Assessing Household Climate Resilience Index to Climate Change Impacts in DKI Jakarta: A Robust Geographically Weighted PCA Approach

Marta Sundari (Study Program on Statistics and Data Science, IPB University, Gd. FMIPA Jl. Meranti Kampus Dramaga Bogor)
Kusman Sadik (Study Program on Statistics and Data Science, IPB University, Indonesia 2, Gd. FMIPA Jl. Meranti Kampus Dramaga Bogor)
Aji Hamim Wigena (Study Program on Statistics and Data Science, IPB University, Indonesia 2, Gd. FMIPA Jl. Meranti Kampus Dramaga Bogor)
Anwar Fitrianto (Study Program on Statistics and Data Science, IPB University, Indonesia 2, Gd. FMIPA Jl. Meranti Kampus Dramaga Bogor)
Rizaldi Boer (Department of Meteorology and Geophysics, IPB University 5, Gd. FMIPA Jl. Meranti Kampus Dramaga Bogor)
Zulkarnain Zulkarnain (Department of Forestry, Faculty of Forestry and Environmental Science, Universitas Halu Oleo. Jl. H.E.A. Mokodompit, Kendari 93121, Southeast Sulawesi)

Forum Geografi
Vol. 40 No. 3 (2026): December 2026

Household climate resilience is spatially heterogeneous because households experience different combinations of exposure, sensitivity, and adaptive capacity across urban settings. This study developeds a local Household Climate Resilience Index (HCRI) for DKI Jakarta using Robust Geographically Weighted Principal Component Analysis based on the Minimum Covariance Determinant estimator (Robust GWPCA-MCD). The analysis used 221 mainland household observations from 17 urban villages in five municipalities of DKI Jakarta, based on the 2022 household resilience survey conducted by CCROM-SEAP IPB University and DLH DKI Jakarta. The final index was constructed using 30 indicators representing exposure, sensitivity, incremental adaptation, and transformational adaptation. Classical GWPCA and Robust GWPCA-MCD were compared to evaluate whether robust local covariance estimation improved the local information structure of the index. The results showed that the Robust GWPCA-MCD produced a substantially stronger local PC1 structure than the classical GWPCA. The classical GWPCA generated a median local PC1 variance of only 14.81%, whereas the Robust GWPCA-MCD increased the median local PC1 variance to 83.66%, with values ranging from 63.42% to 99.999%. The selected model used an adaptive bisquare kernel with 96 nearest neighbors, MCD α = 0.75, and an approximate local h-subset of 72 observations. The HCRI classification identified 43 households (19.46%) as extremely low resilience, 12 households (5.43%) as very low, 52 households (23.53%) as low, 58 households (26.24%) as high, and 56 households (25.34%) as very high, with no household classified as extremely high. Spatial mapping showed that the lower and higher resilience categories were unevenly distributed across sampled urban villages. Local loading analysis further revealed that exposure indicators, particularly distance to flood sources, and sensitivity indicators, including housing characteristics, land area, medication expenditure, and post-drought health burden, dominated different local HCRI structures. These findings demonstrate that the Robust GWPCA-MCD provides a statistically robust and spatially interpretable framework for identifying local household resilience profiles and supporting targeted urban climate adaptation planning.