Sign in Subscribe
Water resource management

Researchers Develop Novel Approach for Data Assimilation in Snow Hydrology

Strategic Insights

— 1 min read

A team of researchers at the CIMA Research Foundation has developed a novel approach for data assimilation in snow hydrology using Long Short-Term Memory (LSTM) networks. This approach aims to improve the accuracy of snow water equivalent and snow depth estimates while reducing computational demands. The research, published in The Cryosphere, found that the LSTM-based data assimilation framework achieved comparable performance to state estimation based on the Ensemble Kalman Filter (EnKF) with a small performance drop in terms of Root Mean Square Error (RMSE).

Key Takeaways:

  • The LSTM-based data assimilation framework achieved comparable performance to the EnKF in improving open-loop estimates with a small performance drop in terms of RMSE for snow water equivalent (+6 mm on average) and snow depth (+6 cm).
  • The inclusion of a memory component further enhanced LSTM stability and performance, particularly in situations of data sparsity.
  • The framework showed promising spatial transferability, with less than a 20% reduction in accuracy for snow water equivalent and snow depth estimation when trained on long datasets (25 years).
  • The research concluded that the approach is robust across various climate regimes and shows limited drop in performance compared to the EnKF.
  • Funding for the research was provided by the Horizon Europe European Research Council, European Space Agency, and National Aeronautics And Space Administration.

Statistics:

  • The LSTM-based data assimilation framework achieved a 70% reduction in computational time compared to a parallelized EnKF.
  • The framework showed less than a 20% reduction in accuracy for snow water equivalent and snow depth estimation when trained on long datasets (25 years).
  • The EnKF-based state estimation had an RMSE of +18 cm for snow depth and +6 mm for snow water equivalent.
  • The LSTM-based data assimilation framework achieved an RMSE of +6 mm for snow depth and +6 cm for snow water equivalent.

Sources:

  • "Learning to filter: snow data assimilation using a Long Short-Term Memory network." The Cryosphere, 2025, 19():4759-4783.
  • VerticalNews editors (2025). Data on Information and Data Aggregation Discussed by Researchers at CIMA Research Foundation (Learning to filter: snow data assimilation using a Long Short-Term Memory network). Information Technology Newsweekly. November 4, 2025; p 104.

Read more