CLIMATE COMFORT INDEX ANALYSIS USING SPATIO-TEMPORAL PCA-FASTMCD METHOD

Abstrak

This study addresses the challenge of modeling spatio-temporal climate data that are often affected by outliers, which significantly bias conventional principal component analysis. The main contribution of this research is not merely the application of Spatio-Temporal Principal Component Analysis (STPCA), but its novel integration with the Fast Minimum Covariance Determinant (FASTMCD) method to obtain robust spatio-temporal components that are resilient to outliers in Bali’s climate data. The core methodology involves transforming four climate variables (thermal comfort, cloud cover, rainfall, and wind speed) from 24 stations in Bali (2010–2019) using Fourier basis expansion, applying spatial weighting, and utilizing robust covariance estimation via FASTMCD. The results indicate that the Inverse Power Distance (IPD) weighting scheme optimally captures the spatial structure. Furthermore, the first robust principal component (STPC1) reveals the dominant climate variability, which is driven primarily by thermal comfort and wind speed. This component successfully highlights a clear spatial differentiation between coastal lowland and highland regions despite the presence of extreme observations. These findings imply that the robust STPCA-FASTMCD framework provides a highly stable representation of regional climate patterns, offering a reliable analytical tool for developing climate comfort indices and supporting climate-informed tourism planning in tropical regions.