County-level disaster resilience to geological hazards in high-altitude mountainous Sichuan, China: an EDE&ndash;TA&ndash;CCD assessment

Xiao, Hai; Dong, Xiangyu; Dong, Manman; Zhang, Shaoliang; Zhao, Qianshu; Jiang, Fengjiao; Liu, Liangzheng; Yu, Mengxi; Yang, Chenglin; Zhang, Minyue

doi:10.5194/egusphere-2026-3640

Preprints

https://doi.org/10.5194/egusphere-2026-3640

Preprints

13 Jul 2026

| 13 Jul 2026

Status: this preprint is open for discussion and under review for Natural Hazards and Earth System Sciences (NHESS).

County-level disaster resilience to geological hazards in high-altitude mountainous Sichuan, China: an EDE–TA–CCD assessment

Hai Xiao, Xiangyu Dong, Manman Dong, Shaoliang Zhang, Qianshu Zhao, Fengjiao Jiang, Liangzheng Liu, Mengxi Yu, Chenglin Yang, and Minyue Zhang

Abstract. High-altitude mountainous counties face compound geological hazards and chronic socioeconomic and ecological pressures, yet existing resilience assessments rarely integrate external disturbance, internal adaptive capacity, and subsystem coordination within a dynamic county-level framework. Taking 73 medium- and high-altitude counties in Sichuan Province, China, as the study area, this study develops an EDE–TA–CCD framework that combines Environmental Disturbance Exposure, Total Adaptive Capacity, and Coupling Coordination Degree. Using panel data from 2007 to 2022, an improved coupling coordination model, spatial autocorrelation analysis, and K-means clustering were applied to examine the spatiotemporal evolution, spatial dependence, and typological differentiation of geological disaster resilience. The results show that regional resilience improved overall during the study period, mainly driven by the continuous enhancement of TA and CCD, while EDE remained relatively stable. Spatially, resilience displayed significant clustering and path dependence, with higher resilience concentrated in counties with stronger development foundations and reconstruction support, and lower resilience persisting in remote mountainous areas. Static classification and dynamic trajectory analysis further identified three resilience patterns and five evolutionary pathways, revealing differentiated processes such as low-level lock-in, stable saturation, post-disaster remodeling, rapid catch-up, and high-pressure stress climbing. The proposed framework provides a practical tool for identifying resilience disparities and supporting differentiated disaster risk governance in high-altitude mountainous regions.

Received: 22 Jun 2026 – Discussion started: 13 Jul 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Hai Xiao, Xiangyu Dong, Manman Dong, Shaoliang Zhang, Qianshu Zhao, Fengjiao Jiang, Liangzheng Liu, Mengxi Yu, Chenglin Yang, and Minyue Zhang

Status: open (until 24 Aug 2026)

Post a comment Subscribe to comment alert

RC1:
'Comment on egusphere-2026-3640', Milad Basirifard, 13 Jul 2026 reply
The manuscript presents a potentially useful county-level framework integrating Environmental Disturbance Exposure (EDE), Total Adaptive Capacity (TA), and Coupling Coordination Degree (CCD). The study area and multidimensional dataset are valuable. However, the central methodology currently contains substantial conceptual, mathematical, validation, and reproducibility problems. In particular, the modified CCD formulation and its correction factors are not sufficiently justified and appear to contain internal inconsistencies. The clustering and trajectory analyses also require stronger validation before the proposed resilience types and policy conclusions can be considered reliable.
Equations (5)–(11) are the methodological core of the manuscript, but several variables and relationships are unclear or mathematically problematic. In Equation (6), D=C⋅T, but T is not consistently defined. In some places, it appears to represent the comprehensive development level or TA, whereas elsewhere it is called the comprehensive threat index. In Equation (7), the numerator T(x,y) is described as the regional resilience level, while the denominator is the threat level. This creates confusion between TA, resilience, threat, and the symbol T. Equation (11) is unclear whether the same adaptive-capacity information is included multiple times. The authors should provide a complete derivation of the proposed model, define every variable consistently, and explain the expected range and interpretation of each term. A flowchart showing exactly which raw indicators enter TA, CCD, EDE, Threat, Exposure, and the final resilience index would be helpful.

The Exposure Normalization Factor becomes extremely large as exposure approaches zero. More importantly, Equation (9) sets exposure to zero when slope is less than or equal to 25∘, which would make Equation (8) undefined because of division by zero. The manuscript does not explain how these cases were treated computationally. Similarly, the topographic correction factor includes max(R)/R(x,y). This term increases when local relief decreases, although the text states that higher relief should amplify geological-hazard pressure. The direction of this term therefore appears inconsistent with its physical interpretation. The notation is also inconsistent: the parameter is introduced as γ, but the text later states that α is set to 0.5.

CCD is calculated directly from the five TA subsystems, and TA or a closely related comprehensive development score appears to be included again in the resilience adjustment factor and final modified CCD equation. Consequently, the reported simultaneous increases in TA and CCD may be partly mathematical rather than evidence of two distinct resilience processes. The authors should perform a systematic redundancy analysis and clearly justify all overlapping indicators. An ablation experiment is needed comparing: the conventional CCD model;the model with only the topographic correction;the model with only threat and exposure corrections; and the complete proposed model.

The manuscript repeatedly refers to county-level “disaster resilience,” but the framework is evaluated primarily through internal index behavior, spatial patterns, and cluster interpretation. No independent outcome is used to demonstrate that counties with higher scores actually experienced lower losses, shorter recovery periods, fewer casualties, less infrastructure disruption, or faster post-disaster recovery.

The analysis uses only four observations, 2007, 2012, 2017, and 2022, but the manuscript does not sufficiently describe how heterogeneous datasets were harmonized for those years. The authors should provide a detailed data dictionary containing the exact product, version, spatial resolution, temporal resolution, processing period, aggregation method, units, and missing-data treatment for every indicator.

The dynamic clustering is based on the 2007 baseline and a linear slope estimated from only four observations. This strongly compresses the temporal information and assumes that changes between 2007 and 2022 were approximately linear. Such a representation cannot reliably distinguish sudden shocks, nonlinear recovery, delayed policy effects, temporary deterioration, or post-disaster reconstruction phases.

For dynamic clustering, the highest reported silhouette coefficient occurs at K=2, not K=5. Moreover, K=8 also has a higher silhouette coefficient than K=5. Therefore, the statement that K=5 is optimal is not adequately supported. The fifth dynamic cluster contains only two counties, making the “Continuous High Pressure-Stress Climbing” category highly unstable and difficult to generalize. The static clustering also treats 292 county-year observations as independent, although four observations come from each county. This ignores repeated-measure dependence and may cause the clusters to reflect time periods rather than true county types. The clustering analysis should be strengthened by reporting: the number of random initializations and the random seed;cluster sizes for every tested K;bootstrap cluster stability;adjusted Rand index or Jaccard stability;results from alternative clustering methods;sensitivity to feature scaling; andwhether clusters are driven primarily by year or geography.

The results depend strongly on the selected spatial-weight matrix, but the manuscript does not clearly specify whether it uses queen contiguity, rook contiguity, distance thresholds, nearest neighbors, or another definition. The authors should report: the construction and row standardization of the spatial-weight matrix; treatment of counties with few or no neighbors; the number of permutation tests; the significance criterion; correction for multiple local Moran’s I tests; and sensitivity to alternative weight matrices. Reporting p-values as 0.0000 is not appropriate; they should be stated as p<0.001 or according to the minimum attainable permutation p-value

Minor comments
Figure 1 contains Chinese words and place labels. All map labels, legends, annotations, and inset text should be translated into English for an international readership. The figure should also be enlarged because several labels are currently difficult to read.

Figure 3 is overly dense, and much of its text and equations are unreadable at normal page size. It should be redesigned as a simpler workflow figure.

Reply
Citation: https://doi.org/10.5194/egusphere-2026-3640-RC1

Hai Xiao, Xiangyu Dong, Manman Dong, Shaoliang Zhang, Qianshu Zhao, Fengjiao Jiang, Liangzheng Liu, Mengxi Yu, Chenglin Yang, and Minyue Zhang

Metrics will be available soon.

Latest update: 13 Jul 2026

Short summary

Landslides hit Sichuan's mountain counties often, but their resilience trends are unclear. This study assessed 73 counties (2007–2022) via exposure, adaptive capacity, and coordination. Exposure stayed stable, while capacity and coordination rose in most counties. Resilience varies across counties, showing spatial clustering. We found three resilience types and five paths to guide governance. These findings help reduce disaster risks in fragile mountain regions.