Next Article in Journal
PM2.5 Speciation of Beta Attenuation Monitor Filters During Wildfire Smoke Events
Previous Article in Journal
Deep Learning in Airborne Particulate Matter Sensing and Surface Plasmon Resonance for Environmental Monitoring
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Atmosphere
Volume 16
Issue 4
10.3390/atmos16040360
atmosphere-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Incorporating Dynamic Factors in Geological Hazard Risk Assessment: Integrating InSAR Deformation and Rainfall Conditions

by
Hui Wang
1,2,
Jieyong Zhu
1,2,*,
Likun Chen
1,2 and
Haohan Shi
1,2
1
Faculty of Land and Resources Engineering, Kunming University of Science and Technology, Kunming 650093, China
2
Key Laboratory of Geohazard Forecast and Geoecological Restoration in Plateau Mountainous Area, Ministry of Natural Resources of China and Yunnan Province, Kunming 650093, China
*
Author to whom correspondence should be addressed.
Atmosphere 2025, 16(4), 360; https://doi.org/10.3390/atmos16040360
Submission received: 14 February 2025 / Revised: 14 March 2025 / Accepted: 20 March 2025 / Published: 22 March 2025
(This article belongs to the Section Biosphere/Hydrosphere/Land–Atmosphere Interactions)
Browse Figures
Versions Notes

Abstract

:
Geological hazards, particularly in mountainous regions, represent significant threats to life, property, and the environment. In this study, we focus on Luoping County, Yunnan Province, China, employing SBAS-InSAR technology to monitor surface deformation between 8 October 2022 and 27 September 2024. By integrating InSAR deformation data with 10 static disaster-causing factors, including elevation, slope, aspect, curvature, distance to faults, distance to rivers, distance to roads, engineering geological rock groups, geomorphological types, and the NDVI, geological hazard susceptibility was assessed using the information value (IV) model and the information value–random forest (IV-RF) coupled model. Accuracy validation using ROC curves indicated that the IV-RF model, integrated with InSAR deformation data, achieved the highest accuracy, with an AUC value of 0.805. Based on the susceptibility evaluation, rainfall intensity was introduced as a triggering factor to assess geological hazard risks under four rainfall conditions: 10-year, 20-year, 50-year, and 100-year return periods. The results demonstrated that incorporating InSAR deformation data significantly improved disaster prediction accuracy, providing more reliable and sustainable risk assessment outcomes. This study underscores the critical role of InSAR technology, combined with rainfall conditions, in enhancing the precision of geological hazard risk assessments, offering a scientific basis for disaster prevention and mitigation strategies in Luoping County and similar regions.

1. Introduction

In China, the frequent occurrence of geological hazards poses a significant threat to people’s lives and property, particularly in mountainous and canyon areas [1,2,3]. Due to steep terrain and unstable geological structures, these regions are highly prone to geological disasters such as landslides, debris flows, and collapses. In recent years, global climate change has intensified, leading to an increase in extreme weather events, especially dramatic changes in rainfall patterns [4,5,6]. This has amplified the triggering effect of rainfall on geological hazards. Rainfall, as a significant triggering factor, plays a vital role in evaluating the risks associated with geological hazards [7]. Previous studies have shown that extreme rainfall significantly increases the likelihood of geological hazards, particularly in areas with steep terrain and unstable geological structures, and there is a significant correlation between rainfall intensity and the occurrence of geological hazards [8,9].
Geological hazard risk assessment refers to the evaluation of the likelihood of hazards occurring, emphasizing the probability of a disaster occurring within a given period of time [10]. The risk assessment builds upon susceptibility evaluation by incorporating temporal probabilities and triggering factors. Currently, susceptibility assessment models primarily include empirical models, statistical models, machine learning models, and coupled models that integrate multiple approaches. Empirical models based on expert scoring and the analytic hierarchy process rely on field experience and the theoretical knowledge of researchers, making them prone to subjective bias. On the other hand, single statistical models, such as the information value method, weight of evidence method, and certainty factor method, are straightforward and intuitive but overly dependent on traditional mathematical formulas, limiting their ability to fully capture the complexity of geological hazards [11,12,13]. To overcome these limitations, many studies have started combining machine learning with statistical models to leverage the advantages of both, thereby enhancing the scientific rigor and accuracy of susceptibility assessments [14,15,16,17,18].
Interferometric Synthetic Aperture Radar (InSAR) technology utilizes radar beams to measure surface deformation, providing high-precision surface displacement information for monitoring dynamic changes in geological hazard areas [19,20,21]. It offers advantages such as high accuracy, all-weather capability, temporal monitoring, low false alarm rates, and cost-effectiveness [22,23]. By capturing subtle surface deformations, InSAR technology reveals potential geological hazard risks, enabling researchers to accurately assess geological hazards on finer spatial scales [24]. Thus, integrating InSAR technology as a dynamic factor into geological hazard assessments allows for a more reasonable evaluation of regional geological hazards. Many scholars have explored the application of InSAR technology in geological hazard research. Stavroula Alatza et al. employed InSAR to monitor ground deformation on Amorgos Island, Greece, revealing small-scale ground movements following the 1956 earthquake, and recommended integrating GPS data to enhance the accuracy of earthquake risk assessments [25]. Gopal Sharma et al. combined InSAR with geological and geophysical data to assess earthquake damage susceptibility in the Barapani Shear Zone in India, revealing higher deformation rates in the northern region, indicating a greater seismic risk [26]. Karan Nayak et al. studied the relationship between ionospheric TEC anomalies and the 2023 Morocco earthquake, discovering that TEC anomalies appeared several days before the earthquake, providing new insights for earthquake early warning systems [27]. Andrea Ciampalini et al. demonstrated that using PS-InSAR data reduces false positives and false negatives in landslide susceptibility maps, improving the reliability of slow-moving landslide predictions, particularly in urbanized areas, thereby providing more precise references for land-use management and decision making [28]. Furthermore, Pengfei Li et al. proposed a high-resolution landslide susceptibility assessment method by integrating land-use changes with InSAR deformation data, significantly improving prediction accuracy [29]. These studies highlight that incorporating InSAR deformation data into geological hazard susceptibility assessments has become a research hotspot. However, studies that integrate InSAR deformation as a dynamic factor alongside rainfall information for geological hazard risk assessment remain relatively rare.
Against this background, this study selected Luoping County, Yunnan Province, as the research area. A total of 54 Sentinel-1A satellite SAR images covering the entire region of Luoping County were obtained, and SBAS-InSAR technology was used to derive surface deformation information from 8 October 2022 to 27 September 2024. Based on the InSAR deformation data, 10 static disaster-causing factors were integrated, including elevation, slope, aspect, curvature, distance to faults, distance to rivers, distance to roads, engineering geological rock groups, geomorphological types, and the NDVI. The information value (IV) model and the information value–random forest (IV-RF) coupled model were employed for the susceptibility assessment of geological hazards. Building on the susceptibility assessment results, rainfall intensity was used as a triggering factor to obtain hazard assessment results under four different rainfall conditions: 10-year, 20-year, 50-year, and 100-year return periods. Additionally, the risk assessment results with and without deformation data were analyzed. The hazard assessment results with and without deformation variables were analyzed, providing crucial support for disaster prevention and mitigation in Luoping County. This effectively reduced the economic losses and social impacts caused by geological hazards.

2. Materials and Methods

2.1. Study Area

Luoping County lies in the eastern region of Yunnan Province, situated at the intersection of Yunnan, Guizhou, and Guangxi provinces. It falls under the administrative jurisdiction of Qujing City, with geographical coordinates of 103°57′–104°43′ E and 24°31′–25°25′ N, as shown in Figure 1. The county extends 75 km wide from east to west and 99 km long from north to south, covering an area of 3018 km2. It borders Xingyi City, Guizhou Province, to the east along the Huangni River, to the southeast along the Nampan and Qingshui Rivers, across the river from Xilin County, Guangxi Zhuang Autonomous Region to the southwest, and Shizong County to the north, and, to the west to the north, it connects with Luliang County, Qilin District, and Fuyuan County, respectively.
The study area has a subtropical plateau monsoon climate, characterized by mild seasons and ample rainfall. The average annual temperature is 15.1 °C, with an average yearly precipitation of 1700 mm. The region has a well-developed surface water system, which belongs to the primary and secondary tributaries of the Nanpan River system within the Pearl River Basin. Major rivers include the Nanpan River, Huangni River, Kuaize River, Jiulong River, Zhuanchang River, and Duoyi River. The total annual runoff of the water system is abundant, reaching 2.418 billion cubic meters. The well-developed hydrological system not only supports industrial and agricultural production but also increases the risk of geological hazards such as landslides and collapses, particularly during periods of heavy rainfall. The combination of intense precipitation and the region’s topography, especially during the monsoon season, contributes to slope instability, which can trigger geological hazards. The study area is primarily dominated by mountainous and basin landforms, with significant topographic relief. The terrain is generally higher in the northwest and lower in the southeast, characterized by plateau mountains and deeply cut river valleys. The high valley density and steep gorges create favorable conditions for geological hazards [3,30,31]. The exposed strata in this area mainly consist of Devonian, Carboniferous, Permian, Triassic, as well as minor Paleogene and Quaternary formations. Magmatic activity is active, with the primary exposure of mafic volcanic rocks concentrated in the northwest. These magmatic activities were primarily concentrated in the Middle to Late Permian, exhibiting certain characteristics of stratigraphic contact. The structural development of this area is mainly attributed to early Himalayan tectonic activity, which can be divided into two primary evolutionary stages. In the early stage, regional NW-SE compression dominated, forming compressional folds and thrust faults as the main structural features. During the late stage, under the background of rapid regional uplift, normal faulting occurred, forming normal fault structures. Overall, the structural features of the early stage predominate, while the distribution of late-stage structural features is relatively limited.
According to field investigations and remote sensing interpretation, a total of 167 geological hazard sites were identified within the study area. The main types of geological hazards include collapses, landslides, and ground subsidence. Geological hazards in the study area are closely related to rainfall, with most occurrences concentrated during the rainy season (May to October). Moreover, the higher the rainfall and the more frequent heavy rainfall events, the greater the susceptibility to geological hazards.

2.2. Data Source and Methodology

The data sources used in this study are listed in Table 1.
The technical flowchart of this study, illustrated in Figure 2, applies various methods to systematically evaluate geological hazard susceptibility and risk in the study area. Initially, ten static factors were chosen for evaluation, and InSAR deformation data were included as a key dynamic factor to better characterize geological hazard sensitivity. On this basis, an IV model and an IV-RF coupled model were developed to generate susceptibility maps of geological hazards. A comprehensive analysis of the evaluation results obtained from different models was conducted. Subsequently, based on the Pearson Type III distribution curve, rainfall conditions corresponding to return periods of 10, 20, 50, and 100 years were predicted. The instability probabilities under various rainfall conditions were superimposed onto the susceptibility maps, ultimately producing hazard zoning maps for geological hazards. By comparing zoning maps across different rainfall conditions, the importance of InSAR deformation data in geological hazard assessment was demonstrated.

2.3. InSAR Data Processing

This study utilized the Sentinel-1A satellite, an Earth observation satellite launched by the European Space Agency in 2014 [32,33]. The satellite is capable of all-weather, all-day Earth observation, equipped with a C-band radar that supports various polarization modes and imaging modes, with a resolution ranging from 5 to 40 m and a revisit period of 12 days for a single satellite [32,34]. A total of 54 SAR images covering the entire jurisdiction of Luoping County were acquired for the study period from 8 October 2022 to 27 September 2024.
Using the newly integrated SARscape module in ENVI 5.6 software by Esri for data processing, the Small Baseline Subset (SBAS) technique was employed to process 54 scenes of Sentinel-1A ascending orbit data along with the corresponding precise orbital data [35,36,37,38]. The results show that the annual average deformation rate of Luoping County ranges from −60.8 to 84.6 mm/a. The processing workflow is illustrated in Figure 3. Positive values indicate surface deformation moving toward the satellite along the radar line of sight, while negative values indicate deformation moving away from the satellite, typically reflecting surface uplift and subsidence [32,39]. To address coherence loss in localized areas that resulted in voids in the deformation rate inversion results, kriging interpolation was applied to the deformation rate data extracted from time-series InSAR to ensure the completeness and accuracy of the deformation rate as a dynamic factor for geological hazards. The resulting deformation magnitude level is shown in Figure 4. By constructing a continuous deformation rate map covering the entire study area, this study not only enhances the timeliness and spatial accuracy of geological hazard risk assessment but also provides a more comprehensive foundational dataset for dynamic monitoring and risk evaluation [40,41,42].

2.4. Evaluation Factors

To conduct the geological hazard risk assessment, this study fully considered the complexity of the geological environment and the main factors influencing geological hazard occurrence. Eleven disaster-causing factors were selected, including ten static factors—elevation, slope, aspect, curvature, distance to faults, distance to rivers, distance to roads, engineering geological rock groups, geomorphic types, and the NDVI—and one dynamic factor, the deformation magnitude. By integrating the static and dynamic factors, a comprehensive susceptibility evaluation was performed. Each evaluation factor was quantified and classified into grades, and the specific results are shown in Figure 5.
Based on the geological hazard susceptibility assessment conducted in the area, rainfall intensity was identified as the triggering factor. Rainfall data from 2001 to 2022 were analyzed, and the Pearson Type III distribution was applied to forecast rainfall patterns. Table 2 presents the rainfall data for each township under four scenarios: 10-year, 20-year, 50-year, and 100-year return periods.

2.5. Evaluation Methods

2.5.1. Information Value (IV) Model

The information value model is a geological hazard risk assessment method based on information theory [43,44]. By analyzing the statistical relationship between hazard occurrence and environmental factors, the model calculates the information value of each factor to quantify its contribution to hazard susceptibility [45]. A larger information value indicates that the factor has a stronger explanatory power for hazard susceptibility. Supported by GIS technology, the information value model efficiently integrates spatial data and enables the division of geological hazard risk zones within a region. The formula is as follows:
I = ∑ i = 1 n I ( x i , H ) = ∑ i = 1 n ln N i / N S i / S
In this formula, I represents the overall information value of the evaluation unit, n denotes the number of evaluation indicators, xi indicates the classification of the selected indicator within the evaluation unit, and I(xi, H) indicates the information value of the evaluation indicator xi for geological hazard H. Ni is the number of geological hazard points within the classified interval of evaluation indicator xi in the study area, while N represents the total number of geological hazard points in the study area. Si is the area corresponding to evaluation indicator xi in the study area, and S refers to the total area of the study region.

2.5.2. Random Forest (RF) Model

The random forest is an ensemble learning algorithm based on decision trees, and its core idea is the integration of multiple decision trees to enhance the model’s predictive performance [46,47,48]. Widely used in geological hazard evaluation, the random forest model excels in handling nonlinear problems in complex geological environments, identifying key influencing factors, and determining the weights of these factors [49,50]. The weights in the random forest were calculated based on the Gini index, which measures the importance of each feature by evaluating its contribution to the reduction in the Gini index during node splitting in decision trees [51]. Specifically, the weight is the cumulative sum of the Gini index reductions caused by the feature at splitting nodes across all decision trees, normalized into a proportion. A higher weight indicates greater importance of the feature in the model’s predictions. The weight calculation formula is as follows:
W f = ∑ i = 1 r ∑ j = 1 s Δ G i n i ( i , j , f ) ∑ f = 1 t ∑ i = 1 r ∑ j = 1 s Δ G i n i ( i , j , f )
In this formula, Wf represents the weight of the f-th geological hazard evaluation indicator, while t, r, and s indicate the total number of geological hazard evaluation indicators, the number of decision trees, and the number of nodes in each decision tree, respectively. ΔGini(i, j, f) denotes the reduction in the Gini index caused by the f-th evaluation indicator at the j-th node of the i-th tree.

2.5.3. Information Value–Random Forest (IV-RF) Coupling Model

The information value model quantifies the contribution of influencing factors to disaster occurrence as information values by analyzing the statistical relationship between disasters and influencing factors [52]. The random forest model further calculates the weights of these factors and optimizes the subjectivity of the evaluation process, ultimately obtaining their combined weights through a linear weighted summation method [53]. The coupling model integrates all factors using the assigned weights, generating a more scientifically sound and reasonable spatial distribution map of hazard risk [54]. The predictive accuracy and reliability of this model in disaster risk assessment have been validated in several studies. For example, Tao Chen et al. applied the IV-RF model to landslide susceptibility mapping in the Three Gorges Reservoir area, and the validation results showed that the model achieved high predictive accuracy [55]. Rongwei Li et al. used the IV-RF model for geological disaster susceptibility assessment in Kang County, Gansu Province, and found that, compared to traditional methods, the IV-RF model demonstrated higher accuracy, thereby confirming its reliability in different geological environments [48]. The equation is as follows:
S = ∑ i = 1 n W i x i
In this formula, S represents the comprehensive information value, n is the number of geological hazard evaluation indicators, Wi is the random forest weight of the i-th geological hazard evaluation indicator, and xi is the information value of the i-th geological hazard evaluation indicator.

2.5.4. Risk Assessment Method

The risk assessment of geological hazards refers to the calculation of the probability of hazard occurrence by incorporating external triggering factors, such as rainfall and earthquakes, into susceptibility analysis through statistical models [56]. This process not only evaluates the potential susceptibility of hazards within a region but also integrates triggering conditions, such as extreme climatic events and seismic activities, to quantitatively assess the actual threat level posed by the hazards. In this study, rainfall was considered the primary triggering factor for geological hazards in the study area. The Pearson Type III distribution curve was utilized to predict rainfall probabilities, calculating rainfall intensities for return periods of 10, 20, 50, and 100 years. Subsequently, the risk indices of evaluation units under various rainfall conditions were calculated. The formula is as follows:
H i = Y i Y max P i
P i = L L max
In this formula, Hi represents the risk index of the i-th evaluation unit under a specific rainfall condition, while Yi denotes its susceptibility index, Ymax signifies the highest susceptibility index within the study area, Pi refers to the probability of instability under a specific rainfall condition, L corresponds to the maximum monthly rainfall for a specific rainfall condition, and Lmax represents the highest monthly rainfall recorded in the study area.

3. Results

3.1. Correlation Analysis of Evaluation Factors

The correlation analysis of evaluation factors is conducted using statistical methods to quantify the degree of correlation between factors, thereby avoiding potential biases in evaluation results caused by multicollinearity or high correlations [57,58,59]. This ensures the scientific validity and reliability of model calculations. In this study, the Pearson Correlation Coefficient (PCC) method was applied to analyze the correlations among evaluation factors, resulting in a correlation coefficient diagram, as shown in Figure 6. It can be observed from the figure that the |PCC| values of all 11 factors are less than 0.5, indicating that the factors are relatively independent. Therefore, they can be collectively incorporated into the model to more comprehensively evaluate geological hazards.

3.2. Statistical Analysis of Evaluation Factor Distribution

Based on the distribution characteristics of geological hazards in the study area, the disaster-causing effects of the 11 aforementioned factors were analyzed by statistically examining the number and density of geological hazards within different classification ranges of each factor. The statistical results of these factors and hazard points are shown in Figure 7.
According to the statistics, in terms of elevation distribution, the number of hazard points shows a significant concentration within the medium elevation range, with the highest number of hazard points distributed in the 1735–1755 m range. Meanwhile, the highest hazard point density is concentrated in the 948–1164 m range. In terms of slope distribution, the number of hazard points reaches its peak within the 12–18° range, showing a trend where hazard point density gradually increases with the slope angle. Regarding aspect distribution, hazard points exhibit significant variation, with the southeast and west slopes as the primary concentration areas, where both the number and density of hazards are significantly higher than in other directions. For curvature distribution, hazard points are mainly concentrated in convex and concave slopes, with convex slopes having the highest number and density of hazard points. In terms of distance to faults, the number and density of hazard points peak within 0–500 m of faults, gradually decreasing with increasing distance. However, there is a slight increase in hazard occurrence in areas more than 2500 m away from faults, indicating a combined influence of faults and other environmental factors on geological hazards. For distance to rivers, geological hazards are most concentrated in the 200–400 m range, showing dual peaks in both number and density. As the distance increases, the frequency of hazards decreases, but there is a slight increase in the number of hazard points beyond 1000 m, possibly influenced by other environmental factors. For distance to roads, the number and density of hazard points are higher within the 0–200 m range, indicating a significant promoting effect of human activities on geological hazards. As the distance increases, the impact gradually weakens, but the number of hazards significantly rises in areas more than 1000 m away from roads, likely due to the combined effects of other geological and environmental conditions. In terms of engineering geological rock groups, the largest number of hazard points is found in the Sub-hard to Soft Layered Clastic Rock Group, but, due to their large distribution range, the hazard density is relatively low. In contrast, the Softer Thin to Medium Layered Clastic Rock Group with Coal Layers and Dense to Fractured Structure Basalt Group with interbedded coal seams show high values in both hazard number and density, marking them as high-risk areas that require special attention. Regarding geomorphological types, the number and density of hazard points vary significantly, with Erosion Valley Landform showing the highest values. For NDVI distribution, hazard points are more concentrated in the medium NDVI range, but hazard density decreases as the NDVI increases. The highest hazard density, with a value of 0.184 points/km2, is observed in the 0.19–0.33 range. For deformation rate distribution, the number of hazard points is highest in the 3.30–6.94 mm/a range, while hazard density gradually increases with the deformation rate, peaking at 0.075 points/km2 in the 11.24–17.18 mm/a range, indicating that geological hazards are more concentrated in areas with high deformation rates.

3.3. Weight Analysis Using Random Forest Model

The study area includes 167 geological hazard points, labeled as “1”. Additionally, 167 non-hazard points were randomly selected from areas beyond 500 m from the hazard points, labeled as “0”, resulting in a total of 334 sample points. To ensure the robustness of the model, the sample points were randomly divided into a training set and a test set, with 70% of the sample points used as training data and 30% as test data [60]. A random forest model was constructed using SPSS 27 software, with parameter optimization and adjustment. Based on this model, the weights of various evaluation indicators were calculated using Formula (2), both with and without the inclusion of InSAR deformation features. The results, presented in Figure 8, illustrate the relative importance of various factors in predicting geological hazards. Among these, the weight of the InSAR deformation feature was 6.3%, indicating that it has a certain level of influence in geological hazard evaluation. Although the contribution of this factor is relatively small, its role should not be overlooked in the assessment process.

3.4. Susceptibility Evaluation Analysis

3.4.1. Susceptibility Evaluation Results

The susceptibility evaluation of geological hazards was conducted using the IV model and IV-RF model, with and without InSAR deformation data. Based on the commonly used natural breaks method in statistics, the evaluation results were categorized into four levels: low susceptibility, moderate susceptibility, high susceptibility, and very high susceptibility. The results of the geological hazard susceptibility evaluation are shown in Figure 9 and Figure 10. The results indicate that the high and very high susceptibility areas are primarily distributed in regions with large terrain undulations and steep slopes, especially in the mountainous and hilly areas of the northern and central regions. These areas experience concentrated rainfall, loose soil and are heavily influenced by engineering activities, making them prone to landslides and other geological disasters. Low susceptibility areas are mainly located in the flat regions of the southern and southeastern parts, where the terrain is relatively gentle, vegetation coverage is high, and human activities are minimal, resulting in a lower likelihood of geological disasters.
A statistical analysis was conducted to compare the susceptibility evaluation results of different models with hazard point density, as shown in Table 3. The introduction of InSAR deformation variables significantly improved the predictive performance of regional hazard susceptibility. For the IV model, the combined hazard point density in high and very high susceptibility areas increased from 0.229 points/km2 without deformation variables to 0.232 points/km2 with deformation variables, indicating that dynamic factors better capture the characteristics of high-risk areas. In the IV-RF coupled model, the combined hazard point density in high and very high susceptibility areas increased from 0.237 points/km2 without deformation variables to 0.255 points/km2 with deformation variables, further confirming the critical role of dynamic factors in enhancing the model’s predictive capability.
Additionally, when comparing models, regardless of the inclusion of dynamic InSAR deformation variables, the combined hazard point density in high and very high susceptibility areas was consistently higher in the IV-RF coupled model than in the IV model. This result demonstrates the superior capability of the coupled model in capturing the distribution characteristics of hazards in complex regions, particularly in identifying high-risk areas.
Overall, the introduction of InSAR deformation variables significantly enhanced the models’ ability to identify high and very high susceptibility areas. The IV-RF coupled model outperformed the IV model across all metrics, demonstrating stronger susceptibility recognition capabilities. Through the synergistic effect of InSAR deformation variables and the coupled model, the accuracy and reliability of regional hazard risk prediction were comprehensively improved. The hazard points in high and very high susceptibility areas were more densely distributed, ensuring a high degree of consistency between the susceptibility evaluation results and actual verification data, as illustrated in Figure 11.

3.4.2. Evaluation of Susceptibility Model Accuracy

In this study, to further analyze the accuracy of the model results, the scikit-learn library in Python 3.12 was used to compute accuracy, precision, recall, F1-score, and the ROC curve for evaluating the accuracy of geological hazard susceptibility assessments [61]. The results are presented in Table 4 and Figure 12. The findings indicate that the IV-RF-InSAR model outperforms others overall, achieving an accuracy of 0.752, a precision of 0.702, a recall of 0.816, and an F1-score of 0.755. These metrics are consistent with the ROC curve evaluation results, further verifying the superiority of this model. By incorporating InSAR deformation variables into the IV-RF model, the model’s reliability was significantly enhanced, particularly by reducing false negatives, thereby improving its predictive capability.
The horizontal axis of the ROC curve represents specificity, while the vertical axis represents sensitivity [62]. The area under the curve (AUC) is used to measure the predictive capability of the model, with larger AUC values indicating higher model accuracy [63]. The analysis results indicate that the IV-RF coupled model, enhanced by the inclusion of InSAR deformation variables, demonstrated the best performance in this study, achieving an AUC value of 0.805. This finding suggests that incorporating InSAR deformation variables effectively addresses the false-negative issues present in traditional models, thereby improving the reliability of geological hazard susceptibility evaluations to a certain extent. Furthermore, regardless of whether InSAR deformation variables were included, the AUC values of the IV-RF coupled model consistently surpassed those of the traditional IV model, further validating the advantages of the coupled model in geological hazard susceptibility assessments. Consequently, the IV-RF coupled model is recommended for subsequent evaluations of geological hazard risks.

3.5. Risk Assessment Analysis

Rainfall is a critical factor in triggering geological hazards and has a significant impact on the risk assessment of geological hazards in the study area [64]. According to investigations, 95% of geological hazards in the region occur between May and October, which aligns closely with the rainy season. Therefore, to more accurately assess the risks of geological hazards under different rainfall conditions, this study used Pearson Type III curves to derive rainfall conditions for different scenarios as triggering factors and integrated them with the susceptibility results of the IV-RF coupled model. The risk assessment results under 10-year, 20-year, 50-year, and 100-year return period rainfall conditions are shown in Figure 13 and Figure 14, with risk levels categorized as low, moderate, high, and very high risk. The area proportions of these risk zones are presented in Figure 15.
The results show that, without considering InSAR deformation variables, the proportions of low, moderate, high, and very high-risk zones under the 10-year return period rainfall condition were 17.52%, 31.34%, 33.33%, and 17.81%, respectively. For the 20-year return period, these proportions were 14.31%, 27.51%, 33.91%, and 24.27%. Under the 50-year return period, they were 11.33%, 23.61%, 32.97%, and 32.09%, while, for the 100-year return period, the proportions were 9.76%, 21.23%, 31.55%, and 37.46%. With increasing rainfall intensity, the very high-risk zones expanded significantly, indicating that rainfall is a primary triggering factor. However, the cumulative effects of surface deformation were not considered, potentially underestimating the risks in certain areas.
After considering InSAR deformation variables, the proportions of low, moderate, high, and very high-risk zones under the 10-year return period condition were 16.06%, 30.20%, 34.95%, and 18.79%, respectively. For the 20-year return period, these proportions were 13.04%, 25.92%, 35.11%, and 25.93%. Under the 50-year return period, they were 10.39%, 21.86%, 33.23%, and 34.52%, while, for the 100-year return period, the proportions were 8.95%, 19.59%, 31.20%, and 40.26%. The introduction of InSAR deformation data further expanded the very high-risk zones and allowed for the more precise identification of potential geological hazard areas.
In summary, under the influence of rainfall, the proportions of high and very high-risk zones increase significantly with rainfall intensity. For the 100-year return period condition with InSAR deformation integrated, the combined proportion of high and very high-risk zones reached its maximum. This highlights the significant role of extreme rainfall events in triggering geological hazards and further validates the importance of InSAR deformation data in hazard risk assessments. For high and very high-risk zones, it is recommended to strengthen long-term monitoring and disaster prevention measures. Areas with high rainfall intensity and existing surface deformation should be prioritized for prevention and control to mitigate potential geological hazard risks.

4. Discussion

This study explores an innovative approach to geological hazard risk assessment by integrating InSAR deformation data with varying rainfall conditions, using Luoping County in Yunnan Province as a case study. By improving traditional evaluation models, this research contributes to advancing hazard risk assessment methodologies. A systematic evaluation model was established by comprehensively considering geological environmental characteristics, rainfall triggers, and dynamic surface deformation data, enabling the more precise identification of potential hazard-prone areas. The results demonstrate that the IV-RF model, enhanced with InSAR deformation data, significantly outperforms traditional models in susceptibility evaluation accuracy. Hazard risk assessments incorporating rainfall intensities further validated the model’s reliability, revealing that very high-risk areas expand significantly with increasing rainfall intensity. This finding highlights the role of dynamic factors in optimizing predictive capabilities.
In recent years, coupling statistical and machine learning models has become a hotspot in geological hazard assessment research. Compared to traditional expert scoring methods, machine learning models derive weights in a data-driven manner, offering greater accuracy and stability [65,66,67]. In particular, the random forest (RF) model demonstrates stronger robustness when handling complex datasets compared to other machine learning models, such as Support Vector Machines (SVMs) and Neural Networks, especially in terms of its higher tolerance to noise and outliers [68,69,70,71]. Consequently, coupled models such as the IV-RF model are more suitable for hazard assessment in complex geological environments, effectively improving the identification of high-risk areas [24,48,72]. Integrating InSAR deformation data into the geological hazard evaluation addresses the inherent false-negative errors in traditional susceptibility assessments [73]. Conventional models typically rely on static factors and historical disaster data, making it challenging to identify potential rock masses undergoing creep or accelerated deformation. By incorporating surface deformation data derived from InSAR, traditional evaluation models can be refined. Moreover, InSAR deformation points can serve as hazard indicators, significantly improving the timeliness of regional hazard assessments [74,75,76].
In the context of geological hazard risk assessments, traditional methods often focus on the combined effects of static factors and triggering factors (e.g., rainfall, earthquakes) on hazards. However, geological hazards often accompany dynamic surface deformation. Solely relying on static factors and triggering factors can easily overlook the temporal dynamics of these changes, thus limiting the comprehensive prediction of potential risks. Therefore, it is necessary to integrate SBAS-InSAR and GPS monitoring technologies for continuous deformation monitoring. This study integrates dynamic factors with rainfall intensities under various scenarios to analyze multiple hazard drivers systematically, reflecting the triggering mechanisms of geological hazards more accurately. This approach significantly enhances the precision and scientific rigor of hazard prediction and risk assessment.
Nonetheless, this study acknowledges certain limitations. While using InSAR deformation data as dynamic factors improves the rationality and accuracy of hazard evaluations, challenges remain in extracting deformation data due to atmospheric disturbances, vegetation cover, and terrain variability, which lead to the uneven distribution of deformation points. Interpolation processes may introduce errors. Future research should aim to improve InSAR processing capabilities to minimize atmospheric and other errors. Additionally, interpolation methods could be optimized using machine learning or deep learning algorithms to achieve higher precision in restoring the spatial distribution characteristics of deformation data. These advancements will enhance the scientific validity and rationality of hazard risk assessment results.

5. Conclusions

This study introduces InSAR deformation data as a dynamic factor, integrating it with 10 static factors, including elevation, slope, and geomorphology, using the IV and IV-RF models to conduct a geological hazard susceptibility study in Luoping County. Different rainfall conditions were then superimposed to propose a novel method for geological hazard risk assessment under varying rainfall conditions combined with InSAR deformation data. The main conclusions are as follows:
In the geological hazard susceptibility assessment of Luoping County, a comprehensive comparison of evaluation models and the impact of introducing deformation data were conducted. The ROC curve was used to analyze the accuracy of susceptibility evaluation results. The results indicate that the IV-RF model with InSAR deformation data achieves the highest evaluation accuracy, with an AUC value of 0.805, demonstrating excellent precision. The susceptibility zoning results under this model show that the disaster point densities for low, moderate, high, and very high susceptibility zones are 0.012 points/km2, 0.015 points/km2, 0.040 points/km2, and 0.215 points/km2, respectively, aligning well with field validation results. This further verifies the model’s accuracy and applicability, providing a reliable tool for regional geological hazard susceptibility assessment.
Drawing on the results of geological hazard susceptibility zoning, rainfall intensity was selected as a key factor for hazard risk assessment. This study shows that, with increasing rainfall intensity, the areas of high-risk and very high-risk zones in the study region exhibit an expanding trend. When InSAR deformation data are considered, the combined area proportions of high-risk and very high-risk zones are 2.60%, 2.86%, 2.69%, and 2.45% higher under the 10-year, 20-year, 50-year, and 100-year rainfall conditions, respectively, compared to conditions without InSAR deformation data. This analysis underscores the importance of InSAR deformation data and rainfall intensity in initiating geological hazards and shaping their spatial patterns, offering a scientific foundation for more effective disaster prevention and mitigation efforts.
Incorporating InSAR deformation data as a dynamic factor into geological hazard assessments demonstrates superior predictive capabilities in both susceptibility and risk assessments. The inclusion of InSAR deformation data effectively addresses the limitations of traditional evaluation models, significantly improving the accuracy of hazard identification and prevention. This provides essential technical support for efficient early warning and the mitigation of geological hazards.
In conclusion, this study reveals the synergistic effects of InSAR deformation data and rainfall intensity in geological hazard risk assessment, offering new technical methods and theoretical support for disaster monitoring and risk management.

Author Contributions

Conceptualization, H.W. and J.Z.; methodology, H.W.; software, H.W., L.C. and H.S.; validation, H.W., J.Z., L.C. and H.S.; formal analysis, H.W.; investigation, H.W., J.Z., L.C. and H.S.; data curation, H.W.; writing—original draft preparation, H.W.; writing—review and editing, H.W. and J.Z.; visualization, H.W.; supervision, J.Z.; project administration, H.W.; funding acquisition, J.Z. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Due to the nature of this research, the participants in this study did not agree for their data to be shared publicly, so supporting data are not available.

Acknowledgments

The authors thank the ESA for providing free Sentinel-1A datasets.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Lu, Z.; Liu, G.; Song, Z.; Sun, K.; Li, M.; Chen, Y.; Zhao, X.; Zhang, W. Advancements in Technologies and Methodologies of Machine Learning in Landslide Susceptibility Research: Current Trends and Future Directions. Appl. Sci. 2024, 14, 9639. [Google Scholar] [CrossRef]
  2. Ke, C.; He, S.; Qin, Y. Comparison of natural breaks method and frequency ratio dividing attribute intervals for landslide susceptibility mapping. B Eng. Geol. Environ. 2023, 82, 384. [Google Scholar] [CrossRef]
  3. Zhang, S.; Tan, S.; Liu, L.; Ding, D.; Sun, Y.; Li, J. Slope Rock and Soil Mass Movement Geological Hazards Susceptibility Evaluation Using Information Quantity, Deterministic Coefficient, and Logistic Regression Models and Their Comparison at Xuanwei, China. Sustainability 2023, 15, 10466. [Google Scholar] [CrossRef]
  4. Moustakis, Y.; Papalexiou, S.M.; Onof, C.J.; Paschalis, A. Seasonality, intensity, and duration of rainfall extremes change in a warmer climate. Earth’s Future 2021, 9, e2020EF001824. [Google Scholar] [CrossRef]
  5. Fowler, H.J.; Lenderink, G.; Prein, A.F.; Westra, S.; Allan, R.P.; Ban, N.; Barbero, R.; Berg, P.; Blenkinsop, S.; Do, H.X. Anthropogenic intensification of short-duration rainfall extremes. Nat. Rev. Earth Environ. 2021, 2, 107–122. [Google Scholar]
  6. Myhre, G.; Alterskjær, K.; Stjern, C.W.; Hodnebrog, Ø.; Marelle, L.; Samset, B.H.; Sillmann, J.; Schaller, N.; Fischer, E.; Schulz, M. Frequency of extreme precipitation increases extensively with event rareness under global warming. Sci. Rep. 2019, 9, 16063. [Google Scholar] [CrossRef]
  7. Yueli, C.; Linna, Z.; Ying, W.; Chengxin, W. Review on forecast methods of rainfall-induced geo-hazards. J. Appl. Meteorolgical Sci. 2019, 30, 142–153. [Google Scholar]
  8. Liu, S.; Zhu, J.; Yang, D.; Ma, B. Comparative study of geological hazard evaluation systems using grid units and slope units under different rainfall conditions. Sustainability 2022, 14, 16153. [Google Scholar] [CrossRef]
  9. Kang, D.; Dan, S.; Hua, Z.; Jingyi, L.; Chenlu, W.; Zhenguo, W.; Shaohua, W. Study on landslide hazard risk in Wenzhou based on slope units and machine learning approaches. Sci. Rep. 2025, 15, 7511. [Google Scholar]
  10. Zhang, M.; Tang, Y. Risk investigation method and practice of geohazards. Geol. Bull. China 2008, 27, 1205–1216. [Google Scholar]
  11. Ma, J.; Wang, X.; Yuan, G. Evaluation of geological hazard susceptibility based on the regional division information value method. Isprs Int. J. Geo-Inf. 2023, 12, 17. [Google Scholar] [CrossRef]
  12. Bai, G.; Yang, X.; Zhu, J.; Zhu, C.; Kang, X.; Sun, B.; Zhou, Y. Susceptibility assessment of geological hazards in Wuhua District of Kuming, China using the weight evidence method. Chin. J. Geol. Hazard Control. 2022, 33, 128–138. [Google Scholar]
  13. Fan, W.; Wei, X.; Cao, Y.; Zheng, B. Landslide susceptibility assessment using the certainty factor and analytic hierarchy process. J. Mt. Sci. 2017, 14, 906–925. [Google Scholar] [CrossRef]
  14. Yuan, X.; Liu, C.; Nie, R.; Yang, Z.; Li, W.; Dai, X.; Cheng, J.; Zhang, J.; Ma, L.; Fu, X. A comparative analysis of certainty factor-based machine learning methods for collapse and landslide susceptibility mapping in Wenchuan County, China. Remote Sens. 2022, 14, 3259. [Google Scholar] [CrossRef]
  15. Chen, X.; Chen, W. GIS-based landslide susceptibility assessment using optimized hybrid machine learning methods. Catena 2021, 196, 104833. [Google Scholar]
  16. Meng, X.; Liu, X.; Wang, Y.; Zhang, H.; Guo, X. Submarine landslide susceptibility assessment integrating frequency ratio with supervised machine learning approach. Appl. Ocean. Res. 2024, 153, 104237. [Google Scholar] [CrossRef]
  17. Dong, L.; Zhang, J.; Zhang, Y.; Zhang, B. Landslide risk assessment in mining areas using hybrid machine learning methods under fuzzy environment. Ecol. Indic. 2024, 167, 112736. [Google Scholar]
  18. Cemiloglu, A.; Zhu, L.; Mohammednour, A.B.; Azarafza, M.; Nanehkaran, Y.A. Landslide susceptibility assessment for Maragheh County, Iran, using the logistic regression algorithm. Land 2023, 12, 1397. [Google Scholar] [CrossRef]
  19. Owerko, T.; Kuras, P.; Ortyl, A. Atmospheric correction thresholds for ground-based radar interferometry deformation monitoring estimated using time series analyses. Remote Sens. 2020, 12, 2236. [Google Scholar] [CrossRef]
  20. Liu, Z.; Qiu, H.; Zhu, Y.; Liu, Y.; Yang, D.; Ma, S.; Zhang, J.; Wang, Y.; Wang, L.; Tang, B. Efficient identification and monitoring of landslides by time-series InSAR combining single-and multi-look phases. Remote Sens. 2022, 14, 1026. [Google Scholar] [CrossRef]
  21. Pieraccini, M.; Miccinesi, L. Ground-based radar interferometry: A bibliographic review. Remote Sens. 2019, 11, 1029. [Google Scholar] [CrossRef]
  22. Morishita, Y.; Lazecky, M.; Wright, T.J.; Weiss, J.R.; Elliott, J.R.; Hooper, A. LiCSBAS: An open-source InSAR time series analysis package integrated with the LiCSAR automated Sentinel-1 InSAR processor. Remote Sens. 2020, 12, 424. [Google Scholar] [CrossRef]
  23. Xie, F.; Wang, W.; Sun, X.; Xie, S.; Wei, L. Optimized Airborne Millimeter-Wave InSAR for Complex Mountain Terrain Mapping. Sensors 2025, 25, 424. [Google Scholar] [CrossRef]
  24. Huang, X.; Hu, Z.; Qiao, P.; Zhang, Y.; Wang, X. Geological disaster identification and monitoring along the Ordos platform pipeline based on SBAS-InSAR. Bull. Surv. Mapp. 2024, 7, 60–64. [Google Scholar]
  25. Alatza, S.; Papoutsis, I.; Paradissis, D.; Kontoes, C.; Papadopoulos, G.A. Multi-temporal InSAR analysis for monitoring ground deformation in Amorgos island, Greece. Sensors 2020, 20, 338. [Google Scholar] [CrossRef]
  26. Sharma, G.; Singh, M.S.; Nayak, K.; Dutta, P.P.; Sarma, K.K.; Aggarwal, S.P. Earthquake Damage Susceptibility Analysis in Barapani Shear Zone Using InSAR, Geological, and Geophysical Data. Geosciences 2025, 15, 45. [Google Scholar] [CrossRef]
  27. Nayak, K.; López-Urías, C.; Romero-Andrade, R.; Sharma, G.; Guzmán-Acevedo, G.M.; Trejo-Soto, M.E. Ionospheric Total Electron Content (TEC) anomalies as earthquake precursors: Unveiling the geophysical connection leading to the 2023 Moroccan 6.8 Mw earthquake. Geosciences 2023, 13, 319. [Google Scholar] [CrossRef]
  28. Ciampalini, A.; Raspini, F.; Lagomarsino, D.; Catani, F.; Casagli, N. Landslide susceptibility map refinement using PSInSAR data. Remote Sens. Environ. 2016, 184, 302–315. [Google Scholar]
  29. Li, P.; Wang, H.; Li, H.; Ni, Z.; Deng, H.; Sui, H.; Xu, G. Refined Landslide Susceptibility Mapping Considering Land Use Changes and InSAR Deformation: A Case Study of Yulin City, Guangxi. Remote Sens. 2024, 16, 3016. [Google Scholar] [CrossRef]
  30. Li, Y.; Zhang, Z.; Gong, S.; Liu, M.; Zhao, Y. Risk assessment of rainstorm disasters under different return periods: A case study of Bohai Rim, China. Ocean. Coast. Manag. 2020, 187, 105107. [Google Scholar]
  31. Qin, Z.; Zhou, X.; Li, M.; Tong, Y.; Luo, H. Landslide susceptibility mapping based on resampling method and FR-CNN: A case study of Changdu. Land 2023, 12, 1213. [Google Scholar] [CrossRef]
  32. Potin, P.; Rosich, B.; Miranda, N.; Grimont, P.; Shurmer, I.; O Connell, A.; Krassenburg, M.; Gratadour, J. Copernicus Sentinel-1 Constellation Mission Operations Status. In Proceedings of the IGARSS 2019—2019 IEEE International Geoscience and Remote Sensing Symposium, Yokohama, Japan, 28 July–2 August 2019; IEEE: Piscataway, NJ, USA, 2019; pp. 5385–5388. [Google Scholar]
  33. Niro, F.; Goryl, P.; Dransfeld, S.; Boccia, V.; Gascon, F.; Adams, J.; Themann, B.; Scifoni, S.; Doxani, G. European Space Agency (ESA) calibration/validation strategy for optical land-imaging satellites and pathway towards interoperability. Remote Sens. 2021, 13, 3003. [Google Scholar] [CrossRef]
  34. Bringi, V.; Zrnic, D. Polarization weather radar development from 1970–1995: Personal reflections. Atmosphere 2019, 10, 714. [Google Scholar] [CrossRef]
  35. Dun, J.; Feng, W.; Yi, X.; Zhang, G.; Wu, M. Detection and mapping of active landslides before impoundment in the Baihetan Reservoir Area (China) based on the time-series InSAR method. Remote Sens. 2021, 13, 3213. [Google Scholar] [CrossRef]
  36. Li, X.; Li, W.; Wu, Z.; Xu, Q.; Zheng, D.; Dong, X.; Lu, H.; Shan, Y.; Zhou, S.; Yu, W. Identification and Deformation Characteristics of Active Landslides at Large Hydropower Stations at the Early Impoundment Stage: A Case Study of the Lianghekou Reservoir Area in Sichuan Province, Southwest China. Remote Sens. 2024, 16, 3175. [Google Scholar] [CrossRef]
  37. Marchese, F.; Filizzola, C.; Lacava, T.; Falconieri, A.; Faruolo, M.; Genzano, N.; Mazzeo, G.; Pietrapertosa, C.; Pergola, N.; Tramutoli, V. Mt. Etna paroxysms of February–April 2021 monitored and quantified through a multi-platform satellite observing system. Remote Sens. 2021, 13, 3074. [Google Scholar] [CrossRef]
  38. Park, S.W.; Hong, S.H. Nonlinear modeling of subsidence from a decade of InSAR time series. Geophys. Res. Lett. 2021, 48, e2020GL090970. [Google Scholar]
  39. Brandt, J.T.; Sneed, M.; Danskin, W.R. Detection and measurement of land subsidence and uplift using interferometric synthetic aperture radar, San Diego, California, USA, 2016–2018. Proc. Int. Assoc. Hydrol. Sci. 2020, 382, 45–49. [Google Scholar] [CrossRef]
  40. Roberts, N.J.; Rabus, B.T.; Clague, J.J.; Hermanns, R.L.; Guzmán, M.; Minaya, E. Changes in ground deformation prior to and following a large urban landslide in La Paz, Bolivia, revealed by advanced InSAR. Nat. Hazards Earth Syst. Sci. 2019, 19, 679–696. [Google Scholar] [CrossRef]
  41. Mancini, F.; Grassi, F.; Cenni, N. A workflow based on SNAP–StaMPS open-source tools and GNSS data for PSI-Based ground deformation using dual-orbit sentinel-1 data: Accuracy assessment with error propagation analysis. Remote Sens. 2021, 13, 753. [Google Scholar] [CrossRef]
  42. Matano, F. Analysis and classification of natural and human-induced ground deformations at regional scale (Campania, Italy) detected by satellite synthetic-aperture radar interferometry archive datasets. Remote Sens. 2019, 11, 2822. [Google Scholar] [CrossRef]
  43. Yang, K.; Xue, X.; Li, S. Geological hazard susceptibility assessment by incorporating information value into GA optimized SVM model. Saf. Environ. Eng. 2022, 29, 109–118. [Google Scholar]
  44. Jingsong, Y.I.; Feng, W.; Cheng, Y.; Zhang, Y. Study on the risk assessment of geological disasters in alpine valley area: A case study in Aba County, Sichuan Province. Chin. J. Geol. Hazard Control. 2022, 33, 134–142. [Google Scholar]
  45. Skilodimou, H.D.; Bathrellos, G.D.; Chousianitis, K.; Youssef, A.M.; Pradhan, B. Multi-hazard assessment modeling via multi-criteria analysis and GIS: A case study. Environ. Earth Sci. 2019, 78, 1–21. [Google Scholar]
  46. Zhi, Z.; Liu, G.; Chen, Q.; Xia, X.; Zhou, Q. Spatial Prediction of Potential Property Loss by Geological Hazards based on Random Forest–A Case Study of Chamdo, Tibet. Plateau Sci. Res. 2023, 7, 21–30. [Google Scholar]
  47. Shu, B.; Liu, Y.; Wang, C.; Zhang, H.; Amani-Beni, M.; Zhang, R. Geological hazard risk assessment and rural settlement site selection using GIS and random forest algorithm. Ecol. Indic. 2024, 166, 11255pr4. [Google Scholar]
  48. Li, R.; Tan, S.; Zhang, M.; Zhang, S.; Wang, H.; Zhu, L. Geological Disaster Susceptibility Evaluation Using a Random Forest Empowerment Information Quantity Model. Sustainability 2024, 16, 765. [Google Scholar] [CrossRef]
  49. Su, Y.; Rong, G.; Ma, Y.; Chi, J.; Liu, X.; Zhang, J.; Li, T. Hazard assessment of earthquake disaster chains based on deep learning—A case study of Mao County, Sichuan province. Front. Earth Sci. 2022, 9, 683903. [Google Scholar] [CrossRef]
  50. Sahin, E.K. Assessing the predictive capability of ensemble tree methods for landslide susceptibility mapping using XGBoost, gradient boosting machine, and random forest. SN Appl. Sci. 2020, 2, 1308. [Google Scholar] [CrossRef]
  51. Arsham, A.; Rosenberg, P.; Little, M. Effects of stopping criterion on the growth of trees in regression random forests. New Engl. J. Stat. Data Sci. 2022, 1, 46. [Google Scholar] [CrossRef]
  52. Liu, Y.; Yuan, A.; Bai, Z.; Zhu, J. GIS-based landslide susceptibility mapping using frequency ratio and index of entropy models for She County of Anhui Province, China. Appl. Rheol. 2022, 32, 22–33. [Google Scholar]
  53. Hasanah, S.H.; Julianti, E. Analysis of CART and Random Forest on Statistics Student Status at Universitas Terbuka. INTENSIF J. Ilm. Penelit. Dan Penerapan Teknol. Sist. Inf. 2022, 6, 56–65. [Google Scholar] [CrossRef]
  54. Zhao, B.; Ge, Y.; Chen, H. Landslide susceptibility assessment for a transmission line in Gansu Province, China by using a hybrid approach of fractal theory, information value, and random forest models. Environ. Earth Sci. 2021, 80, 441. [Google Scholar]
  55. Chen, T.; Zhu, L.; Niu, R.; Trinder, C.J.; Peng, L.; Lei, T. Mapping landslide susceptibility at the Three Gorges Reservoir, China, using gradient boosting decision tree, random forest and information value models. J. Mt. Sci. 2020, 17, 670–685. [Google Scholar] [CrossRef]
  56. Shuai, L.; Jie, Z.; De, Y. Geological hazard risk assessment of collapse and landslide under different rainfall conditions. Bull. Geol. Sci. Technol. 2024, 43, 253–267. [Google Scholar]
  57. Li, Y.; Mei, H.; Ren, X.; Hu, X.; Li, M. Geological disaster susceptibility evaluation based on certainty factor and support vector machine. J. Geo-Inf. Sci. 2018, 20, 1699–1709. [Google Scholar]
  58. Wang, H.; Wang, X.; Zhang, C.; Wang, C.; Li, S. Analysis on the susceptibility of environmental geological disasters considering regional sustainable development. Environ. Sci. Pollut. R. 2023, 30, 9749–9762. [Google Scholar]
  59. Bravo-López, E.; Fernández Del Castillo, T.; Sellers, C.; Delgado-García, J. Analysis of conditioning factors in cuenca, ecuador, for landslide susceptibility maps generation employing machine learning methods. Land 2023, 12, 1135. [Google Scholar] [CrossRef]
  60. Liu, S.; He, B.; Wang, T.; Liu, J.; Cao, J.; Wang, H.; Zhang, S.; Li, K.; Li, R.; Zhang, Y. Development characteristics and susceptibility assessment of coseismic geological hazards of Jishishan M S 6.2 earthquake, Gansu Province. China Int. J. 2024, 30, 314–331. [Google Scholar]
  61. Xing, F.; Li, N.; Zhao, B.; Xiang, H.; Chen, Y. An Investigation into the Susceptibility to Landslides Using Integrated Learning and Bayesian Optimization: A Case Study of Xichang City. Sustainability 2024, 16, 9085. [Google Scholar] [CrossRef]
  62. Ullah, I.; Aslam, B.; Shah, S.H.I.A.; Tariq, A.; Qin, S.; Majeed, M.; Havenith, H. An integrated approach of machine learning, remote sensing, and GIS data for the landslide susceptibility mapping. Land 2022, 11, 1265. [Google Scholar] [CrossRef]
  63. Zhang, J.; Qian, J.; Lu, Y.; Li, X.; Song, Z. Study on Landslide Susceptibility Based on Multi-Model Coupling: A Case Study of Sichuan Province, China. Sustainability 2024, 16, 6803. [Google Scholar] [CrossRef]
  64. Picarelli, L.; Olivares, L.; Damiano, E.; Darban, R.; Santo, A. The effects of extreme precipitations on landslide hazard in the pyroclastic deposits of Campania Region: A review. Landslides 2020, 17, 2343–2358. [Google Scholar]
  65. Huang, S.; Dou, H.; Jian, W.; Guo, C.; Sun, Y. Spatial prediction of the geological hazard vulnerability of mountain road network using machine learning algorithms. Geomat. Nat. Hazards Risk 2023, 14, 2170832. [Google Scholar]
  66. Benbouras, M.A.; Sadoudi, L.; Leghouchi, A. Prediction of the Resilient Modulus of Subgrade Soil Using Machine-Learning Techniques. Urbanism. Arhitectura. Constr. 2025, 16, 1–14. [Google Scholar]
  67. Fanos, A.M.; Pradhan, B.; Alamri, A.; Lee, C. Machine learning-based and 3d kinematic models for rockfall hazard assessment using LiDAR data and GIS. Remote Sens. 2020, 12, 1755. [Google Scholar] [CrossRef]
  68. Bezdan, A.; Bezdan, J.; Marković, M.; Mirčetić, D.; Baumgertel, A.; Salvai, A.; Blagojević, B. An objective methodology for waterlogging risk assessment based on the entropy weighting method and machine learning. Catena 2025, 249, 108618. [Google Scholar]
  69. Matougui, Z.; Zouidi, M. A temporal perspective on the reliability of wildfire hazard assessment based on machine learning and remote sensing data. Earth Sci. Inform. 2025, 18, 1–20. [Google Scholar]
  70. Al-Sheriadeh, M.S.; Daqdouq, M.A. Robustness of machine learning algorithms to generate flood susceptibility maps for watersheds in Jordan. Geomat. Nat. Hazards Risk 2024, 15, 2378991. [Google Scholar]
  71. Moghim, S.; Mehrabi, M. Wildfire assessment using machine learning algorithms in different regions. Fire Ecol. 2024, 20, 1–20. [Google Scholar]
  72. Huang, W.; Qiu, H.; Wu, W.; Qin, Y.; Zhou, X.; Zhang, Y.; Ullah, M.; He, Y. Enhancing the Performance of Landslide Susceptibility Mapping with Frequency Ratio and Gaussian Mixture Model. Land 2024, 13, 1039. [Google Scholar] [CrossRef]
  73. Dai, C.; Li, W.; Lu, H.; Zhang, S. Landslide hazard assessment method considering the deformation factor: A case study of Zhouqu, Gansu Province, Northwest China. Remote Sens 2023, 15, 596. [Google Scholar]
  74. Ge, Q.; Sun, Q.; Zhang, N.; Hu, J. Evaluation of geological hazard susceptibility of Baiyin City based on MT-InSAR deformation measurements. Geomat. Inf. Sci. Wuhan Univ. 2023, 49, 1434–1443. [Google Scholar]
  75. Hu, X.; Ming, L.; Wu, T.; Liu, B.; Pang, D.; Yin, J.; Song, B.; Ke, F. Evaluation of geological hazard vulnerability in Xining city based on InSAR and informativeness-hierarchical analysis coupled modeling. Bull. Surv. Mapp. 2023, 12, 51–5675. [Google Scholar]
  76. Zhou, D.Y.; Zuo, X.; Xi, W.; Xiao, B.; You, H. Combined SBAS-InSAR and PSO-BP algorithm for evaluating the risk of geological disasters in alpine valley regions. Trans. Chin. Soc. Agric. Eng. 2021, 37, 108–116. [Google Scholar]
Atmosphere 16 00360 g001
Figure 1. Overview of the study area.
Figure 1. Overview of the study area.
Atmosphere 16 00360 g001
Atmosphere 16 00360 g002
Figure 2. Technical flowchart of this study.
Figure 2. Technical flowchart of this study.
Atmosphere 16 00360 g002
Atmosphere 16 00360 g003
Figure 3. SBAS-InSAR technical flowchart.
Figure 3. SBAS-InSAR technical flowchart.
Atmosphere 16 00360 g003
Atmosphere 16 00360 g004
Figure 4. Deformation magnitude level of the study area.
Figure 4. Deformation magnitude level of the study area.
Atmosphere 16 00360 g004
Atmosphere 16 00360 g005
Figure 5. Geological hazard assessment factor maps in this study: (a) elevation, (b) slope, (c) aspect, (d) curvature, (e) distance to faults, (f) distance to rivers, (g) distance to roads, (h) engineering geological rock groups, (i) geomorphology, (j) NDVI, (k) deformation magnitude level.
Figure 5. Geological hazard assessment factor maps in this study: (a) elevation, (b) slope, (c) aspect, (d) curvature, (e) distance to faults, (f) distance to rivers, (g) distance to roads, (h) engineering geological rock groups, (i) geomorphology, (j) NDVI, (k) deformation magnitude level.
Atmosphere 16 00360 g005
Atmosphere 16 00360 g006
Figure 6. Heat map of correlation coefficients of evaluation factors: A–elevation, B–slope, C–aspect, D–curvature, E–distance to faults, F–distance to rivers, G–distance to roads, H–engineering geological rock groups, I–geomorphology, J–NDVI, K–deformation magnitude level.
Figure 6. Heat map of correlation coefficients of evaluation factors: A–elevation, B–slope, C–aspect, D–curvature, E–distance to faults, F–distance to rivers, G–distance to roads, H–engineering geological rock groups, I–geomorphology, J–NDVI, K–deformation magnitude level.
Atmosphere 16 00360 g006
Atmosphere 16 00360 g007
Figure 7. Distribution of evaluation factors and hazards in the study area: (a) elevation, (b) slope, (c) aspect, (d) curvature, (e) distance to faults, (f) distance to rivers, (g) distance to roads, (h) NDVI, (i) engineering geological rock groups, (j) geomorphology, (k) deformation magnitude level.
Figure 7. Distribution of evaluation factors and hazards in the study area: (a) elevation, (b) slope, (c) aspect, (d) curvature, (e) distance to faults, (f) distance to rivers, (g) distance to roads, (h) NDVI, (i) engineering geological rock groups, (j) geomorphology, (k) deformation magnitude level.
Atmosphere 16 00360 g007
Atmosphere 16 00360 g008
Figure 8. Weight map of evaluation factors in the study area: (a) without InSAR deformation and (b) considering InSAR deformation.
Figure 8. Weight map of evaluation factors in the study area: (a) without InSAR deformation and (b) considering InSAR deformation.
Atmosphere 16 00360 g008
Atmosphere 16 00360 g009
Figure 9. Geological hazard susceptibility map based on IV: (a) without InSAR deformation and (b) considering InSAR deformation.
Figure 9. Geological hazard susceptibility map based on IV: (a) without InSAR deformation and (b) considering InSAR deformation.
Atmosphere 16 00360 g009
Atmosphere 16 00360 g010
Figure 10. Geological hazard susceptibility map based on IV-RF: (a) without InSAR deformation and (b) considering InSAR deformation.
Figure 10. Geological hazard susceptibility map based on IV-RF: (a) without InSAR deformation and (b) considering InSAR deformation.
Atmosphere 16 00360 g010
Atmosphere 16 00360 g011
Figure 11. Comparison between susceptibility evaluation results and field survey observations.
Figure 11. Comparison between susceptibility evaluation results and field survey observations.
Atmosphere 16 00360 g011
Atmosphere 16 00360 g012
Figure 12. ROC curves of different models.
Figure 12. ROC curves of different models.
Atmosphere 16 00360 g012
Atmosphere 16 00360 g013
Figure 13. Geological hazard risk map under different rainfall conditions without considering InSAR deformation: (a) 10-year return period risk level, (b) 20-year return period risk level, (c) 50-year return period risk level, (d) 100-year return period risk level.
Figure 13. Geological hazard risk map under different rainfall conditions without considering InSAR deformation: (a) 10-year return period risk level, (b) 20-year return period risk level, (c) 50-year return period risk level, (d) 100-year return period risk level.
Atmosphere 16 00360 g013
Atmosphere 16 00360 g014
Figure 14. Geological hazard risk map under different rainfall conditions considering InSAR deformation: (a) 10-year return period risk level, (b) 20-year return period risk level, (c) 50-year return period risk level, (d) 100-year return period risk level.
Figure 14. Geological hazard risk map under different rainfall conditions considering InSAR deformation: (a) 10-year return period risk level, (b) 20-year return period risk level, (c) 50-year return period risk level, (d) 100-year return period risk level.
Atmosphere 16 00360 g014
Atmosphere 16 00360 g015
Figure 15. Proportion of geological hazard risk levels under different rainfall conditions: (a) without InSAR deformation and (b) considering InSAR deformation.
Figure 15. Proportion of geological hazard risk levels under different rainfall conditions: (a) without InSAR deformation and (b) considering InSAR deformation.
Atmosphere 16 00360 g015
Table 1. List of data sources.
Table 1. List of data sources.
Data NameData Source
Elevation, Slope, Aspect, and CurvatureBased on ASF 12.5 m DEM
Faults and Lithology1:200,000 geological map
River DistributionNational catalogue service for geographic information
Road NetworkOpen street map
NDVIDerived from Landsat 8 remote sensing imagery
Deformation MagnitudeBased on Sentinel-1A satellite data
RainfallNational Tibetan Plateau/Third Pole Environment Data Center
Table 2. The maximum monthly rainfall under different rainfall conditions in each township of the study area.
Table 2. The maximum monthly rainfall under different rainfall conditions in each township of the study area.
Rainfall Stations10-Year (mm)20-Year (mm)50-Year (mm)100-Year (mm)
Zhongshan324.88340.12356.82367.70
Laochang320.78336.93354.82366.59
Lubuge312.00329.47349.50363.08
Banqiao317.34333.47351.46363.37
Agang300.97317.29335.83348.30
Majie310.90327.22345.58357.81
Changdi321.67337.38354.67365.98
Fule321.15337.08354.73366.33
Dashuijing307.85324.84344.27357.39
Jiuwuji314.89331.11349.37361.54
Lushan308.11325.31344.94358.20
Jiulong312.44329.35348.44361.21
Luoxiong308.11325.31344.94358.20
Table 3. Division table of geological hazard susceptibility zones in the study area.
Table 3. Division table of geological hazard susceptibility zones in the study area.
IV
Susceptibility LevelArea ProportionNumber of Hazard PointsDisaster Point Density (Units/km2)
Without InSAR deformationLow16.27%50.010
Moderate32.25%150.015
High32.88%430.043
Very high18.60%1040.186
Considering InSAR deformationLow16.43%50.010
Moderate32.28%150.015
High32.79%400.040
Very high18.50%1070.192
IV-RF
Susceptibility LevelArea ProportionNumber of Hazard PointsDisaster Point Density (Units/km2)
Without InSAR deformationLow17.41%70.013
Moderate31.41%160.017
High33.41%370.037
Very high17.77%1070.200
Considering InSAR deformationLow16.19%60.012
Moderate31.80%140.015
High36.11%440.040
Very high15.90%1030.215
Table 4. Comparison of accuracy evaluation metrics for the susceptibility assessment model in the study area.
Table 4. Comparison of accuracy evaluation metrics for the susceptibility assessment model in the study area.
MetricsAccuracyPrecisionRecallF1 Score
IV0.7230.7060.7350.720
IV-InSAR0.7130.7000.7140.707
IV-RF0.7230.6910.7760.731
IV-RF-InSAR0.7520.7020.8160.755
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Wang, H.; Zhu, J.; Chen, L.; Shi, H. Incorporating Dynamic Factors in Geological Hazard Risk Assessment: Integrating InSAR Deformation and Rainfall Conditions. Atmosphere 2025, 16, 360. https://doi.org/10.3390/atmos16040360

AMA Style

Wang H, Zhu J, Chen L, Shi H. Incorporating Dynamic Factors in Geological Hazard Risk Assessment: Integrating InSAR Deformation and Rainfall Conditions. Atmosphere. 2025; 16(4):360. https://doi.org/10.3390/atmos16040360

Chicago/Turabian Style

Wang, Hui, Jieyong Zhu, Likun Chen, and Haohan Shi. 2025. "Incorporating Dynamic Factors in Geological Hazard Risk Assessment: Integrating InSAR Deformation and Rainfall Conditions" Atmosphere 16, no. 4: 360. https://doi.org/10.3390/atmos16040360

APA Style

Wang, H., Zhu, J., Chen, L., & Shi, H. (2025). Incorporating Dynamic Factors in Geological Hazard Risk Assessment: Integrating InSAR Deformation and Rainfall Conditions. Atmosphere, 16(4), 360. https://doi.org/10.3390/atmos16040360

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop