A Tool for Surface Reflectance Retrieval from Google Earth Engine for the estimation of river flow variation
Abstract
River discharge monitoring is crucial for understanding and mitigating the impacts of climate change, such as changes in the frequency and duration of floods and droughts. However, the high costs associated with field surveys and the installation of river gauge networks make it difficult to obtain real-time observations worldwide. Satellite observations are therefore increasingly used, thanks to advancement in sensor technology. This study presents a methodology for estimating river discharge using the so-called Calibration/Measurement (CM) approach, which leverages multispectral data from the Sentinel-2 (S2) and MODIS sensors, processed through the Google Earth Engine (GEE) platform. The developed Python code is made public to encourage open community development. The methodology was first adapted to various river characteristics, such as sediment transport and the presence of riverside vegetation, and then applied to 62 river sites worldwide to test its robustness and applicability. The results confirm that the CM approach can be used to reliably estimate river discharge, with Sentinel-2 data providing superior performance (median 0.9 Spearman correlation) compared to MODIS (median 0.73 Spearman correlation) due to its higher spatial resolution. Nevertheless, the analysis also identifies challenges to our methodology, such as cloud shadows, unmasked clouds and image co-registration issues, which particularly affect small rivers, offering change opportunities in the masking procedure for more accurate discharge estimates.
