Regionalization of Rainfall Regimes Using Hybrid RF-Bs Couple with Multivariate Approaches

ISPRS International Journal of Geo-Information

Sumayyah Aimi Mohd Najib¹  Sri Andayani²  Muhamad Afdal Ahmad Basri³  Nurul Hila Zainuddin³  Shazlyn Milleana Shaharudin³  Kismiantini⁴  Mou Leong Tan⁵ 

[1] Department Geography and Environment, Faculty of Human Sciences, Universiti Pendidikan Sultan Idris, Tanjong Malim, Perak 35900, Malaysia;Department of Mathematics, Faculty of Mathematics and Natural Sciences, Universitas Negeri Yogyakarta, Yogyakarta 55281, Indonesia;Department of Mathematics, Faculty of Science and Mathematics, Universiti Pendidikan Sultan Idris, Tanjong Malim, Perak 35900, Malaysia;Department of Statistics, Faculty of Mathematics and Natural Sciences, Universitas Negeri Yogyakarta, Yogyakarta 55281, Indonesia;GeoInformatic Unit, Geography Section, School of Humanities, Universiti Sains Malaysia, Gelugor, Pulau Pinang 11800, Malaysia;

关键词: rainfall;    principal component analysis (PCA);    hierarchical clustering analysis (HCA);    imputation method;    random forest-bootstrap algorithm (RF-Bs);   

DOI  :  10.3390/ijgi10100689

来源: DOAJ

【 摘 要 】

Monthly precipitation data during the period of 1970 to 2019 obtained from the Meteorological, Climatological and Geophysical Agency database were used to analyze regionalized precipitation regimes in Yogyakarta, Indonesia. There were missing values in 52.6% of the data, which were handled by a hybrid random forest approach and bootstrap method (RF-Bs). The present approach addresses large missing values and also reduces the Root Mean Square Error (RMSE) and the Mean Absolute Error (MAE) in the search for the optimum minimal value. Cluster analysis was used to classify stations or grid points into different rainfall regimes. Hierarchical clustering analysis (HCA) of rainfall data reveal the pattern of behavior of the rainfall regime in a specific region by identifying homogeneous clusters. According to the HCA, four distinct and homogenous regions were recognized. Then, the principal component analysis (PCA) technique was used to homogenize the rainfall series and optimally reduce the long-term rainfall records into a few variables. Moreover, PCA was applied to monthly rainfall data in order to validate the results of the HCA analysis. On the basis of the 75% of cumulative variation, 14 factors for the Dry season and the Rainy season, and 12 factors for the Inter-monsoon season, were extracted among the components using varimax rotation. Consideration of different groupings into these approaches opens up new advanced early warning systems in developing recommendations on how to differentiate climate change adaptation- and mitigation-related policies in order to minimize the largest economic damage and taking necessary precautions when multiple hazard events occur.

【 授权许可】

Unknown