Thematic Modules: Environment


Climate analytics and data processing


Providing a python package implementing an unsupervised Deep Learning method, a Convolutional Variational Auto-Encoder (CVAE) that can characterise generic climate extreme events

Base methods and functions to provide the extraction of generic characteristics of climate extremes. It will enable users to explore the impacts on extreme events on specific users’ applications in the context of selected climate simulations.

Release Notes

The development of the xtclim Python package is currently under heavy development. The level of the software can be considered alpha, as many aspects are still under development.

Currently the xtclim package can identify generic combined climate extremes of temperature, wind, and precipitation for four seasons using a Convolutional Variational Auto-Encoder (CVAE) using pyTorch, on global climate simulations over a selected geographical region.

The main workflow consists of data pre-processing to tailor data appropriately for the CVAE method, training, then applying the method. End-user products are then generated at the last step of the workflow such as plots and graphs.

Future Plans

First Release will happen no later than the end of the first quarter of 2024.

  • Assess required training sample size.
  • Validate against icclim results (analytical method).
  • Fix issues related to the land-sea mask.
  • Implement spatial extent characteristic retrieval.
  • Implement time duration characteristic retrieval.
  • Assess the method performance (computing time, memory needed).
  • Assess different climate variables.
Target Audience
  • DT Developers
  • Expert Scientists

Apache 2

Created by

Related Use Cases

interTwin Use Case: Extreme rainfall, temperature and wind – weather extremes as a result of climate change

Using AI to create a Digital Twin Application that can detect and characterize the changes...