Skip to Main content Skip to Navigation
Journal articles

Classification climatique méditerranéenne pour l'hydrologie

Abstract : The Mediterranean region is a temperate zone where various populations live, making it one of the most sensitive regions to anthropic and climatic changes. The objective of this study is to establish a fine climate classification in the Mediterranean region that highlights the climatic continuity from one place to another and could evaluate changes trends. The proposed approach includes a Principal Component Analysis to reduce the number of climate indices to consider only the most contributory, a classification in K-means to distribute the stations into 5 classes here and finally the construction of a decision tree based on the distances to the climatic classes kernels to determine whether a place belongs to a Mediterranean climate or not, and to which type it belongs to if so. Data from 144 stations at monthly time steps in 20 Mediterranean countries, inside and outside the watersheds periphery are used to calibrate the classification, 36 stations to verify it within Mediterranean region and 21 stations to verify it on regions known for their Mediterranean climate as Chile and South Africa. The results show that the distribution into 5 classes coincides with a geographical distribution in the Mediterranean and the interclass connectivity analysis shows that the evolution of the climate is done in a continuous way, since the same station can obey the criteria of membership of several classes.
Complete list of metadata

https://hal.inrae.fr/hal-02944621
Contributor : Hélène Lesur <>
Submitted on : Monday, September 21, 2020 - 3:39:48 PM
Last modification on : Friday, February 5, 2021 - 4:02:34 AM

Identifiers

Citation

Antoine Allam, Jinane El Hassan, Wajdi Najem, Claude Bocquillon, Roger Moussa. Classification climatique méditerranéenne pour l'hydrologie. La Houille Blanche - Revue internationale de l'eau, EDP Sciences, 2020, pp.60-69. ⟨10.1051/lhb/2020008⟩. ⟨hal-02944621⟩

Share

Metrics

Record views

25