Spatiotemporal Heterogeneity of the Standardized Summer Resort Index Under Normal and Extreme High-Temperature Climates Based on a PCA–Copula Model: A Case Study of Shaanxi Province

This study aims to address the limitations of conventional summer-resort suitability evaluation, which fails to characterize the nonlinear responses of multi-factor variables and suffers from subjectivity in linear weighting, by constructing, based on observational records from 101 meteorological stations across Shaanxi Province spanning 1981–2024, an evaluation system that covers four dimensions (physiological thermal comfort, environmental amenity, extreme-weather risk and climatic stability) with 13 indicators. Group-based principal component analysis (PCA) is applied for dimensionality reduction. A multi-dimensional joint distribution is further established by integrating the empirical cumulative distribution function (ECDF) and the C-vine copula function to derive the Standardized Summer-Resort Index (SSRI). Independent external validation is implemented using outputs of the Universal Thermal Climate Index (UTCI) from ERA5-HEAT, and leave-one-out cross-validation (LOOCV) is adopted to assess the generalization performance and responses to extreme climates. The results reveal that the C-vine copula effectively captures the conditional dependence and nonlinear structures among multi-dimensional climatic variables (AIC = −4490.36, CvM test p = 0.175). The SSRI exhibits a significant positive correlation with the fraction of days without thermal stress derived from UTCI (Kendall’s τ = 0.289, p < 0.001), and the average year-to-year classification accuracy of LOOCV reaches 66.3%. Plains are dominated by wind-driven heat dissipation, whereas high-altitude mountainous regions are sensitive to temperature–humidity variations, and statistical buffering signals induced by vegetation can be detected. The SSRI in Shaanxi presents a “three-high and two-low” spatial pattern: high-value clusters occur in the Qin-Ba Mountains, the northern slope of the Daba Mountains and the Huanglong Forest Region, while low-value areas are distributed in the Guanzhong Basin and the wind-sand zones of northern Shaanxi. Relatively high suitability within forested areas during extreme high-temperature events can be statistically identified, which is jointly modulated by altitude–terrain–vegetation interactions, with a balanced accuracy of 0.771. The complete modelling workflow of this study can provide methodological references for other complex terrain regions. Nevertheless, all Shaanxi SSRI parameters, normalisation thresholds and classification criteria are trained on Shaanxi-specific datasets and cannot be directly transferred; local observational data must be used for refitting and calibration when applying this index to other regions.