharing and the long latency in data dissemination worsen the situation, preventing the use of methods for natural hazard forecasting in many regions.

 In this framework, over the last few decades, satellite data have been used to support the ground network information thanks to the strong growth in technologies, data processing and applications that fostered their use for the water cycle monitoring. In particular, considering the daily river discharge measurement, the recent advances in near-infrared (NIR) satellite sensors encouraged their use for the river discharge estimation, due to their frequent revisit time and wide spatial coverage. Therefore, passive remote sensing data from multiple sensors such as MODIS, MERIS and OLCI (spatial resolution of about 250 - 300 m) have been used to develop a non-linear regression model to estimate the river flow in medium-sized catchments (around 100’000 km²). The model is based on the different behavior in the NIR band between a calibration pixel C, selected over land, and a measurement pixel M, selected over the river boundaries. The ratio of the two pixels is indeed well correlated with in situ river discharge, but the methodology still needs to calibrate the pixel locations by using observed data, limiting the usefulness of the methodology to the gauged areas.

More recently, Sentinel-2 satellites of the European Union’s Earth observation COPERNICUS programme, foster the monitoring of narrow rivers (< 150 m wide) thanks to the high spatial resolution (10 m). The high resolution enables to better identify the main geographical features (e.g., water, vegetation, urban area, river boundaries) and to better monitor the effect of several factors (vegetation and sediments) in the river discharge estimation. An important contribution has been found in the sediment component, affecting the reliable reproduction of high flow due to the high reflectance of turbid water sensed by the satellite. For this reason, the original formulation for the estimation of river discharge has been modified and tested over several rivers worldwide to assess its influence in different environments.

Here, we show the results of the analysis applying the new approach for the estimation of the river discharge to both Sentinel-2 and MODIS data, in order to evaluate the advantages of the use of high spatial resolution information. Furthermore, results and limitations of the uncalibrated version of the algorithm are also shown underling the possibility to use the methodology over ungauged rivers, where the absence of observed data prevents the applicability of the classical satellite methods for river discharge estimation.