Robustness and evolution of climate types in France between 1961 and 2020
Abstract
This study assesses the robustness of the climate classification proposed by Joly et al. (2010) in the context of climate changes observed in France between 1961 and 2020. Using SAFRAN reanalysis data, we applied this climate typology to an 8km grid, whereas the original classification was based on interpolated observations at a 250m resolution. We tested two classification methods: a Naïve Bayes model, similar to that used in the original study, and a Random Forest model, which is better suited to capturing complex interactions among climate variables. Our results show that the Random Forest model reproduces the Joly et al. (2010) classification applied to SAFRAN data much more accurately, achieving an overall accuracy of 95%, compared with 23.7% for the Naïve Bayes model. This result indicates the greater robustness of the Random Forest approach to changes in both spatial resolution and data source. The analysis of four successive climate normals (1961-1990, 1971-2000, 1981-2010 and 1991-2020) reveals several major climate dynamics. Between the first and last reference periods, mean annual temperature increased by 0.9°C, reflecting the ongoing warming trend, with particularly pronounced increases in southern and eastern France. Oceanic influence strengthened in the northwestern part of the country, contributing to reduced seasonal temperature contrasts and higher annual precipitation. At the same time, Mediterranean climate characteristics expanded, particularly along the Rhône Valley, in Languedoc and in Corsica, accompanied by a marked increase in the frequency of very hot days and shifts in precipitation seasonality. These findings indicate that the Joly et al. (2010) classification remains broadly relevant for characterizing climate types in France. However, transitional areas, particularly those in semi-continental and Mediterranean regions, appear to be the most sensitive to ongoing climatic changes. As such, they provide valuable indicators of the progressive reorganization of climate types under recent climate change.
Related
- Determining the exposure of forest production zones to climate change in France?
- Climatology and trends of extreme precipitation in France: evaluation of an explicit-convection regional climate model
- Regional Impact of Climate Change on the Spatio-Temporal Characteristics of Droughts in France
- Evolution and Analysis of Landslides in Lowland Areas: The Case Study of Reuil in the Champagne Vineyard Region (Marne, France)
- Climate Change and the Evolution of Cultural Transmission in the Human Lineage
BibTeX
@article{hilal2026robustness,
title = {{Robustness and evolution of climate types in France between 1961 and 2020}},
author = {Mohamed Hilal and Daniel Joly and Léa Laurent and Yves Richard},
journal = {Cybergeo : Revue européenne de géographie / European journal of geography},
year = {2026},
month = oct,
url = {https://hal.inrae.fr/hal-05800885v1},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Robustness and evolution of climate types in France between 1961 and 2020
AU - Mohamed Hilal
AU - Daniel Joly
AU - Léa Laurent
AU - Yves Richard
PY - 2026
DA - 2026/10/13
JO - Cybergeo : Revue européenne de géographie / European journal of geography
UR - https://hal.inrae.fr/hal-05800885v1
AB - This study assesses the robustness of the climate classification proposed by Joly et al. (2010) in the context of climate changes observed in France between 1961 and 2020. Using SAFRAN reanalysis data, we applied this climate typology to an 8km grid, whereas the original classification was based on interpolated observations at a 250m resolution. We tested two classification methods: a Naïve Bayes model, similar to that used in the original study, and a Random Forest model, which is better suited to capturing complex interactions among climate variables. Our results show that the Random Forest model reproduces the Joly et al. (2010) classification applied to SAFRAN data much more accurately, achieving an overall accuracy of 95%, compared with 23.7% for the Naïve Bayes model. This result indicates the greater robustness of the Random Forest approach to changes in both spatial resolution and data source. The analysis of four successive climate normals (1961-1990, 1971-2000, 1981-2010 and 1991-2020) reveals several major climate dynamics. Between the first and last reference periods, mean annual temperature increased by 0.9°C, reflecting the ongoing warming trend, with particularly pronounced increases in southern and eastern France. Oceanic influence strengthened in the northwestern part of the country, contributing to reduced seasonal temperature contrasts and higher annual precipitation. At the same time, Mediterranean climate characteristics expanded, particularly along the Rhône Valley, in Languedoc and in Corsica, accompanied by a marked increase in the frequency of very hot days and shifts in precipitation seasonality. These findings indicate that the Joly et al. (2010) classification remains broadly relevant for characterizing climate types in France. However, transitional areas, particularly those in semi-continental and Mediterranean regions, appear to be the most sensitive to ongoing climatic changes. As such, they provide valuable indicators of the progressive reorganization of climate types under recent climate change.
SN - 1278-3366
ER -