Loading...

Climate Classification of the MENA (Middle East and North Africa) by Introducing a New Index for Clustering Validation

Rajabi, Reza | 2020

548 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 53481 (09)
  4. University: Sharif University of Technology
  5. Department: Civil Engineering
  6. Advisor(s): Moghim, Sanaz
  7. Abstract:
  8. Clustering presents valuable information in discovery of the climatic zones. To use clustering approaches, similarity measure, clustering algorithm, and clustering validity index should be determined. To find climatic zones over Middle East nad North Africa (MENA), this study performs k-means clustering with Euclidean distance as the similarity measure on four monthly precipitation datasets (CRU, GPCC, UDEL, and PREC/L) and two monthly temperature datasets (CRU, NOAA GHCN-CAMS). This study aims to validate clustering results and find a proper number of clusters. For this purpose, five traditional validity indices are examined on experimental datasets. Results show significant differences between suggested optimal clusters for the experimental datasets. Furthermore, complexity of the climate datasets (data without clear cluster boundaries) encouraged us to introduce a more robust validity index. Thus, concepts of spectral clustering and neighborhood network of data points are used to adjust DB index such that the new validity index (DBR) can control the two main clustering problems (unnecessary cluster merging and unnecessary cluster division). The DBR index suggests 6, 6, 5, and 4 precipitation clusters for CRU, GPCC, PREC/L, and UDEL, respectively. In addition, this index proposes 3 and 5 temperature clusters for CRU and NOAA GHCN-CAMS datasets. Results show that the GPCC clustering results are more consistent with those of the Köppen-Geiger classification system and also previous studies. Combining precipitation clusters of the GPCC with temperature ones obtained from CRU and NOAA GHCN-CAMS leads to 16 and 20 different climatic zones over MENA region, respectively
  9. Keywords:
  10. Clustering ; Themperature ; Validation ; K-means Clustering ; Precipitation ; Density-based Spatial Clustering Applicoction with Noise (DBSCAN) ; Davies-Bouldin (DB)Index ; Climatic Zones ; Middle East nad North Africa (MENA)Zone

 Digital Object List

 Bookmark

No TOC