Adaptive assimilation of spatiotemporal sparse satellite-derived snow cover data into hydrological modelling in the Rhône River basin, France

Accurate estimation of spatiotemporal snowpack is crucial for understanding the hydrological processes associated with snowmelt in mountainous regions. Incorporating in-situ and remote sensing observations into physics-based snowpack models through data assimilation techniques can mitigate model uncertainties and improve estimates of snow water equivalent (SWE). However, implementing data assimilation techniques over a large spatial domain remains challenging, due to the sparse and uneven availability of observations across varying spatial and temporal scales. Remote sensing data are also constrained by gaps caused by revisit intervals, cloud cover, and complex topography in mountains. Therefore, this study proposes an adaptive snow data assimilation framework with satellite remote sensing data utilizing the ensemble Kalman filter (EnKF). The adaptive EnKF assimilates daily sparse high-resolution, remote-sensed snow cover data into the snow model of the distributed wflow_sbm hydrological model, applied to the Rhône River basin—a region in France heavily dependent on snow and glacier meltwater for runoff across multiple scales. Using this adaptive EnKF, we simulate spatiotemporal SWE and river runoff in the Rhône River basin from 2016 to 2019. Results demonstrate that snow data assimilation significantly improves streamflow predictions in both spatial and temporal dimensions. Compared to the simulations without assimilation, our model indicates a spatial decrease in snowmelt runoff during winter (October to March) and a spatial increase during the melt season (April to June). These results demonstrate that adaptive data assimilation not only effectively integrates high-resolution satellite data with hydrological models but also enhances the representation of snowmelt processes, leading to more accurate forecasts of river runoff. This approach paves the way for developing snow reanalysis and forecasting tools, seamlessly integrating sparse information from high-resolution satellite observations into physics-based models, offering valuable insights for water resource management in basins governed by snowmelt-driven hydrological processes.