Exploring the Impact of Multitemporal DEM Data on the Susceptibility Mapping of Landslides

Featured Application: Landslide susceptibility assessment and other geological disaster assessment issues. Abstract: Digital elevation models (DEMs) are fundamental data models used for susceptibility assessment of landslides. Due to landscape change and reshaping processes, a DEM can show obvious temporal variation and has a signiﬁcant inﬂuence on assessment results. To explore the impact of DEM temporal variation on hazard susceptibility, the southern area of Sichuan province in China is selected as a study area. Multitemporal DEM data spanning over 17 years are collected and the topographic variation of the landscape in this area is investigated. Multitemporal susceptibility maps of landslides are subsequently generated using the widely accepted logistic regression model (LRM). A positive correlation between the topographic variation and landslide susceptibility that was supported by previous studies is quantitatively veriﬁed. The ratio of the number of landslides to the susceptibility level areas (RNA) in which the hazards occur is introduced. The RNA demonstrates a general decrease in the susceptibility level from 2000 to 2009, while the ratio of the decreased level is more than ﬁfteen times greater than that of the ratio of the increased level. The impact of the multitemporal DEM on susceptibility mapping is demonstrated to be signiﬁcant. As such, susceptibility assessments should use DEM data at the time of study.


Introduction
Geohazards are some of the most uncontrolled impacts on local and global economies, as well as on people's livelihoods. Almost nowhere on the planet is free from the damage of geohazards [1]. An essential component of predicting possible geohazard zones is the identification of an area that is vulnerable to future landslides [2,3]. Landslides are common geohazards that destroy local resources and environments. Therefore, susceptibility assessments of landslides have been widely investigated to improve their capability for use with these hazards [4,5].
Current studies commonly focus on two major issues regarding the susceptibility assessments of landslides-selection of evaluation indices and establishment of a rational assessment model. Relevant reviews [6][7][8][9] demonstrate that exiting studies have proposed various remarkable models for susceptibility assessment of landslides, such as logistic regression models (LRMs) [10]. Meanwhile, in recent decades, the rapid development of machine learning algorithms in susceptibility assessments The southern area of Sichuan province in China is a region with frequently occurring landslides. The study area covers the three districts of Panzhihua, Liangshan, and Ya'an ( Figure 1). The area is located at a latitude of 26°03'-30°56' north and a longitude of 100°03'-103°52' east. The area also belongs to the Sichuan Basin, adjacent to the Qinghai-Tibet Plateau in the west. It has a complex topography mainly composed of plains, hills, mountains, and plateaus; furthermore, the relative elevation is more than 3200 m. The climate in the study area is a typically humid subtropical monsoonal climate that is hot in summer and mild and humid in winter, with plentiful rainfall and mist. The vertical climate differs greatly, and the annual average temperature distribution gradually increases from northwest to southeast. The average temperature ranges from 5.3 to 15.7 °C, with annual average rainfall ranging from 1500 to 1800 mm. The study area is, therefore, extremely rainy. The rainfall is mainly concentrated from June to October, which accounts for approximately 70%-75% of the annual rainfall. The maximum daily precipitation reaches 300-500 mm. Meanwhile, part of the study area is located in the Himalaya earthquake zone, which suffers from permafrost, avalanches, and landslides. The stratum of the study area is complete (from Archean to Quaternary). The dominant lithology is sedimentary rock, majorly consisting of dolomite, limestone, siliceous rock, shale, and sandstone [33]. The surface water system is developed, and there are many tributaries of the Yangtze River. The groundwater is distributed widely and buried shallowly, which is affected significantly by the rainfall and landform [34].
The Chinese Geological Environment Monitoring Institute collected the landslides information in this region through the China Geological Survey, including locations and occurrence times, and released it to public in its Bulletin of National Geological Hazards [35]. The annual total number of historical landslide events is shown in Figure 2. More than 85% of the historical landslide events were induced by natural environmental factors, such as rainfall and earthquakes, while only a few events were induced by human factors, such as mining and slope cutting. As the source data in the Bulletin of National Geological Hazards are not distinguished by landslide type, the general landslides analyzed in the study are of various types.
In this study, the total number of historical landslide events from 2000 to 2016 and the number of those hazards in various years after 2004 are obtained. However, we can only obtain the total number of historical landslide events for the period between 2000 and 2004. Thus, to better analyze the impact of a DEM on susceptibility assessment, we divide the historical landslide events into two The climate in the study area is a typically humid subtropical monsoonal climate that is hot in summer and mild and humid in winter, with plentiful rainfall and mist. The vertical climate differs greatly, and the annual average temperature distribution gradually increases from northwest to southeast. The average temperature ranges from 5.3 to 15.7 • C, with annual average rainfall ranging from 1500 to 1800 mm. The study area is, therefore, extremely rainy. The rainfall is mainly concentrated from June to October, which accounts for approximately 70%-75% of the annual rainfall. The maximum daily precipitation reaches 300-500 mm. Meanwhile, part of the study area is located in the Himalaya earthquake zone, which suffers from permafrost, avalanches, and landslides. The stratum of the study area is complete (from Archean to Quaternary). The dominant lithology is sedimentary rock, majorly consisting of dolomite, limestone, siliceous rock, shale, and sandstone [33]. The surface water system is developed, and there are many tributaries of the Yangtze River. The groundwater is distributed widely and buried shallowly, which is affected significantly by the rainfall and landform [34].
The Chinese Geological Environment Monitoring Institute collected the landslides information in this region through the China Geological Survey, including locations and occurrence times, and released it to public in its Bulletin of National Geological Hazards [35]. The annual total number of historical landslide events is shown in Figure 2. More than 85% of the historical landslide events were induced by natural environmental factors, such as rainfall and earthquakes, while only a few events were induced by human factors, such as mining and slope cutting. As the source data in the Bulletin of National Geological Hazards are not distinguished by landslide type, the general landslides analyzed in the study are of various types.

Influencing Factors
Selecting the influencing factors is a key step in a landslide susceptibility assessment [36,37]; however, the causes of landslides are complex. To date, hundreds of influencing factors have been identified that are potentially important to susceptibility assessments [38]. For example, areas with

Influencing Factors
Selecting the influencing factors is a key step in a landslide susceptibility assessment [36,37]; however, the causes of landslides are complex. To date, hundreds of influencing factors have been

Influencing Factors
Selecting the influencing factors is a key step in a landslide susceptibility assessment [36,37]; however, the causes of landslides are complex. To date, hundreds of influencing factors have been identified that are potentially important to susceptibility assessments [38]. For example, areas with high elevation and steep slopes are highly prone to landslides [39]. Lithology is also related to landslide occurrences, because different lithologies can withstand different levels of triggering factors. The distances to structure lines, rivers, and roads have important impacts on the spread and size of landslides in the study area [40]. Pourghasemi et al. investigated global susceptibility during the period of 2005-2016 [28]. Nearly 100 factors were summarized to reveal the commonly used factors. In accordance with the relevant studies [29,41,42] and the study area, we select eight influencing factors: elevation, topography, slope, lithology, distance to a structure line, distance to a river, average annual rainfall, and distances to roads. Table 1 lists all the DEM data for the study area and their sources. The Open Topography Facility provided the GMRT data, which is hosted at the San Diego Supercomputer Center, University of California San Diego. This facility has built a strong cyberinfrastructure framework for managing and processing high-resolution topography data from light detection and ranging (LiDAR) (http://opentopo.sdsc.edu/datasets?listAll=true). Meanwhile, the Shuttle Radar Topography Mission (SRTM) DEM and Global Digital Elevation Model (GDEM) DEM were provided by the Geospatial Data Cloud site of the Computer Network Information Center in the Chinese Academy of Sciences (http://www.gscloud.cn/). The DEM data at the same resolution for 1992, 2000, and 2009 were then obtained using a resampling tool in the GIS environment ( Figure 4). There are three resampling methods in the resampling tool, namely nearest neighbor, bilinear interpolation, and cubic convolution methods. Nearest neighbor is selected in the present study because the method is simple, fast, and applicable. high elevation and steep slopes are highly prone to landslides [39]. Lithology is also related to landslide occurrences, because different lithologies can withstand different levels of triggering factors. The distances to structure lines, rivers, and roads have important impacts on the spread and size of landslides in the study area [40]. Pourghasemi et al. investigated global susceptibility during the period of 2005-2016 [28]. Nearly 100 factors were summarized to reveal the commonly used factors. In accordance with the relevant studies [29,41,42] and the study area, we select eight influencing factors: elevation, topography, slope, lithology, distance to a structure line, distance to a river, average annual rainfall, and distances to roads.  Table 1 lists all the DEM data for the study area and their sources. The Open Topography Facility provided the GMRT data, which is hosted at the San Diego Supercomputer Center, University of California San Diego. This facility has built a strong cyberinfrastructure framework for managing and processing high-resolution topography data from light detection and ranging (LiDAR) (http://opentopo.sdsc.edu/datasets?listAll=true). Meanwhile, the Shuttle Radar Topography Mission (SRTM) DEM and Global Digital Elevation Model (GDEM) DEM were provided by the Geospatial Data Cloud site of the Computer Network Information Center in the Chinese Academy of Sciences (http://www.gscloud.cn/). The DEM data at the same resolution for 1992, 2000, and 2009 were then obtained using a resampling tool in the GIS environment ( Figure 4). There are three resampling methods in the resampling tool, namely nearest neighbor, bilinear interpolation, and cubic convolution methods. Nearest neighbor is selected in the present study because the method is simple, fast, and applicable.  The DEM data for 1992, 2000, and 2009 were used to analyze factors affected by the DEM. Meanwhile, the data of other factors, namely lithology, distance to the structure line, distance to the river, and average annual rainfall, remained the same and were provided by China Railway Number 4 Engineering Group Co., Ltd. (745 Heping Road, Wuhan, Hubei, China), and the Roads and Traffic Authority of China. Because of the diversity of lithological layers in the study area, the lithology is divided into four groups in Table 2 [43]. Thus, the zoning maps of the four factors were obtained ( Figure 5). The DEM data for 1992, 2000, and 2009 were used to analyze factors affected by the DEM. Meanwhile, the data of other factors, namely lithology, distance to the structure line, distance to the river, and average annual rainfall, remained the same and were provided by China Railway Number 4 Engineering Group Co., Ltd. (745 Heping Road, Wuhan, Hubei, China), and the Roads and Traffic Authority of China. Because of the diversity of lithological layers in the study area, the lithology is divided into four groups in Table 2 [43]. Thus, the zoning maps of the four factors were obtained ( Figure 5).

Flowchart
The research methodologies applied in the present study are as shown in Figure 6. The flowchart consists of four major steps, as follows: Zoning map of (a) lithology, (b) distance to a structure line, (c) distance to a river, (d) and average annual rainfall.

Flowchart
The research methodologies applied in the present study are as shown in Figure 6. The flowchart consists of four major steps, as follows: a) Data preparation. The multitemporal DEM data, influencing factors, and historical landslide events are prepared; b) Data preprocessing. The multitemporal DEM data is compared to eliminate the noise based on coverage probabilities for confidence intervals; c) Susceptibility assessment of landslides. The assessment results for 1992, 2000, and 2009 are obtained using a LRM; d) Comparison and analysis. The assessment results for 1992, 2000, and 2009 are compared to obtain the impacts of the elevation difference and the historical landslide events on the susceptibility assessment.

Confidence Interval
In the present study, because of the noise in the DEM data reducing the accuracy of data analysis significantly, not all DEM data are reliable. Therefore, the noise needs to be eliminated to determine the reliability of the DEM data. The confidence level of the DEM data is assessed to improve the accuracy of data analysis.
The confidence interval is a commonly-used interval estimation method used for sample statistics, which shows the confidence probability of the measured parameter value [42,44]. The elevation difference is considered as the sample data in the present study. The average value of elevation difference is , and the standard deviation is . The confidence probability is obtained as follows: where is the significance level and the interval (c1, c2) is the confidence interval. Therefore, the confidence interval of the average value is − / , + / , where = / is the corresponding standard score. In general, the confidence probability in the literature [42,45,46] is 90% or 95%, and / is 1.645 or 1.96, respectively.

Confidence Interval
In the present study, because of the noise in the DEM data reducing the accuracy of data analysis significantly, not all DEM data are reliable. Therefore, the noise needs to be eliminated to determine the reliability of the DEM data. The confidence level of the DEM data is assessed to improve the accuracy of data analysis.
The confidence interval is a commonly-used interval estimation method used for sample statistics, which shows the confidence probability of the measured parameter value [42,44]. The elevation difference is considered as the sample data in the present study. The average value of elevation difference is µ, and the standard deviation is σ. The confidence probability is obtained as follows: where α is the significance level and the interval (c 1 , c 2 ) is the confidence interval. Therefore, the confidence interval of the average value is (µ − σZ α/2 , µ + σZ α/2 ), where = Z α/2 is the corresponding standard score. In general, the confidence probability in the literature [42,45,46] is 90% or 95%, and Z α/2 is 1.645 or 1.96, respectively.

LRM
The LRM is a statistical model used to predict the probability of a categorical occurrence using one or more independent variables [47]. The purpose of LRM is to obtain the relationship between a dependent variable and multiple independent variables that have been identified.
In the present study, the independent variables are the influencing indices of landslides, and the dependent variable is the probability of the landslide occurring. By transforming the universal formula of LRM, Y is obtained: where C 0 is the LRM constant coefficient, C i is the LRM coefficient, I i is the landslide index, Y = ln(P/(1 − P)), and P is the probability of the landslide occurring. The LRM coefficient C i and the LRM formula can be obtained using Statistical Product and Service Solutions (SPSS) software.

Ratio of Number of Landslides to Area (RNA)
RNA denotes the ratio of the number of historical landslide events to the area of the susceptibility levels at which historical landslide events are located. The RNA of the level i is obtained with the following equation: where RNA i is the number of landslides per area at a particular level i; n i is the number of historical landslide events at level i; and A i is the area of level i in the assessment.

Elimination of Noise
As mentioned above, the DEM has a significant effect on susceptibility assessment through three indices, namely elevation, topography, and slope. The changes in the DEM from 1992 to 2000 and from 2000 to 2009 are obtained to explore the impact of the temporal variation of DEM data on susceptibility assessment. Meanwhile, the average values and standard deviation of the elevation differences are obtained from the topography changes revealed by the DEM data. The confidence probability is considered as 95% in this study. The corresponding standard scores and confidence intervals are, therefore, calculated (Table 3)

Susceptibility Assessments of Landslides
The susceptibility assessment maps of the study area in 1992, 2000, and 2009 are obtained in the GIS environment using LRM (Figure 8 and Table 4). The susceptibility is classified into four levels based on the natural breaks classification (NBC) method, which is based on natural groupings inherent in data. The NBC identifies groups of similar values and maximizes the differences between

Susceptibility Assessments of Landslides
The susceptibility assessment maps of the study area in 1992, 2000, and 2009 are obtained in the GIS environment using LRM (Figure 8 and Table 4). The susceptibility is classified into four levels based on the natural breaks classification (NBC) method, which is based on natural groupings inherent in data. The NBC identifies groups of similar values and maximizes the differences between classes. The features are divided into classes whose boundaries are set based on relatively big differences in the data values. Here, level I, level II, level III, and level IV denote low, moderate, high, and very high susceptibility, respectively. The area percentages of various susceptibility levels in the study are shown in Table 4.

Susceptibility Assessments of Landslides
The susceptibility assessment maps of the study area in 1992, 2000, and 2009 are obtained in the GIS environment using LRM (Figure 8 and Table 4). The susceptibility is classified into four levels based on the natural breaks classification (NBC) method, which is based on natural groupings inherent in data. The NBC identifies groups of similar values and maximizes the differences between classes. The features are divided into classes whose boundaries are set based on relatively big differences in the data values. Here, level I, level II, level III, and level IV denote low, moderate, high, and very high susceptibility, respectively. The area percentages of various susceptibility levels in the study are shown in Table 4.    As shown in Figure 8 and Table 4, the results of the susceptibility assessment vary from year to year. The areas with low and moderate susceptibility (levels I and II, respectively) in 1992 are the largest, while the area with very high susceptibility (level IV) is the largest in 2000. The rapid decrease in the level IV area from 2000 to 2009 is noticeable. One possible explanation for the decrease is that the slopes where landslides occurred became stable in the short term, owing to the excessive landslides that occurred from 2000 to 2009; thus, the susceptibility level of the study area decreased after the hazards.
The susceptibility maps are obtained using different DEM-dependent factors, while the DEM-independent factors are constant in the assessment. Comparing these results illustrates the significant influence of the DEM data from different years on the susceptibility assessment. Figure 8 and Table 5 show the susceptibility assessment levels at which historical landslide events occurred from 2010 to 2016. It can be clearly seen in Table 5 that the combined number of historical landslide events designated as levels III and IV is 106, 140, and 154, respectively, in 1992, 2000, and 2009. The RNA for level IV in 2009 is 5.44 × 10 −3 /km 2 , which is the largest among the RNA values over the three years. The most ideal assessment of landslides would be most landslides occurring in level IV areas. There is a significant positive correlation between the RNA and the accuracy of the assessment. Therefore, the results of the susceptibility assessment using the DEM for 2009 are more accurate than the results using the DEM for 1992 and 2000.

Assessment Levels of Historical Landslide Events
It is apparent from Table 5 that the small change in elevation over time has a great influence on the susceptibility assessment of landslides. However, in previous research, the survey data are not explained or the data are significantly different from the assessment time.

Differences between the Assessment Levels
Comparisons are performed using the susceptibility assessments of landslides in the study area from different periods. Figure 9 and Table 6 compare the differences between the assessment results for 1992 and 2000 and between those for 2000 and 2009, showing that the area where the level difference is zero is the largest area. Meanwhile, the area with increased susceptibility is much larger than the area with decreased susceptibility in the period ranging from 1992 to 2000, which is contrary to the results for the period ranging from 2000 to 2009.

Comprehensive Comparison of Assessment Results for 1992, 2000, and 2009
We compare the areas and RNA values of various assessment levels for the years 1992, 2000, and 2009 (Figure 10a,b), and the level differences from 1992 to 2000 and from 2000 to 2009 ( Figure  10c). Figure 10 is quite revealing in several ways. First, the RNA for level IV in 2009 is larger than that in 1992 or 2000. Because of the positive correlation between the RNA and the accuracy of the assessment, the accuracy of the assessment is further exemplified in studies using the DEM for 2009. As Figure 10c shows, there is a significant difference between the two results. The area with increased susceptibility in the period ranging from 1992 to 2000 is larger than that in in the period ranging from 2000 to 2009. The observed increase in the area with increased susceptibility could be attributed to the frequent landslides from 2000 to 2009 (e.g., the various landslides caused by the 2008 earthquake).

Comprehensive Comparison of Assessment Results for 1992, 2000, and 2009
We compare the areas and RNA values of various assessment levels for the years 1992, 2000, and 2009 (Figure 10a,b), and the level differences from 1992 to 2000 and from 2000 to 2009 (Figure 10c). Figure 10 is quite revealing in several ways. First, the RNA for level IV in 2009 is larger than that in 1992 or 2000. Because of the positive correlation between the RNA and the accuracy of the assessment, the accuracy of the assessment is further exemplified in studies using the DEM for 2009. As Figure 10c shows, there is a significant difference between the two results. The area with increased susceptibility in the period ranging from 1992 to 2000 is larger than that in in the period ranging from 2000 to 2009. The observed increase in the area with increased susceptibility could be attributed to the frequent landslides from 2000 to 2009 (e.g., the various landslides caused by the 2008 earthquake).

Impact of Elevation Difference on the Susceptibility Assessment
In addition to the above analysis, we analyze the impact of elevation differences on the susceptibility assessment and the relationships between the elevation differences and the differences in assessment levels. Therefore, the areas with varying level and elevation differences are presented

Impact of Elevation Difference on the Susceptibility Assessment
In addition to the above analysis, we analyze the impact of elevation differences on the susceptibility assessment and the relationships between the elevation differences and the differences in assessment levels. Therefore, the areas with varying level and elevation differences are presented (Table 7).  Table 7 demonstrates that the decrease in elevation results in a decrease in the susceptibility assessment level. There is a positive correlation between the elevation difference and the difference in the assessment level. Regardless of the level difference, the elevation difference is much larger in the area ranging from −20 to 20 m than in other areas.

Impact of Historical Landslide Events on Susceptibility Assessment
Most of the factors influencing susceptibility assessments have been explored in several studies [48][49][50]. However, much of the historical research overlooks the impact of historical landslide events on the assessment [51]. In contrast, because the slopes where landslides occur become stable in the short term following these events, the susceptibility of the area will consequently decrease. Therefore, historical landslide events have a significant effect on susceptibility assessments.
The numbers and percentages of historical landslide events at various assessment levels from 2000 to 2009 and the RNA values of the various assessment levels are investigated, allowing the change in the assessment levels of the historical landslide events to be obtained (Table 8 and Figure 11).
In Figure 11, the circles represent the percentages of historical landslide events or the RNA values at various assessment levels. It can be seen from Figure 11 that over one-third of the historical landslide events (38.71%) occur at a decreasing assessment level. A total of 54.24% of the historical landslide events are within the same assessment level. If the assessment level area is larger, then there may be more historical landslide events within that assessment level. To rule out the possible influence of area size, the RNA is considered to be the most important factor. The RNA at decreased levels (81.31 × 10 −3 /km 2 ) is more than fifteen times greater than the RNA at increased levels (5.25 × 10 −3 /km 2 ). Previous studies [52,53] do not take into account the impact of historical landslide events; therefore, the above results can establish the importance of historical landslide events for susceptibility assessments of landslides.

Advantages and Limitations
The elevation, topography, and slope-depending on the DEM-are usually the main factors affecting susceptibility [23]. The existing studies assume that the influencing factors for susceptibility are constant over time [15,54]. However, the DEM data always change significantly over time owing to landscape reshaping, such as landslides and engineering construction [30,31,55]. A susceptibility assessment is, therefore, closely related to the data source [56], and as such the susceptibility maps of the study area are not consistent.
Despite this, the existing studies have analyzed the susceptibility of landslides using a single time-related data source without considering whether the date of the data source is consistent with the study date [57,58]. Therefore, it is necessary to understand the temporal evolution of influencing factors (e.g., topography) and the influencing mechanisms of the factors on the susceptibility assessment [59,60]. The changes in influencing factors are significant in earthquake-and landslideprone areas [61], such as the study area. Therefore, more research regarding the impacts of factors on the susceptibility of landslides needs to be undertaken. It is better not to evaluate susceptibility simply using constant factors.
The present study performs a preliminarily analysis of the impact of multitemporal DEM on susceptibility and shows that DEM variation has an impact on the susceptibility assessment using DEM-dependent factors. The analysis undertaken here can extend the knowledge regarding the impacts of multitemporal data on susceptibility and can aid preliminary studies on the mechanism causing the influence of factor variation.
DEM variation closely relates to local geotechnical properties, which change the susceptibility assessment of landslides. However, the reasons for DEM variation are debated. The mechanism

Advantages and Limitations
The elevation, topography, and slope-depending on the DEM-are usually the main factors affecting susceptibility [23]. The existing studies assume that the influencing factors for susceptibility are constant over time [15,54]. However, the DEM data always change significantly over time owing to landscape reshaping, such as landslides and engineering construction [30,31,55]. A susceptibility assessment is, therefore, closely related to the data source [56], and as such the susceptibility maps of the study area are not consistent.
Despite this, the existing studies have analyzed the susceptibility of landslides using a single time-related data source without considering whether the date of the data source is consistent with the study date [57,58]. Therefore, it is necessary to understand the temporal evolution of influencing factors (e.g., topography) and the influencing mechanisms of the factors on the susceptibility assessment [59,60]. The changes in influencing factors are significant in earthquake-and landslide-prone areas [61], such as the study area. Therefore, more research regarding the impacts of factors on the susceptibility of landslides needs to be undertaken. It is better not to evaluate susceptibility simply using constant factors.
The present study performs a preliminarily analysis of the impact of multitemporal DEM on susceptibility and shows that DEM variation has an impact on the susceptibility assessment using DEM-dependent factors. The analysis undertaken here can extend the knowledge regarding the impacts of multitemporal data on susceptibility and can aid preliminary studies on the mechanism causing the influence of factor variation.
DEM variation closely relates to local geotechnical properties, which change the susceptibility assessment of landslides. However, the reasons for DEM variation are debated. The mechanism explaining the consequent influences of geotechnical properties on DEM variation remains a scientific challenge. In addition, considering the large spatial extent and period (about 77,000 km 2 , 17 years) in the present study, changes between the three susceptibility maps could not be completely explained by DEM variation. There are also data limitations and difficulties in analyzing the interactions of various factors.

Conclusion
In the present paper, DEM data for 1992, 2000, and 2009 are obtained to evaluate the susceptibility of landslides. Based on the DEM-dependent factors and the same DEM-independent factors, the susceptibility results are evaluated using LRM.
From the assessment, we find that the results are different by using DEM data from different times. The rapid decrease in the area of level IV from 2000 to 2009 is noticeable. The RNA for level IV in 2009 is larger than in 1992 and 2000. The area of increased susceptibility is much larger than that of decreased susceptibility based on the results of the assessment level differences from 1992 to 2000, which differ from the results from 2000 to 2009.
From the assessment results, we conclude that the DEM data have an impact on the susceptibility assessment in the study area. It is also worth noting that the influences of the specific values of the elevation differences on the assessment results are obtained. Meanwhile, the influence of historical landslide events on the susceptibility assessment is obtained by analyzing the hazard data and the differences between the assessment results. The most obvious finding to emerge from the analysis is that the assessment level of the area with historical landslide events decreases. The present study goes some way towards enhancing our understanding of the impacts and the mechanism of multitemporal factors on susceptibility.