Abstract
Data assimilation of snow observations significantly improves the accuracy of snow cover simulations. However, remotely-sensed snowpack observations made in areas of complex topography are typically subject to large error and biases, creating a challenge for data assimilation. To improve the reliability of ensemble snowpack simulations, this study investigated the appropriate conditions for assimilating MODIS-like synthetic surface reflectances. We used a simulation system that included the Particle Filter data assimilation technique. More than 270 ensemble simulations involving assimilation of synthetic observations were conducted in a twin experiment procedure for three snow seasons. These tests were aimed at establishing the spectral combination of MODIS-like reflectances that convey the more information in the assimilation system, rendering the most reliable snowpack simulation, and determining the maximum observation errors that the assimilation system could tolerate. The assimilation of the first seven MODIS-like bands, covering visible and near-infrared wavelengths, provided the best scores compared with any other band combination, and thus are highly recommended for use when possible. The simulation system tolerated a maximum deviation from ground truth of 5% without loss of performance. However, the assimilation of the first seven bands of true MODIS surface of reflectance fails on improving simulation results in rouged mountain areas.
Highlights
Some of the major natural hazards in mountain areas are directly linked to the snowpack distribution and its evolution over time, and include snow avalanches (Schweizer et al, 2003, 2008) and floods in downstream areas (Gaal et al, 2015)
As the synthetic observations of surface reflectances were not used for evaluating the simulation, and were only considered in the simulation system when a potential satellite observation was possible, these observations are only included in the graph when the assimilation took place
Our results demonstrate that the assimilation of snowpack surface reflectances using the Particle Filter algorithm improves simulation of the temporal evolution of snowpack bulk variables
Summary
Some of the major natural hazards in mountain areas are directly linked to the snowpack distribution and its evolution over time, and include snow avalanches (Schweizer et al, 2003, 2008) and floods in downstream areas (Gaal et al, 2015). Snowpack modelling is affected by forcing and model errors that result in discrepancies between the real state and the simu lated snowpack (Morin et al, 2020) The accumulation of such dis crepancies over time decreases forecasting capabilities, leading to large uncertainties in risk management. Active satellite sensors have been assimilated in large study areas (Cortes et al, 2016; Larue et al, 2018; Margulis et al, 2019), and in some cases combined with in–situ observations (Piazzi et al, 2019), showing encouraging results These works demonstrate the effort of the snow modeling community on assimilating snow observations to improve forecasting capabilities of snow models. Remote sensing techniques capa bility of retrieving information over extended (and remote) areas is the main reason that has motivated this attempt
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
