Séminaire OSE - Observations Signal & Environment

OSE
Jeudi 30.03.2023
Horaires :
De 14:00 à 16:00

Adresse :

Visioconférence

https://imt-atlantique.webex.com/imt-atlantique/j.php?MTID=m5cbdfd9904392d980e66d8ff59a06285

Public cible

Enseignants chercheurs
Doctorants
Elèves ingénieurs 3ème année

Title: Combining machine learning and extreme value theory to characterize CMIP6 ensembles' respective performances - Application to Sea Surface Temperature in the Mediterranean Sea

Abstract: Climate models are useful tools to investigate the future of environmental variables of interest. However, they are plagued with structural uncertainty: a computational representation of the dynamics and the variability of the ocean-earth system is necessarily incomplete and truncated at some scales. The choice of prognostic equations, the numerical methods to solve them, the selection of processes to be parameterized, their physical parameterization, not to mention the initialization, all participate in the evolution of climate model outputs for a given forcing scenario. Model ensembles have become common practice to both assess this structural uncertainty and the inherent chaotic variability of the climate system. However, the question of how to compare the performance of these ensembles, let alone how to use them in practice beyond the popular duo of "ensemble mean" and “ensemble anomalies”, remains challenging.
On the case study of the Mediterranean sea surface temperature and marine heat waves, we leverage functional clustering and extreme value theory, combined with the Wasserstein distance, to produce an assessment and comparison tool for a suite of climate model outputs (namely CMIP6 ensembles) and a reference (30-year of reprocessed observations from the High Resolution L4 Sea Surface Temperature Copernicus Mediterranean dataset). This tool allows in particular the joint visualization of model performance w.r.t. the bulk and extreme behaviors of the variable of interest respectively, as well as displaying the spatial intra-model and inter-model variability. In light of this assessment, an approach based on multi-criteria decision-making is implemented and discussed, offering another operational combination of multiple model outputs and an alternative interpretation of model ensembles.

Keywords: Model validation, Model ranking, Extremes, Functional clustering, Marine heat waves.

Authors:  N. Le Carrer, P. Tandeo, F. Sevellec, C. Gaetan, P. Girardi,  P. Naveau

 

En savoir +

Voici le lien Webex et les codes d'accès pour la visio

Meeting link:
https://imt-atlantique.webex.com/imt-atlantique/j.php?MTID=m5cbdfd9904392d980e66d8ff59a06285

Meeting number:
2732 336 2015

Meeting password:
Fm9uSWRQ3P3

Join from a video or application
Dial 27323362015@imt-atlantique.webex.com
You can also dial 62.109.219.4 and enter your meeting number.

Join by phone
+33-1851-48835 France Toll
+33-1-7091-8646 France Toll 2
Access code: 27323362015

Global call-in numbers
https://imt-atlantique.webex.com/imt-atlantique/globalcallin.php?MTID=m5ea6fc5b1eb4fa7b35291648dd33bb47

Organisateur(s)

Publié le 28.03.2023
 
 
 
1
 
2
 
3
 
4
 
5
 
6
 
7
 
8
 
9
 
10
 
11
 
12
 
13
 
14
 
15
 
16
 
17
 
18
 
19
 
20
 
21
 
22
 
23
 
24
 
25
 
26
 
27
 
28
 
29
 
30
 
31