Thesis Code: T2

Correction of ERA5 Snow data using a U-NET-based framework and SNOW-IT reanalysis

Short description:

The management of water resources, especially in regions reliant on seasonal snowpacks, hinges critically on accurate and high-resolution data on Snow Water Equivalent (SWE) and snow depth. These variables serve as key indicators of water availability, influencing decisions in agriculture, hydroelectric power generation, and flood risk management. However, the acquisition of reliable and finely resolved SWE and snow depth data poses significant challenges, particularly in mountainous areas where snow accumulation patterns are both critical and complex. The inherent limitations in spatial resolution and accuracy of existing datasets underscore a critical gap in our hydrological understanding and operational capabilities. This thesis is motivated by the need to enhance the resolution and accuracy of SWE and snow depth data, leveraging downscaling and bias correction methodologies to better inform water management practices. 

To improve the accuracy and resolution of snow data in mountainous territories, our approach involves the development of a U-Net-based bias correction framework [1]. This framework is designed to process spatial data derived from the ERA5 reanalysis dataset [2], which is acclaimed for its comprehensive global coverage and consistent temporal data, but presents coarse spatial resolution and significant biases for complex terrain regions. By employing SNOW-IT as a reference dataset [3], known for its accurate and region-specific snow measurements across Italy, we aim to refine and scale down ERA5 data. Once optimized, this framework has the potential to be extrapolated to additional Alpine areas outside of the regions covered by SNOW-IT.

To achieve these goals, we expect the student to carry out the following activities:

  1. Literature review: reviewing the state-of-the-art for downscaling and bias correction in geosciences application, with a focus on hydrology tasks. Reviewing of spatial snow climatology dataset for the region of interest.
  2. Data Collection: acquisition of SWE data from ERA5 and SNOW-IT reanalysis for the region considered. Inspection of the data. Create the training dataset for the network. 
  3. Computational experiments: train the U-net-based framework to correct and downscale the input data from ERA5. Test the model on regions that the model has never seen. 

References:

[1] Ronneberger, O., Fischer, P., & Brox, T. (2015). U-Net: Convolutional Networks for Biomedical Image Segmentation. Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015, 9351, 234-241. DOI: 10.1007/978-3-319-24574-4_28 

[2] Hersbach, H., Bell, B., Berrisford, P., Hirahara, S., Horányi, A., Muñoz-Sabater, J., Nicolas, J., Peubey, C., Radu, R., Schepers, D., Simmons, A., Soci, C., Abdalla, S., Abellan, X., Balsamo, G., Bechtold, P., Biavati, G., Bidlot, J., Bonavita, M., … Thépaut, J.-N. (2020). The ERA5 global reanalysis. Quarterly Journal of the Royal Meteorological Society, 146(730), 1999-2049. DOI: 10.1002/qj.3803[3] Avanzi, F., Gabellani, S., Delogu, F., Silvestro, F., Pignone, F., Bruno, G., Pulvirenti, L., Squicciarino, G., Fiori, E., Rossi, L., Puca, S., Toniazzo, A., Giordano, P., Falzacappa, M., Ratto, S., Stevenin, H., Cardillo, A., Fioletti, M., Cazzuli, O., Cremonese, E., Morra di Cella, U., & Ferraris, L. (2023). IT-SNOW: a snow reanalysis for Italy blending modeling, in situ data, and satellite observations (2010–2021). Earth System Science Data, 15, 639–660. DOI: 10.5194/essd-15-639-2023

Relevant courses and knowledge:

Natural Resources Management

Number of students:

1 or 2

Requisites:

The student should be comfortable with data handling and programming skills in Python and have some experience with Neural Networks (preferably convolutional networks and TensorFlow library).