Supervision: Gionata Ghiggi,Daniele Grattarola

Project type: Master thesis

Available

Background

The high-resolution short-range forecasting (0–6h ahead) of precipitation (nowcasting) is an essential component of severe weather and hydrological early warning systems. Timeliness and accuracy of such forecasts are essential for weather-dependent decision-making to guarantee infrastructure safety and socio-economic operations. State-of-the-art operational nowcasting methods typically advect precipitation fields with radar-derived displacement estimates, but they struggle to capture the initiation of new convective cells and the growth and decay of precipitation cells, especially in regions with complex orography.

Project

In the past years, the community-driven open-source software Pysteps (Pulkinnen et al., 2019) has become the reference standard to benchmark the skill of precipitation nowcasting statistical algorithms. However, the recent work of Ravuri et al., 2021 has set a new milestone by demonstrating the potential superiority of Conditionally Generative Adversarial Network (CGAN) models for nowcasting applications compared to previous-generation Pysteps algorithms. In collaboration with MeteoSwiss, the past year we established a framework to facilitate the design and training of autoregressive convolutional deep learning models for short-term precipitation forecasting over the Swiss territory.

In this master thesis project, the student will investigate the extension of such a framework for the training of deep learning generative models (i.e. CGAN) and the ability to produce a large set of realistic spatio-temporal realizations characterizing the short-term evolution of precipitation over the Swiss territory. The accuracy of the precipitation forecasts will be benchmarked against a comprehensive set of established nowcasting algorithms as well as the operational NowPrecip product of MeteoSwiss (Sideris et al., 2020).

Objectives

  • Development and training of CGAN for precipitation nowcasting
  • Probabilistic verification of the ensemble nowcast

Requirements

  • Good programming skills in Python
  • Previous knowledge of deep learning and PyTorch

Supervisors

Gionata Ghiggi: gionata.ghiggi@epfl.ch

Daniele Grattarola: daniele.grattarola@epfl.ch

References

Pulkkinen, S., Nerini, D., Pérez Hortal, A. A., Velasco-Forero, C., Seed, A., Germann, U., and Foresti, L., 2019. Pysteps: an open-source Python library for probabilistic precipitation nowcasting (v1.0), Geosci. Model Dev., 12, 4185–4219. https://doi.org/10.5194/gmd-12-4185-2019

Ravuri, S., Lenc, K., Willson, M. et al. 2021. Skillful precipitation nowcasting using deep generative models of radar. Nature 597, 672–677. https://doi.org/10.1038/s41586-021-03854-z

Sideris, I. V., L. Foresti, D. Nerini, and U. Germann, 2020: Nowprecip: Localized precipitation nowcasting in the complex terrain of Switzerland. Quart. J. Roy. Meteor. Soc., 146, 1768–1800. https://doi.org/10.1002/qj.3766