IMPORTANCE SAMPLING
Importance sampling is a way to preferentially select out of a range of model simulations the ones that will lead to extremes of the metric of interest in an early stage of simulation, thus increasing computing efficiency for the cases of interest.
Description
Importance sampling is a method of statistical physics that nudges a system (like climate) towards extremes of a given observable (like high temperatures over a region). The idea is to make ensembles of simulations of the system for a short time, delete the simulations that do not increase the chosen observable and replicate those which do increase the observable. This is similar to particular filtering, with weights on temperature.
This procedure has been applied to the IPSL climate model to simulate extremes of temperatures over France.
This type of method is a broad case of “ensemble boosting”. It can be used to design “worst case scenarios”. In the present version, this method is designed for a global climate model (GCM) and requires supercomputing facilities.
Potential User Groups
- Long-term investment in energy sector: French RTE.
- Agriculture & agronomy: French INRAE
- Collaborations with research institutes with access to supercomputing facilities is necessary.
Guide
PhD thesis of Robin Noyelle (https://cnrs.hal.science/tel-04632646/).
Availability
Code is available upon request once the paper (submitted to J. Clim) is accepted. Such codes are meant to run on supercomputing facilities.
Use Cases
Robin Noyelle. Statistical and dynamical aspects of extreme heatwaves in the mid-latitudes. Ocean, Atmosphere. Université Paris-Saclay, 2024. English. ⟨NNT : 2024UPASJ013⟩. ⟨tel-04632646v2⟩
This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 101003469.
Tools
The Artificial Intelligence for Disentangling Extremes (AIDE) toolbox allows for tackling generic problems of detection and impact assessment of events such as tropical cyclones and severe convective storms, heat waves, and droughts, as well as persistent winter extremes, among others
CAUSEME web-platform is a platform to benchmark causal discovery methods based on ground truth benchmark datasets featuring different real data challenges.
Ensemble boosting uses climate models to efficiently generate very intense and rare weather and climate extremes that can be analyzed for planning and stress testing of critical infrastructure.
Importance sampling is a way to preferentially select out of a range of model simulations the ones that will lead to extremes of the metric of interest in an early stage of simulation, thus increasing computing efficiency for the cases of interest.
We contributed to the continuous development of the tigramite (https://github.com/jakobrunge/tigramite/) Python package for causal inference methods.
This package provides a wide range of constraint-based causal discovery and causal effect estimation methods.
In the aftermath of an extreme weather event, the event is studied more and more often by an operational attribution service. Two examples of such operational attribution services are used within XAIDA. These operational services generally answer slightly different questions and are complementary.
ClimaMeter is an international consortium designed to
This tool explores potential energy risk scenarios using weather regimes. This allows users to tailor the tool to their relevant impact and weather regimes.
Stochastic Weather Generation is a computationally light tool to simulate temperatures for worse case heat extremes, at city up to country level.
In the aftermath of an extreme weather event, the event is studied more and more often by an operational attribution service. Two examples of such operational attribution services are used within XAIDA. These operational services generally answer slightly different questions and are complementary.
World Weather Attribution (WWA), an initiative founded
GreenEarthNet is a machine learning-powered toolkit for predicting ecosystem responses to climate changes by leveraging Earth observation data and climate models. It supports effective environmental monitoring and forecasting for improved climate action.