Evaluation of the impact of land-use change on earthquake risk distribution in different periods: An empirical analysis from Sichuan Province

2.1 Overview of the study area

‌Sichuan Province (26°03′-34°19′N, 97°21′-108°33′E; Figure 1) features diverse topography ranging from northwestern highlands to southeastern lowlands, including plains, hills, mountains, and plateaus. The climate varies from subtropical humid to alpine plateau, with mean annual temperatures of 14–19°C and precipitation of 900–1,200 mm. As China’s most disaster-prone region, Sichuan experienced 19 M ≥ 5.0 earthquakes (2008–2018), causing 460,000 casualties and ¥856.8 billion in losses [33]. The 2008 Wenchuan (M8.0) and 2013 Lushan (M7.0) earthquakes were particularly devastating [34].

Figure 1

Overview of the study area.

2.2 Data

2.2.1 Land use data

We used the China Land Cover Dataset (https://doi.org/10.5281/zenodo.4417810), providing 30 m resolution annual land cover data (1985–2023) with 85% overall accuracy and Kappa >0.82 [35]. This dataset has been widely applied in disaster research, including floods, earthquakes, and landslides [36,37]. Hydrological data were obtained from the National Geospatial Information Center (https://www.webmap.cn/commres.do?method=dataDownload). Figure 2 shows land cover evolution across four periods.

Figure 2

Spatial distribution of land-use data in different periods. (a) 1990, (b) 2000, (c) 2010, and (d) 2023.

To validate the suitability of the acquired land-use data for research purposes, we assessed data quality using confusion matrices [38,39,40], comparing classifications with high-resolution Google Earth imagery (Figure 3). Validation showed >85% accuracy and Kappa >0.82 for all periods (Table 1), meeting research requirements [41].

Figure 3

Spatial distribution of samples in 2020. (a) 1990, (b) 2000, (c) 2010, and (d) 2023.

Table 1

Accuracy evaluation results of land-use data

Year	1990	2000	2010	2023
OA (%)	85.68	87.25	85.90	89.12
Kappa	0.82	0.84	0.83	0.87

2.2.2 Natural environment data

The natural environment dataset comprises four critical components: topographic elevation, slope gradient, earthquake fault zones, and historical earthquake records. The elevation data, acquired from the United States Geological Survey’s Earth Resources Observation and Science Center, features a spatial resolution of 30 m. Derived slope data were generated through digital terrain analysis of the elevation dataset. Both fault zone and seismic records were obtained from the National Earthquake Data Center (https://data.earthquake.cn/). Figure 4 displays these environmental variables.

Figure 4

Spatial distribution of natural environment data. (a) DEM, (b) slop, (c) fault zone, (d) earthquake point, and (e) water.

3.1 Land use data change analysis method

3.2 Earthquake risk assessment

This study integrates three key components: (1) earthquake hazard analysis, (2) vulnerability assessment, and (3) comprehensive risk evaluation (Figure 5).

Figure 5

Earthquake risk assessment technical flow chart.

3.2.2 Earthquake vulnerability

In accordance with the evaluation criteria of earthquake vulnerability [47], vulnerability encompasses numerous aspects: population, economy, environment, and ecology [48]. In this study, a vulnerability model defined as a function of exposure, sensitivity, and adaptive capacity [49], Formula (2), was adopted.

(2) V i = E i × S i A i ,

where V i is the vulnerability indicator for county i , E i is the exposure of county i , S i is the sensitivity of county i , and A i is the adaptability of county i .

Exposure ( E i ): It is defined as the nature and extent to which an ecosystem system is exposed to significant climate change [50]. Data on land-use classification were obtained following the disaster risk assessment guide released by China’s State Oceanic Administration (http://www.soa.gov.cn/zwgk/zcgh/ybjz/201601/t20160115_49734.html). Based on the guidelines and the classification system, the different land types correspond to one or more of the following types of hazard-affected bodies. For instance, impervious surfaces received the highest exposure value (1.0) due to their high population density and infrastructure concentration, making them particularly vulnerable to earthquake-induced casualties and economic losses [51]. Land use exposure values were assigned as follows: cropland (0.8) due to loose soil structure prone to secondary disasters [52]. Vegetation (0.6) for its stabilizing effect on geological structures [53]. Water/barren areas (0.4) for potential flood/landslide risks; and snow/wetlands (0.2) for avalanche/liquefaction hazards [54]. Environmental exposure factors included: (1) elevation, (2) slope, (3) water proximity, (4) fault zone distance, and (5) historical epicenter proximity. Specifically, the higher the elevation of an area, the steeper the slope, or the closer it is to water systems, fault zones, and earthquake points, the higher the exposure, and the lower the vice versa [55]. The AHP to determine factor weights [56]. This method mainly includes the following four aspects [57]: developed hierarchical model (goal-criterion-solution layers), conducted pairwise comparisons via expert consultation, calculated eigenvalues and eigenvectors (CR < 0.1), and derived normalized weights (Table 4). Exposure was calculated as formula 3. In this study, the classification determination of each index was referred to [58,59] (Table 4). Spatial distributions are shown in Figure 6

(3) E i = w i × l + w e × e + w s × s + w n × n + w d × d + w q × q ,

where w i represents the exposure value of land-use data, and l represents the weight of land-use data. w e represents the exposure value of the elevation data, and e represents the weight of the elevation data. w s represents the exposure value of slope data, and s represents the weight of the slope data. w n represents the exposure value of distance to water data, and n represents the weight of the distance to water data. w d represents the exposure value of the distance to fault zone data, and d represents the weight of the distance to fault zone data. w q represents the exposure on behalf of the distance to earthquake point data, and q represents the weight of the distance to earthquake point data.

Table 4

Exposure indicators and weights

Indicator	Judgment criterion	Exposure value	Weight	Indicator	Judgment criterion	Exposure value	Weight
Elevation (km)	<1.0	0.2	0.111	Distance to fault zone (km)	<1	1	0.115
	1.0–2.0	0.4			1.0–3.0	0.8
	2.0–3.0	0.6			3.0–5.0	0.6
	3.0–4.0	0.8			5.0–10.0	0.4
	>4.0	1			>10.0	0.2
Slope	<5°	0.2	0.078	Distance to earthquake point (km)	<10	1	0.131
	5°–8°	0.4			10–20	0.8
	8°–15°	0.6			20–30	0.6
	15°–25°	0.8			30–40	0.4
	>25°	1			>40	0.2
Distance to water (km)	<0.5	1	0.065	Land use	Impervious	1.0	0.5
	0.5–1.0	0.8			Cropland	0.8
	1.0–2.0	0.6			Vegetation	0.6
	2.0–5.0	0.4			Water	0.4
	>5.0	0.2			Bareland	0.4
					Snow/ice	0.2
					Wetland	0.2

Figure 6

Spatial distribution map of exposure in different periods. (a) 1990, (b) 2000, (c) 2010, and (d) 2023.

Sensitivity ( S i ): It represents a system’s susceptibility to disasters [60], while adaptability ( A i ) measures its capacity to respond to and recover from adverse impacts [61]. Following Liu et al. [62], we incorporated key economic indicators (public budget expenditure and GDP) to reflect regional disaster resilience. Our evaluation system comprised 14 standardized indicators assessing county-level seismic vulnerability (Table 5). Data normalization ensured comparability across indicators [63]:

(4) x i j ′ = x i j − MIN ( x j ) MAX ( x j ) − MIN ( x j ) ,

(5) x i j ′ = MAX ( x j ) − x i j MAX ( x j ) − MIN ( x j ) ,

Table 5

Indicators for assessing vulnerability index to earthquake disaster

Vulnerability dimension	No.	Indicator	Vulnerability impact1	Weight
Exposure	1	Areas of land use with an exposure value of 0.2–0.4	+	0.089
	2	Areas of land use with an exposure value of 0.4–0.6	+	0.074
	3	Areas of land use with an exposure value of 0.6–0.8	+	0.177
	4	Areas of land use with an exposure value of 0.8–1.0	+	0.224
	5	Total population	+	0.051
	6	Percentage of females	+	0.035
Sensitivity	7	Percentage of population aged 14 and under	+	0.093
	8	Percentage of population aged 65 and above	+	0.112
Adaptability	9	General public budget expenditure	—	0.002
	10	GDP	—	0.105
	11	Urban disposable income per capita	—	0.016
	12	Rural disposable income per capita	—	0.005
	13	Number of hospital medical staff	—	0.008
	14	Number of medical institutions	—	0.009

¹ “+” indicates the indicator tends to increase vulnerability; “−” indicates the indicator tends to decrease vulnerability.

where formula (4) represents a positive indicator, formula (5) represents a negative indicator, x i j is the raw data value, x i j ' is the standardized value of x i j , MAX ( x j ) is the maximum value of the jth indicator, and MIN ( x j ) is the minimum value of the j th indicator.

To improve assessment robustness, we employed an entropy-weighted Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) method for vulnerability evaluation. This approach calculates distances to both ideal and negative-ideal solutions, providing comprehensive and objective results by (1) minimizing subjective bias in weight determination through entropy weighting, and (2) systematically comparing alternatives against optimal benchmarks. The calculation formula can be referred to in the literature [64].

The calculated vulnerability index values are graded, as shown in Table 6.

Table 6

Comparison of vulnerability index values and vulnerability levels

Vulnerability value	Vulnerability level	Vulnerability explanation
(0.0, 0.079]	1	Very low
(0.079, 0.16]	2	Low
(0.16, 0.32]	3	Moderate
(0.32, 0.54]	4	High
(0.54, 1.0]	5	Very high

3.3 Spatial autocorrelation analysis of earthquake risk

Spatial autocorrelation analysis evaluates the spatial dependence and clustering patterns of geographical variables, comprising both global and local components [66]. Global spatial autocorrelation assesses overall spatial association patterns across the study area.

where n represents the total number of counties, and X i and X j represent the risk index values of cities i and j , respectively. X ¯ represents the average of the risk index of all cities and W i j is the spatial weight matrix, representing the spatial relationship between cities i and j . When cities i and j are adjacent in space, W i j = 1; otherwise, W i j = 0.

Global spatial autocorrelation analysis cannot identify localized clusters or spatial anomalies. Therefore, we employed the local Moran’s I index to assess spatial risk correlations and detect significant clusters. The calculation is as follows:

where n , X ¯ , X i , X j , and W i j are the same as in formula (3). Based on the local Moran’s I values, the spatial risk patterns are classified into five types: “high-high” (HH) and “low-low” (LL) represent spatial clusters with positive spatial autocorrelation; “high-low” (HL) and “low-high” (LH) can be described outliers with negative spatial autocorrelation; “Not significant” means that there is no significant spatial difference in the risk index between a city and its neighbors.

4.1 Analysis of land-use change from 1990 to 2023

Table 8 presents the land-use area changes in Sichuan Province from 1990 to 2023, revealing several key patterns: cropland area decreased by 9,616 km² (24.74–22.76%) due to urbanization and ecological policies, while forest cover expanded by 17,702 km² (37.10–40.74%) reflecting conservation efforts, grassland diminished by 10,534 km² (35.03–32.87%). And the area of impervious surfaces increased significantly, from 1246.82 km² in 1990 to 4805.36 km² in 2023 (0.26–0.99%), reflecting accelerated urbanization and increased demand for construction land. In summary, the area of cropland, grassland, and wetland in Sichuan Province showed a decreasing trend. The type area of forest and impervious surfaces showed an increasing trend. Meanwhile, complex fluctuation patterns for shrubland (decrease–increase–increase), water bodies (increase–increase–decrease), snow/ice (stable–increase–decrease), and barren land (increase–decrease–increase). These changes collectively illustrate the dynamic interplay between anthropogenic activities and natural systems in shaping regional land cover dynamics.

Table 9 reveals land-use transitions in Sichuan Province (1990–2023), with the unchanged area of land-use type is 428277.24 km², of which the relatively large cropland and forest unchanged areas are 98332.60 km² and 168,748 km², respectively. The most dynamic transitions occurred among cropland, forest, grassland, shrubland, and barren land, particularly involving cropland-to-forest (16,341 km²) and cropland-to-impervious (3654.91 km²) conversions, reflecting simultaneous ecological restoration and urban expansion. Forest gains are derived not only from cropland but also grassland (636.035 km²) and shrubland (1744.23 km²) conversions, while grassland losses primarily transitioned to forest and cropland despite 154,707 km² remaining stable. Notably, impervious surface expansion stemmed chiefly from cropland and grassland conversions, underscoring urbanization’s profound transformation of land cover patterns during this 33-year period.

Table 10 demonstrates temporal changes in land-use landscape indices for Sichuan Province (1990–2023), revealing a clear trend toward landscape consolidation and simplification: PD declined from 11.76 to 6.68, indicating reduced landscape fragmentation and increased land-use centralization; ED decreased from 52.24 to 41.59, reflecting a decrease in the length of the landscape edge, which further indicated that the fragmentation degree of the landscape was reduced. LSI dropped from 914.50 to 728.76, demonstrating geometric simplification and regularization of patch forms. While Aggregation Index (AI) increased from 92.16 to 93.76, signifying enhanced spatial connectivity and clustering of land-use types. These systematic shifts toward lower fragmentation, simplified geometries, and higher aggregation likely result from the combined effects of urbanization, ecological conservation policies, and land-use planning adjustments. These changes have important implications for regional ecosystem services and disaster risk assessment.

4.2 Earthquake risk analysis

Figure 7 presents the earthquake hazard distribution derived from ground shaking parameters (PGA), which visually expresses the degree of spatial variability of earthquake hazards in the study area. Low-hazard zones (PGA < 0.09 g) predominantly occur in the geologically stable Sichuan Basin, suggesting minimal seismic risk, moderate-hazard areas (0.09–0.19 g) form transitional belts along basin margins and hilly regions. High-hazard zones (0.19–0.38 g) characterize the tectonically active western Sichuan Plateau and mountainous areas, while very high-hazard regions (PGA >0.38 g) concentrate near major fault systems in the western Sichuan Plateau, particularly around Xichang City. This east-to-west escalating hazard gradient fundamentally reflects the province’s geological setting – the stable underlying the eastern basin contrasts sharply with the intensely deforming eastern Tibetan Plateau margin in the west, explaining both the spatial hazard distribution and varying seismic activity levels observed across different physiographic regions.

Figure 7

Spatial distribution of earthquake hazard levels.

Integrated assessment of earthquake vulnerability in Sichuan Province, incorporating land-use patterns, socio-environmental indicators, and natural environmental parameters, demonstrates a general downward trend in vulnerability values across most counties during 1990–2023, though with notable spatial and temporal variations. While the majority of counties exhibited decreasing vulnerability, certain localities experienced periodic fluctuations or even increasing vulnerability levels. Quantitative analysis reveals distinct temporal patterns (Figure 8): very high vulnerability areas decreased progressively (5 → 4 → 4 → 2 counties across four periods), while high vulnerability zones showed an initial decline followed by resurgence (5 → 3 → 2 → 3 counties). Moderate vulnerability increased during 1990–2000 (6 → 7 counties) before stabilizing. Spatially, very low vulnerability regions predominated in eastern Sichuan, contrasting with fewer occurrences in central, southern, and western areas – a distribution strongly correlated with lower building density, reduced population concentrations, and limited government fiscal capacity in these less vulnerable zones. These patterns collectively highlight the complex interplay between anthropogenic factors and geographical determinants in shaping regional seismic vulnerability dynamics.

Figure 8

Spatial distribution of earthquake vulnerability levels: (a) 1990, (b) 2000, (c) 2010, (d) 2023.

Based on a comprehensive assessment of earthquake hazard and earthquake vulnerability, reveals pronounced spatiotemporal variations in risk distribution (Figure 9), demonstrating a clear east-west gradient with escalating risk levels from the stable basin regions to the tectonically active western plateau and mountainous areas. Initial 1990 data showed basin regions dominated by “very low” to “low” risk classifications, contrasting sharply with western areas characterized by “moderate” to “very high” risk levels. Temporal analysis indicates significant spatial reorganization of risk zones: very high-risk areas migrated from central-southern westward regions with decreasing coverage, while high-risk zones shifted from southern to western areas with similar areal reduction. These spatial transformations correlate with observed land-use changes, particularly forest expansion (increasing by 17,702 km² during 1990–2023) and landscape pattern modifications (PD decreasing from 11.76 to 6.68, AI increasing from 92.16 to 93.76), suggesting that enhanced vegetation cover and reduced landscape fragmentation have contributed to regional risk mitigation. The risk redistribution patterns align closely with provincial land-use policy implementations, particularly the Grain-for-Green program.

Figure 9

Spatial distribution of earthquake risk levels: (a) 1990, (b) 2000, (c) 2010, and (d) 2023.

Quantitative analysis of earthquake risk distribution in Sichuan Province (1990–2023) reveals distinct temporal patterns (Table 11): very low-risk areas exhibited gradual expansion (143–145 counties), while low-risk zones showed more pronounced growth (14–18 counties). Conversely, both moderate and very high-risk areas demonstrated consistent decline throughout the 33-year period, with high-risk areas following a unique trajectory of initial reduction followed by resurgence. The provincial risk profile remains dominated by very low-risk classifications, accounting for the majority of spatial coverage, with other risk categories collectively representing secondary components of the overall risk distribution. This pattern reflects the combined effects of landscape stabilization and targeted risk mitigation measures.

4.3 Land-use composition dynamics along earthquake risk gradients

This study examines the spatiotemporal dynamics of earthquake risk distribution across various land-use types from 1990 to 2023, revealing shifts in landscape composition under different seismic risk levels (Table 12). Key findings indicate that forest cover expanded notably in very low (36.29–40.67%) and very high-risk zones (39.22–67.69%), suggesting potential linkages to reforestation policies or climate-driven vegetation changes. Conversely, cropland declined in very low risk areas (42.91–38.33%), while impervious surfaces increased (0.44–1.73%), reflecting urbanization pressures. Grasslands dominated moderate to high-risk regions but exhibited volatility, peaking in 2000 (68.23%) before declining to 55.57% (moderate) and 65.13% (high) by 2023, possibly due to land degradation or agricultural conversion. Notably, wetlands nearly vanished in very low and moderate risk categories, underscoring ecological vulnerability. The abrupt rise of forests in very high-risk zones, alongside shrinking grasslands, may signal ecosystem resilience or anthropogenic interventions. Quantitatively, each unit increase in risk level elevates the probability of expansion for human-dominated land types (e.g., cropland and impervious) while reducing the likelihood of expansion for natural ecological covers (e.g., grassland and wetland). The anomalous growth of grasslands in moderate-risk zones (+10.13%) may reflect localized buffering effects from ecological restoration efforts. These trends highlight the interplay between seismic hazards and land-use change, emphasizing the need for risk-sensitive territorial planning to mitigate future vulnerabilities.

4.4 Global and local spatial autocorrelation analysis

Based on the 2023 earthquake risk results, the spatial autocorrelation statistical technique is used to determine whether the risk has statistical significance in the spatial clustering, as shown in Figures 10 and 11. First, the global Moran’ I index is calculated as 0.58 (z-value = 12.99, p-value <0.001), indicating that there are significant positive spatial correlations and risk spatial clustering among the 183 cities analyzed. Figure 10 shows that most of the points are located in the first and third quadrants, representing clusters of HH and LL cities, respectively. In addition, a small number of points are distributed in the second quadrant, indicating that low-risk cities are surrounded by high-risk cities; that is, these cities show a negative spatial correlation in terms of risk.

Figure 10

Global Moran index scatter plot.

Figure 11

Spatial clustering map of earthquake risk in the counties of Sichuan Province.

Local spatial autocorrelation analysis (Figure 11) provides an intuitive representation of earthquake risk clusters in the “HH,” “LL,” “HL,” and “LH” cluster types. First, the western region cities are identified as HH risk clusters, corresponding to the higher risk of these cities. Second, the eastern region has been identified as an LL risk cluster, and these cities are resilient and not vulnerable to earthquake disasters. In addition, there are LH clusters centered on Xiangcheng, Daocheng, Hongyuan, and Miyi. In other words, these cities are surrounded by cities with high risk. Compared with the surrounding cities, the risk and vulnerability of these cities are below the middle and low levels, which makes these cities have strong adaptability. Finally, the other cities showed no obvious agglomeration, and the spatial autocorrelation was not significant, indicating that the risk was randomly distributed. Given that this result is a well-documented risk distribution among prefecture-level cities in Sichuan Province, it is necessary to give priority to these high-risk cities in risk reduction planning and management.

5.1 Verification of earthquake risk assessment results

To validate our earthquake risk assessment, we compared results with historical seismic data from the China Earthquake Networks Center and the National Earthquake Data Center. As shown in Figure 12, the earthquake zoning assessed in this study aligns well with the distribution of historical earthquake epicenters. While M >6 earthquakes are rare, western Sichuan exhibits significant latent risk due to high altitude, sparse population, and poor socioeconomic conditions [67]. Our findings show strong agreement (80% overlap) with Shao et al.’s [68] predicted strong earthquake zones for 2021–2030 (Figure 12). Shao et al. predicted four strong earthquake risk zones in Sichuan Province, namely the Eastern section of the Kunlun fault belt – northern section of Longriba fault zone, the middle and south segment of Xianshuihe fault belt – the south segment of Longmenshan fault zone, Litang fault zone Shawan segment–Lijiang–Xiaojinhe fault zone, and East of Sichuan–Yunnan border. A comparison revealed that these strong earthquake hazard zones largely coincide with the moderate or higher earthquake risk zones assessed in this study. Although Shao et al.’s study primarily relied on fault detection and geophysical observations, the results of this study suggest that the northwestern region of Sichuan Province also faces a high earthquake risk. Therefore, it is recommended that future planning for earthquake hazard mitigation prioritize the northwestern region of Sichuan Province as a key area of focus.

Figure 12

Historical earthquake records and prediction maps of strong earthquake danger areas in Sichuan Province.

Earthquake prediction remains inherently uncertain due to the stochastic nature of seismic events [69]. Our risk assessment model, while incorporating key seismic parameters, contains simplifications that may limit its application across diverse geological settings, particularly regarding wave attenuation variations [70]. Future refinements will focus on integrating additional empirical data and optimizing attenuation and fault activity parameters to improve model robustness.

5.2 Impact analysis of land-use change on earthquake risk assessment

From 1990 to 2023, earthquake risk levels in Sichuan Province are closely related to the change of land-use type (Figure 13). Among them, the number of very low-risk areas and low-risk areas increased from 143 and 14 in 1990 to 145 and 18 in 2023, respectively. Significant changes have taken place in the land-use structure, mainly represented by the decline of the proportion of cropland area and the increase of forest area. The increase in forests may reduce earthquake risk by enhancing ecological stability and disaster resilience [71]. For example, forests can effectively reduce the occurrence of lifetime disasters such as landslides and mudslides, thus reducing the vulnerability to earthquake disasters. Grassland expansion (45.44–55.57%, 1990–2023) correlated with reduced high-risk areas (16 to 6), suggesting improved soil stability mitigated secondary seismic hazards. Conversely, cropland expansion in high-risk zones (4.14–15.44%) increased vulnerability through: (1) loose soil structure prone to liquefaction and sliding during earthquakes, and (2) denser infrastructure amplifying potential losses [72].

Figure 13

Changing trend of area proportion of different land-use types and number of earthquake risk grades from 1990 to 2023.

The land-use intensity index in Sichuan Province exhibited a fluctuating trend during 1990–2023, peaking in 2000 (224.51) before declining to 223.63 by 2023 (Table 13). This decrease reflects reduced land-use intensity, likely associated with enhanced ecological protection measures and land-use optimization [73]. For example, the increase of forest and grassland, the reduction of cropland, and the control of construction land have reduced the intensity of land use, resulting in a decrease in the number of very high-risk areas in Sichuan Province.

A reduction in the number of patches, decreased edge complexity, more regular shapes, and increased aggregation indicate the optimization and stabilization of land-use landscape patterns, which contribute to lowering regional earthquake risk levels [74]. Future research should further investigate the underlying mechanisms through high-resolution land-use mapping coupled with earthquake case analyses to assess impacts on surface structures, population distribution, and infrastructure density, as well as enhanced spatiotemporal modeling incorporating real-time monitoring and uncertainty quantification to improve assessment accuracy.

5.3 Analysis of the importance of different environmental factors on earthquake risk

Land use significantly influences earthquake risk due to varying vulnerability across different types [11]. Urban areas with dense infrastructure exhibit higher vulnerability, while natural landscapes like forests and grasslands demonstrate greater resilience. Our analysis prioritized land use (weight = 0.5) given its direct impact on exposure and susceptibility. Environmental factors including proximity to epicenters, fault zones, steep slopes, and water bodies (potential liquefaction) [75], further modulate risk, though certain land uses (e.g., vegetation) may mitigate secondary hazards through soil stabilization [76]. These observations align with our AHP-derived factor weights (Table 14): land use > epicenter distance > fault zone distance > elevation > slope > water proximity.

Sichuan Province’s earthquake risk with high-risk zones concentrated in western and central regions characterized by complex topography and active tectonics. The western region, located on the Tibetan Plateau’s eastern margin, shows particularly elevated risk due to: (1) ongoing crustal deformation from Bayan Har and Sichuan–Yunnan Block interactions [77], and (2) concentration of major fault zones (Longmenshan, Xianshuihe, Anninghe-Zemuhe) that have hosted 81.8% of the province’s M ≥ 7 earthquakes historically (Figure 14). These faults form a “Y”-shaped seismic framework where crustal thickness gradients facilitate frequent moderate-to-strong earthquakes.

Figure 14

Distribution map of major fault zones in Sichuan Province.

Socioeconomic factors exacerbate vulnerability in western regions, where (1) limited healthcare/education infrastructure, (2) challenging plateau terrain hindering rescue operations, and (3) frequent secondary hazards (landslides, debris flows) compound disaster impacts. Recent M∼6 events in densely populated eastern areas (e.g., Leshan, Yibin) highlight emerging risks along lesser-studied faults, necessitating enhanced preparedness measures.

5.4 Policies and Suggestions

Landscape type-earthquake risk relationships are critical for guiding land-use planning and risk mitigation. Rapid urbanization often exacerbates landscape fragmentation and seismic vulnerability, necessitating optimized land-use strategies with ecological buffers. For instance, Chengdu’s “ecological redlines” and green infrastructure demonstrate effective urban boundary control, reducing geological disturbances while enhancing earthquake resilience [78]. These initiatives not only enhanced the city’s ecological resilience but also reduced the risk of secondary earthquake disasters, demonstrating significant socio-economic feasibility.

In agricultural regions, excessive reclamation increases soil erosion and geological risks. Liangshan Prefecture’s “Grain for Green” and terracing projects significantly decreased landslides while improving land productivity [79], showcasing balanced ecological–economic benefits. For natural ecosystems like Aba Prefecture, forest protection and grassland restoration policies enhanced vegetation coverage and surface stability, mitigating earthquake-induced landslides [80]. Ganzi Prefecture’s grassland restoration and ecological migration improved landscape connectivity while reducing soil erosion. These measures not only enhance the disaster resilience of regional ecosystems but also reduce the risks of landslides and collapses triggered by earthquakes, providing critical support for sustainable regional development.

Our interdisciplinary study combines long-term time-series analysis with entropy-weighted TOPSIS methods to objectively assess Sichuan’s land-use-earthquake risk relationships. Furthermore, through the analysis of typical cases and policy evaluations, practical and actionable recommendations are proposed. The findings not only enrich the theoretical framework of earthquake risk prevention and control but also provide a scientific basis and practical guidance for land-use planning and disaster mitigation in Sichuan Province and similar regions.