GIS-based earthquake-triggered landslide susceptibility mapping 1 with an integrated weighted index model in Jiuzhaigou region of 2 Sichuan Province , China 3

A Mw 6.5 earthquake struck the Jiuzhaigou region of Sichuan Province, China at 21:19 pm on Tuesday, 8 August 11 2017, and triggered a large number of landslides. For mitigating the damages of earthquake-triggered landslides to 12 individuals and infrastructures of the earthquake affected region, a comprehensive landslide susceptibility mapping was 13 attempted with an integrated weighted index model by combining the frequency ratio and the analytical hierarchy process 14 approaches under GIS-based environment in the earthquake heavily attacked Zhangzha town of the Jiuzhaigou region. For 15 this purpose, a total number of 842 earthquake-triggered landslides were visually interpreted and located from Sentinel-2A 16 images acquired before and after the earthquake at first, and then the recognized landslides were randomly split into two 17 groups to establish the earthquake-triggered landslide inventory, among which 80 % of the landslides was used for training 18 the integrated model and the remaining 20 % for validation. Nine landslide controlling factors were considered including 19 slope, aspect, elevation, lithology, distance from faults, distance from rivers, land-use/cover, normalized difference 20 vegetation index and peak ground acceleration. The frequency ratio was utilized to evaluate the contribution of each 21 landslide controlling factor on landslide occurrence, and the analytical hierarchy process was used to analysis the mutual 22 relationship between landslide controlling factors. Finally, the landslide susceptibility map was produced by using the 23 weighted overlay analysis. Furthermore, an area under the curve approach was adopted to comprehensively evaluate the 24 performance of the integrated weighted index model, including the degree of model fit and model predictive capability. The 25 results demonstrated the reliability and feasibility of the integrated weighted index model in earthquake-triggered landslide 26 susceptibility mapping at regional scale. The generated map can help engineers and decision makers assess and mitigate 27 hazards of the earthquake-triggered landslides to individuals and infrastructures of the earthquake affected region. 28 29

Over the last decades, many approaches for landslide susceptibility mapping were proposed, among which the application of (red, green, blue) with the spatial resolution of 10 m were adopted to analysis the image characteristics of earthquake-1 triggered landslides. With the aid of ArcGIS and ENVI tools, the landslide information of the study area was extracted using 2 on-screen visual interpretation on pre-and post-earthquake Sentinel-2A images. In order to ensure the quality of visual 3 interpretation, GF-1 images with spatial resolution of 2 m on January 15, 2017, were used to verify the results. Consequently, 4 a total number of 842 earthquake-triggered landslides were recognized and positioned. Smaller landslides with total pixels 5 less than 20 were not included as they were not clear enough in visual features. We assumed that the distribution of the 6 earthquake-triggered landslides was reasonably accurate and complete at regional scale in order to make the problem 7 tractable. For earthquake-triggered landslide susceptibility mapping, the landslide inventory dataset was randomly split into 8 two groups, among which 80 % (673 landslides) of the recognized landslides was used for training the integrated weighted 9 index model and the remaining 20 % (169 landslides) for validation. 10

Landslide controlling factors 11
The occurrence of landslides is a consequence of geological, meteorological, anthropogenic and triggering factors, 12 commonly referred to as landslide controlling factors (Bai et al., 2010). Standard guidelines for choosing the optimal 13 landslide controlling factors are unavailable, but the scale of analysis, the nature of the study area, the data availability and 14 the quasi-empirical and statistical criterions in literatures can be referenced (Romer and Ferentinou, 2016;Zhou et al., 2016). 15 In this study, slope, aspect, elevation, lithology, distance from faults, distance from rivers, land-use/cover (LULC), 16 normalized difference vegetation index (NDVI) and peak ground acceleration (PGA) were selected as the landslide 17 controlling factors, as shown in Fig. 3 (Su et al., 2015). Landslides become more possible once the slope gradient is higher 21 than 15° (Lee and Min, 2001). In the study area, the slopes were generally steep, with an average slope angle of about 30°. 22 Aspect, referred to the direction of slope faces, is related to soil moisture, surface runoff and vegetation, which indirectly 23 affects landslide development (Zhang et al., 2016). The elevation, as the measure of the land surface height, is a key factor 24 determining gravitational potential energy of terrain and is often considered in relevant studies (Conforti et al. sandstone also exhibited a large number of landslides. In this study, the lithological data was obtained from the geological map at 1: 500,000 scale and was digitized in ArcGIS for further analysis. The distances of a slope from faults as well as from 1 the river channels are also important factors in terms of slope stability (Kanungo et al., 2006). In addition, earthquake-2 triggered landslides are usually found in the vicinity of active faults. Hence, the distances of a slope from geological tectonic 3 zone were often taken into account in slope stability analysis. Fan et al. (2018) had revealed that this earthquake occurred 4 along a previously unknown blind fault probably belonging to a south branch of the Tazang fault or north part of the Huya 5 fault. However, due to its great uncertainty, this blind fault was not taken into account in the study area. In this study, the 6 faults were digitized from the geological map at 1: 500,000 scale, and the river channels were interpreted from remote 7 sensing images. And the LULC map is one of controlling factors that pose direct impact on the occurrence of landslides 8 (Song et al., 2012;Mansouri Daneshvar, 2014). In this study, the LULC map was downloaded from the Geographical 9 Information Monitoring Cloud Platform. 10 Vegetation coverage poses effect on soil water erosion, which indirectly affects the occurrence of landslides. NDVI, as the 11 measure of vegetation coverage, is usually adopted in landslide susceptibility analysis (Siqueira et al., 2015). The NDVI is 12 calculated from these individual measurements as follows: 13 Where, DNNIR stands for the spectral reflectance derived from the measured radiances in the near-infrared regions (NIR), and 15 DNR stands for the spectral reflectance derived from the measured radiances in the visible (Red) regions. 16 In this study, the NDVI map was generated from the Landsat-8 image acquired on April 8, 2017 over the study area. To ensure the consistency and easy process of these data, all factor layers were converted into raster data format (GeoTIFF) 22 with an identical spatial projection (WGS84 datum) and resampled to a resolution of 30 m by ENVI 5.3 and ArcGIS 10.2. 23

Methodology 24
In this study, an integrated weighted index model was developed as a complete landslide susceptibility model by combining 25

Analytical hierarchy process (AHP) 1
The AHP method, developed by Saaty (Saaty, 1977), is an important multiple criteria decision-making method (Vaidya and 2 Kumar, 2006), which has been applied for landslide susceptibility assessment for many years (Akgun, 2012; Barredo et al., In the AHP, a complex non-structural problem is first broken down into several component factors. Then, based on the 5 expert's prior experience and knowledge, a pair-wise comparison matrix can be constructed through comparing the relative 6 importance of each factor (Vargas, 1990). An underlying 9-point recording scale is used to rate the relative importance of 7 factors (Mansouri Daneshvar, 2014). Specifically, when a factor is more important than another, the score varies between 1 8 and 9. Conversely, the score varies between 1/2 and 1/9. The higher the score, the greater the importance of the factor. With 9 the help of a pair-wise comparison matrix, the contribution of factors can be converted into numerical values. Note that a 10 consistency check of comparison matrix needs to be carried out, and the Consistency Ratio (CR) of less than 0.1 is generally 11 accepted. 12 In this study, the relative importance of landslide controlling factors was determined from the prior experience and 13 knowledge of experts. Since the knowledge source varies from person to person, the best judgment always comes from an  (Table 2) and a 17 general consensus achieved by experts. Weights of factors were determined in the process of a pair-wise comparison matrix 18 using Python software, as shown in Table 2. The Consistency Ratio (CR) for this study was 0.017, which showed that the 19 pair-wise comparison matrix satisfied the consistency requirement. 20

Frequency ratio (FR) 21
The FR method is one of the most widely used approaches to assess the landslide susceptibility at regional scale (Guo et al., be investigated by using the FR method. Therefore, the FR values of each controlling factor category were calculated from 29 their relationship with landslide occurrence locations as illustrated in Table 3. The average value of FR is 1 so that a value 30 larger than one represents a higher correlation and those less than it, a lower correlation (Romer and Ferentinou, 2016). 31 The FR value can be calculated as follows (Ghobadi et al., 2017): 32 (2) 1 Where, ( ) represents number of grid cells recognized as landslides in class i, and ( ) represents total number 2 of grid cells belonging to class i in the whole area; while ∑ ( ) stands for the total number of grid cells recognized as 3 landslides in the whole area, and ∑ ( ) represents total number of grid cells in the whole area. 4

Integrated weighted index 5
The integrated weighted index is considered to measure the probability of slope failures. By combining FR and AHP 6 methods, the integrated weighted model can assess the correlation between the controlling factors and also the influence of 7 each landslide controlling factor on landslide occurrence. 8 The integrated weighted index can be calculated as follows: 9 Where, m stands for number of controlling factors, Wi is the weight of each controlling factor calculated by the AHP method, 11 FRi is the FR value of the controlling factor calculated by the FR method. 12 In this study, the values of Wi and FRi were used to obtain the integrated weighted index of each grid cell in the study area, 13 and the final landslide susceptibility map was generated by using Weighted Overlay Analysis tool of ArcGIS. 14 5 Results and discussions 15

Landslide susceptibility mapping 16
The AHP method was used to assign the weights for each controlling factor. The higher the weight was, the more impacts on 17 landslide occurrence could be expected. As shown in Table 2, the weight of slope was highest, implying the most significant 18 influence of slope on the landslide occurrence, and the weights of aspect and NDVI were the lowest, which indicated that 19 these two factors played the least role in the landslide occurrence. 20 The FR values of each controlling factor category were calculated by using the Eq. (2) (as shown in Table 3). Table 3 clearly  21 shows the relationship between each controlling factor and the landslide occurrence. In the term of the relationship between 22 landslide occurrence and slope, landslides mostly occurred in the slope ranging from 40° to 60°. For the elevation, landslides 23 mostly occurred below the elevation of 3400 m, which implied that the probability of landslide occurrence was higher in 24 moderate steep mountainous region. In terms of the aspect, the FR value was very high for the class of E, N, SE and NE, and 25 it was lowest for the class of Flat. For the lithology, the highest FR value was achieved for Permian System which influenced 26 the landslide occurrence. For the factor of distance from faults, the highest FR value belonged to the area higher than 2000 m. 27 The distance from rivers with the highest FR value for frequent landslide occurrence was found usually between 0 and 600 m, 28 and landslides mostly occurred in the region with low vegetation cover of less NDVI value. In the case of PGA, the value of 29 0.26 g had the highest FR value, which indicated the significant influence of the earthquake on the landslide occurrence. In 1 general, our results were basically consistent with the previous study (Fan et al., 2018), which found that most of the 2 landslides mainly occurred in proximity of rivers and the epicentre, with an elevation of 2600 m to 3200 m and a slope of 35° 3 to 55°. 4 Finally, the landslide susceptibility map of study area was generated by using Weighted Overlay Analysis tool of ArcGIS, 5 and the study area was classified into seven categories of landslide susceptibility levels as presented in Fig. 5: very high,  6 high, relatively high, moderate, relatively low, low and very low by using Natural Breaks (Jenks) method with ArcGIS, 7 respectively. 8 According to the landslide susceptibility map, the location close to the epicentre and rivers was classified as the most 9 susceptible areas for landslides, and the high and very high landslide susceptible areas mostly located in the middle central 10 mountainous region. The low and very low susceptibility areas far from the epicentre and less affected by the earthquake, 11 mainly distributed in the North and South-West parts of the study area. Table 4 presented the distribution of seven landslide 12 susceptibility levels. As indicated in Table 4 To investigate the prediction performance of the integrated weighted index model, we also adopted five sub-datasets 1 containing 20 %, 40 %, 60 %, 80 % and 100 % of validation dataset (i.e., 169 landslides) respectively, to estimate the 2 prediction rates. Note that the validation dataset (i.e., 20 % of the landslide inventory dataset) was not used in the training 3 process. The AUC values of five sub-datasets, as presented in Fig. 6(b), were 78.71 %, 81.66 %, 84.27 %, 86.09 % and 4 87.16 %, respectively. With the increase of input data, the performance of the integrated weighted index model was 5 significantly improved, which indicated a reliable predicting capability of the integrated weighted index model adopted in 6 this study. 7 In addition, the landslide density distribution of each susceptibility level was computed by associating landslides with the 8 classified landslide susceptibility map (as shown in Table 4). There was a clear trend that the increase in the level of 9 landslide susceptibility was highly correlated with the density of landslides. The high and very high susceptibility levels had 10 the significant high landslide density values, while the low susceptibility categories were just the opposite, which also 11 implied the effectiveness of the generated landslide susceptibility map of the study area. 12 accurately identify the landslides before the Jiuzhaigou earthquake due to the limitations of historical images, and smaller 2 landslides were also not completely identified. Future work should focus on the preparation of more detailed landslide 3 inventories. Secondly, in this study, as the proposed method was applied to medium-scale datasets, the results may not be 4 suitable for specific analysis of large or detailed scale. At large or detailed scales, more detailed landslide inventory dataset 5 and controlling factor layers are required. Additionally, the assumption behind much of the landslide susceptibility mapping 6 is that future landslides will occur under similar environmental conditions as historical landslides (Guzzetti et al., 1999;7 Pourghasemi and Rahmati, 2018). However, results obtained in the past environmental conditions are not a guarantee for the 8 future (Guzzetti et al., 2005). In this study, we used a weighted index model by integrating the AHP and FR approaches to 9 map the earthquake-triggered landslides susceptibility and the generated susceptibility map of the study area was made for 10 the present situation. The susceptibility results need to be adapted as soon as environmental conditions or their causal 11 relationships obviously change in the future. Despite its limitations, the integrated method can generate a reliable landslide 12 susceptibility map at regional scale which can provide rapid assessment for reconstruction of tourism facilities, regional 13 disaster management etc. 14

Conclusions 15
Earthquake is one of the dynamic causes in landslide occurrence. Earthquake-triggered landslides can cause extensive and 16 significant damages to both lives and properties. In this study, given the main motivation to adopt an integrated weighted 17 index model based on FR and AHP methods for earthquake-triggered landslide susceptibility mapping at the Zhangzha town landslide susceptibility mapping at regional scale. 25 Even some limitations do exist, the integrated weighted index model can generate a reliable landslide susceptibility map at 26 regional scale that is useful for engineers and decision makers to understand the probability of landslides and mitigate 27 hazards. Furthermore, the integration of some machine learning techniques should be taken into account in the integrated 28 weighted index model for advancement in future studies.