Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Applied Rheology 2022; 32: 22–33

Research Article

Yu Liu*, Anying Yuan, Zhigang Bai, and Jingzhong Zhu

GIS-based landslide susceptibility mapping


using frequency ratio and index of entropy
models for She County of Anhui Province, China
https://doi.org/10.1515/arh-2022-0122 rivers plays a major role in landslide occurrence and dis-
received October 09, 2021; accepted April 11, 2022 tribution. The research results can be used for preliminary
Abstract: Landslides caused countless economic and casualty land use planning and hazard mitigation purposes.
losses in China, especially in mountainous and hilly areas. Keywords: landslide susceptibility, mountainous areas
Landslide susceptibility mapping is an important approach
and tool for landslide disaster prevention and control. This
study presents a landslide susceptibility assessment using fre-
quency ratio (FR) and index of entropy (IOE) models 1 Introduction
within a geographical information system for She County
in the mountainous region of South Anhui, China. First, Landslide disasters frequently occur in mountainous areas,
the landslide locations were ascertained in the study area causing secondary geological hazards, enormous casual-
using historical landslide records, aerial photographs, and ties, and losses of eco-environmental and infrastructure
multiple field surveys. In all, 502 landslides were identified [1]. Landslide disasters cause thousands of casualties and
and randomly divided into two groups as training (70%) several hundred billion dollars in direct economic losses
and validation (30%) datasets. Additionally, the landslide- worldwide [2]. In China, 6,000 landslide accidents occur
influencing factors, including slope angle, slope aspect, annually, resulting in an estimated 5–8 billion CNY in direct
curvature, landform, lithology, distance to faults, distance economic losses and hundreds of deaths, and most of the
to roads, distance to rivers, rainfall, and normalized dif- landslides accidents occur in mountainous and hilly areas
ference vegetation index, were selected and their relative [3]. She County is located in the mountainous region of
importance and weights were determined by FR and IOE South Anhui, China, where hilly and mountainous areas
models. The results show that the very high and high sus- account for 95%, which is the most serious landslide dis-
ceptibility classes cover nearly 50% of the study area. aster area in Anhui Province and even China. Until now, the
Finally, the comprehensive performance of the two models landslides have caused 68 casualties, 1,098 houses have
was validated and compared using receiver operating been destroyed, and 1018.4 million CNY of direct economic
characteristic curves. The results demonstrated that the loss [4,5]. Therefore, it is necessary to scientifically assess
IOE model with the area under the curve (AUC) of 0.802, the regional landslide susceptibility and help local govern-
which is slightly better in prediction than the FR model ment work out effective landslide prevention and control
(AUC = 0.786). The interpretation of the susceptibility map plans to reduce negative effects.
indicated that landform, slope degree, and distance to Currently, there exists a preliminary but reasonable
and applicable procedure for landslide susceptibility map-
ping based on the geographical information system (GIS).

The main core of landslide sensitivity assessment is to
* Corresponding author: Yu Liu, State Key Laboratory of Mining
Response and Disaster Prevention and Control in Deep Coal Mines, determine the influencing factors of regional landslides
Anhui University of Science and Technology, Huainan, Anhui, and appropriate models to produce the landslide suscept-
232001, China, e-mail: yliu@aust.edu.cn ibility maps [6]. Landslide is the external manifestation
Anying Yuan: State Key Laboratory of Mining Response and Disaster of the comprehensive action of primary geological condi-
Prevention and Control in Deep Coal Mines, Anhui University of
tions and external environmental factors. Generally, the
Science and Technology, Huainan, Anhui, 232001, China
Zhigang Bai, Jingzhong Zhu: School of Earth and Environment,
factors considered in the regional landslide susceptibility
Anhui University of Science and Technology, Huainan, Anhui, analysis mainly include lithology, elevation, slope angle,
232001, China slope aspect, human activities, water system, vegetation,

Open Access. © 2022 Yu Liu et al., published by De Gruyter. This work is licensed under the Creative Commons Attribution 4.0 International
License.
GIS-based landslide susceptibility mapping using FR and IOE models  23

rainfall, topographical structure, and geological structure. have been few studies comparing these methods, espe-
Due to the differences in regional conditions and scholars’ cially in the mountainous and hilly areas of China.
views, there is still no unified factors catalogue [4]. Lithology, The main purpose of this study is to form a landslide
slope angle, water system, and human activities are consid- susceptibility map of She County in Anhui Province,
ered basic factors by most scholars [7]. Zou et al. considered China, which is a landslide-prone area. For this purpose,
that 8–12 factors are sufficient to meet the requirements of 10 landslide-related factors are considered and overlaid by
landslide susceptibility evaluation, not the more the better using FR and IOE models based on the GIS. Additionally,
[8]. The evaluation models widely used in landslide sensi- the predictive performance of these models has been eval-
tivity mapping can be divided into qualitative and quantita- uated and compared using receiver operating characteristic
tive. As the most prevailing tools in landslide susceptibility (ROC) curves. The results of this study have implications for
modeling, quantitative approaches can be divided into landslide prevention and mitigation in the study area and
three groups: heuristic, deterministic, and statistical other similar terrains.
methods [9,10]. The statistical methods are the most widely
used, including statistical index (SI), weight of evidence
(WOE), certainty factor (CF), index of entropy (IOE), analy-
tical hierarchy process (AHP), logistic regression (LR), and 2 Study area
frequency ratio (FR) [11–16]. As the main technical sources
of data-driven modeling, machine-learning methods are She County is situated at longitudes 118°15′00ʺ to 118°53′
popularly applied to predict regional landslide suscept- 50ʺE and latitudes 29°30′25ʺ to 30°07′00ʺN in the moun-
ibility [7,17,18]. Some researchers have developed hybrid tainous region of South Anhui, China (Figure 1). The
models, combining statistical and machine-learning approaches, county encompasses approximately 2,122 km2, and hilly
to obtain more accurate results and overcome the short- and mountainous areas account for about 95% of the
comings of individual methods [19,20]. Each method has study area.
its inherent advantages and limitations; the prior step is to Topographically, the altitudes gradually decrease from
comprehensively understand the application of statistical the southeast to the northwest, in which the maximum and
methods [12]. It is beneficial for an in-depth understanding minimum are 1,787 and 100 m according to a DEM++
of the application and combination with each method. In with a grid of 30 m. Geologically, the tectonic structure
some of the literature, FR and IOE are most frequently used within the study area is complex, with abundant folds
to evaluate landslide susceptibility [21–24]. However, there and faults formed by multi-stage tectonic movements,

Figure 1: Location of the study area and landslide inventory map.


24  Yu Liu et al.

mainly including Indo-china Movement, Yanshanian of landslides, mainly including stratigraphic lithology, topo-
Movement, and Wannan Orogeny. The exposed strata graphy, and geological structure. The latter refers to the
include Sinian of Proterozoic, Cambrian of Paleozoic, factors that trigger the occurrence of landslides, such as
Ordovician, Jurassic and Cretaceous of Mesozoic, and the hydrogeological environment and human engineering
Quaternary loose soil and deposition. With regard to cli- activities. According to relevant research [28–30], the
mate, the study area belongs to the subtropical monsoon availability of data sources, and the characteristics of local
climate zone, which provides abundant rainfall in spring geological environments, 10 influencing factors including
and summer. It is reported that the annual average precipi- slope angle, slope aspect, curvature, landform, lithology,
tation is 1582.7 mm, and the annual average temperature is distance to faults, distance to roads, distance to rivers,
16.3°C. Xin’an, Lian, Fengle, Fengyuan, Changyuan, Jieyuan rainfall, and normalized difference vegetation index (NDVI),
Rivers, and their tributaries form a network drainage system. are considered for landslide susceptibility analysis. The digital
In 2020, the population of this county was approximately elevation model (DEM), which depicts the accurate represen-
362.9 thousand. The coupling effect of special geological and tation of the land surface, is suitable for medium-scale
climatic conditions and the influence of human engineering mapping [11]. Geomorphological-related thematic data
activities make this area prone to landslides. The coupling layers, including slope angle, slope aspect, and curvature,
effects of special geology, climate, and human engineering are extracted from the 30 m × 30 m DEM covering our
activities make this area prone to landslides. study area. Other parameters are mainly collected from
available resources, such as geological map, environment
geology map, road map, and drainage map. All of the
landslide-influencing factors are reclassified and expressed
3 Data used as corresponding thematic maps (Figure 3) with an identical
resolution of 30 m × 30 m.
The spatial characteristics of landslides in the study area Slope angle is an indispensable parameter in land-
were ascertained using 1:50,000 scale aerial photograph slide susceptibility evaluation, and its important influ-
interpretation, historical landslide records, and extensive ence on landslide occurrence has been widely discussed
field surveys and observations. Resultantly, a total of 502 [31]. Generally, the slope angle is firmly connected with
landslides (source areas) were detected in the study area, the stress distribution, groundwater, loose deposits, and
including 463 slides and 39 falls. In terms of size, the human engineering activities, thus affecting the failure
smallest landslide volume is about 1,880 m3, and the lar- modes and dynamic characteristics of a landslide [32,33].
gest is larger than 9 × 106 m3, and large-scale (106–107 m3), In this study, the slope angle was reclassified into five
medium-scale (105–106 m3), and small-scale (≤105 m3) categories: 0–10°, 10–18°, 18–26°, 26–35°, and 35–73°.
landslides (Figure 2) accounted for about 1, 8, and 91% The slope aspect determines the rainfall direction,
of the total, respectively. The detected landslides were solar radiation intensity, and the morphologic structure
randomly divided into two groups: 351 cases (70%) were of the area, which impacts the physical environment and
randomly selected for modeling and the remaining 151 vegetation around the slope [34]. For the landslide sus-
(30%) cases were used for validation [25]. ceptibility assessment, the slope aspect is classified into
The landslide is caused by the combined action of the nine categories: Flat, North, North-East, East, South-East,
internal elementary geological conditions and external South, South-West, West, and North-West.
environmental factors of the slope [26,27]. The former refers Plane curvature is defined as the curvature of contour
to the factors that control the occurrence and development formed by the intersection of horizontal plane and

Figure 2: Landslide photos: (a) small-scale landslide; (b) medium-scale landslide; and (c) medium-scale landslide (part).
GIS-based landslide susceptibility mapping using FR and IOE models  25

Figure 3: Landslide influencing factor maps: (a) slope angle; (b) slope aspect; (c) curvature; (d) landform; (e) distance to faults; (f) lithology;
(g) distance to roads; (h) distance to rivers; (i) rainfall, and (j) NDVI.
26  Yu Liu et al.

surface [35]. It affects the convergence and divergence of study, a total of four road buffer zones have been generated:
water flowing through the surface. Curvature is employed <400 m, 400–1,200 m, 1,200–2,000 m, and >2,000 m.
and arranged into three categories: <(–1), –1 to 1, and >1 River erosion is one of the important factors scouring
in the study. the slope toe and reducing slope stability [41]. The char-
The dependence of landslides on landform has been acteristics and mechanism of landslides triggered by river
demonstrated by relevant studies, and mechanisms behind erosion have been revealed through investigations and
the effects of various landform types, such as hilly and experiments [42,43]. Typically, it can be concluded that
mountain areas, on landslide activity have been discussed the probability of landslide occurrence decreases with
as well [36]. Landforms in She County include hills and increasing distance to rivers. In the present study, five
mountains. Mountainous and hilly areas account for 95% buffer zones are generated with an interval of 400 m,
of the county area, and plains account for 5% [5]. More including <400 m, 400–800 m, 800–1,200 m, 1,200–2,000 m,
specifically, the landforms are divided into eight cate- and >2,000 m.
gories: 1 (plain), 2 (shallow hill plain), 3 (middle hill), 4 Rainfall can accelerate the rate of slope erosion and is
(high hill), 5 (low undulating low mountain), 6 (high always considered one of the most crucial negative factors in
undulating low mountain), 7 (low undulating mountain), the stability of the landslide, especially in mountainous and
and 8 (high undulating mountain). hilly areas [14]. In this study, data from 18 meteorological
Faults usually break rock mass structure and reduce stations provided by the Anhui Provincial Meteorological
rock strength. In complicated structure areas, landslides Bureau (http://www.weather.org.cn) are used to produce
often occur along the faults and decrease sharply with the mean annual rainfall map. Thus, the rainfall is reclassi-
the distance to the faults [37]. The distance from faults, fied into four classes: <1,530 mm/year, 1,530–1,580 mm/year,
formed by multi-stage tectonic movements, is calculated 1,580–1,650 mm/year, and >1,650 mm/year.
at 400-m intervals using the geological map at a scale of Vegetation fixes the soil through roots and improves
50,000. Euclidean metric and visual inspection were the shear capacity of soil, which can effectively prevent
implemented to analyze the correlation between faults the occurrence of landslides [44]. NDVI is always regarded
and landslides. as a critical index reflecting vegetation characteristics in
Lithology, which can reflect rock physical and mechan- landslide susceptibility mapping [45]. In the current study,
ical properties including strength, weathering resistance, the NDVI map of She County is obtained from the opera-
etc., is generally considered to be the decisive interfering tional land image (OLI) of Landsat 8. The value is calcu-
factor for landslide stability [32,38]. In the study area, a total lated by the following formula:
of six categories (Table 1) are identified according to the
(IR − R)
geological ages and lithofacies using the bedrock geological NDVI = , (1)
(IR + R)
map with a 1:50,000 scale.
Generally, landslides tend to spread along road cuts where IR is the infrared band of the electromagnetic spec-
due to the changed slope conditions and artificial free trum, and R is the red band of the electromagnetic spectrum.
surfaces during road construction [39]. Many investigations Ultimately, the NDVI value is reclassified into five
and research results show that the probability of landslides categories including <0.2, 0.2–0.3, 0.3–0.35, 0.35–0.4,
decreases with the distance to roads [15,29,40]. In this and >0.4.

Table 1: Lithology classification

Categories Stratum and lithology

1 Mesoproterozoic. Gray purple and gray-green medium-thick lithic sandstone – laminated fine sandstone and siliceous
slate composition
2 Lower sinian. Composition of brown – yellow dolomite, gray – black streaked argillaceous slate, calcareous slate,
dolomitic, and carbonaceous siliceous slate
3 Middle sinian. Carbonated siliceous slate and striated siliceous slate in the gray-black middle layer
4 Cambrian. Gray black, black thin carbonaceous siliceous rocks intercalated with carbonaceous shale
5 Ordovician. Gray-green, light gray calcareous shale and nodular calcareous shale
6 Jurassic period. Gray – white quartz conglomerate with conglomerate lens, gray – white fine sandstone, lithic sandstone,
sandy mudstone, carbonaceous mudstone
GIS-based landslide susceptibility mapping using FR and IOE models  27

4 Landslide susceptibility mapping 4.2 Application of ratio index of entropy


model
4.1 Application of frequency ratio model
Another model used for evaluating landslide suscept-
The movement of rock-soil mass and controlling factors ibility is the index of the entropy model based on the
of landslide occurrence is assumed to be similar to those bivariate analysis principle. The entropy indicates the
observed in the past. Based on this assumption, future extent of instability, disorder, imbalance, and uncertainty
landslides occurring in an unspecified time span can be of a system [48]. The model can calculate the weight of
predicted [12]. The frequency ratio (FR) is defined as the various factors referring to the extent that each landslide
ratio between the percentage of landslides and the per- conditioning factor influences the occurrence of landslides
centage of pixels within a category. The FR is the ratio of in the natural environment [49]. The weight value for each
the area where landslides have occurred to the total area variable obtained separately is expressed as the entropy
and is also the ratio of the probability of the landslide index [10]. The equations used to calculate the weight
occurrence to nonoccurrence for a given attribute [46]. In value for each variable as a whole are as follows:
the present study, the FR for each factor’s type or range is b
Pij = , (4)
calculated based on its relationship with landslides. The a
larger the value, the stronger the correlation between the Pij
given factor and the landslide [47]. The landslide sus- (Pij ) = Sj
, (5)
∑ j = 1Pij
ceptibility index (LSI) has been acquired by summing
the ratios of each factor as the given equation: Sj
Hj = − ∑ (Pij ) log2(Pij ) , j = 1,…, n , (6)
LSI = ∑FR, (2) j=1

where FR is the ratio of each factor’s type or class and can Hj max = log2Sj, Sj is the number of classes, (7)
be expressed by the following equation:
Hj max − Hj
NLSpix
Ij = , I = (0, 1) , j = 1,…, n , (8)
Hj max
∑1n NLSpix
FR = , (3)
NC pix Wj = IjPijW , (9)
∑1m NC pix
where a and b are the percentages of domain and land-
where NLSpix is the number of pixels of landslides, and
slide, respectively, (Pij) represents the probability den-
NCpix is the number of pixels of a class.
sity, Hj and Hjmax are entropy values, Ij represents the
Based on the FR model, each class of the factors used
information coefficient, and Wj is the resultant weight
in this study is characterized by a certain landslide occur-
value for the factor as a whole. The LSM is finally pre-
rence density (Table 2).
pared by the following equation:
As a result of the FR model, an LSI map (Figure 4) has
n
been constructed by ArcGIS 10.0 software, in which the z
YIOE = ∑ × C × Wj, (10)
LSI values vary from 2.88 to 8.01. Obviously, larger LSI m
i=1 i
values indicate a higher susceptibility for land sliding.
where YIOE is the sum of all classes, i is the number of
For the visual interpretation of LSI maps, considering
specific parameter mappings, z is the number of classes
data distribution histogram, the quantile and natural
with the largest number of classes in a parameter map-
break methods are applied to classify the data in this
ping, mi is the number of classes in a specific parameter
study. The comparison results show that the quantile
map, C represents the value of the class after reclassifica-
data classification method is able to produce better clas-
tion, and Wj represents the weight of the parameter. The
sification results. After which the LSI map is classified
summation results reflect the different sensitivities of
into four sensitivity classes (low, moderate, high, and
landslide susceptibility.
very high) using the quantile classifier. The corresponding
The weight Wj of each conditioning factor in land-
area proportions for low, moderate, high, and very high
slides is calculated and presented in Table 2. The results
susceptibility classes are 25.67, 25.33, 22.60, and 25.50%,
show that each class of the selected factors has a certain
respectively.
28  Yu Liu et al.

Table 2: Spatial relationship between conditioning factor and landslide by FR and IOE models

Conditioning factor Class % Total of area (a) % Total of landslide area (b) FR (b/a) (Pij) Hj Hjmax Ij Wj

Slope angle (°) 0–10 8.55 6.77 0.79 0.21 1.98 2.32 0.15 0.55
10–18 24.79 25.50 1.03 0.28
18–26 44.26 50.80 1.15 0.31
26–35 22.15 16.93 0.76 0.20
35–73 0.25 0 0 0.00
Slope aspect F 0.31 0 0 0.00 2.99 3.17 0.06 0.46
N 10.97 10.96 1.00 0.12
NE 11.17 8.96 0.80 0.10
E 12.48 10.76 0.86 0.11
SE 13.47 11.75 0.87 0.11
S 12.38 14.74 1.19 0.15
SW 12.81 14.74 1.15 0.14
W 13.20 12.95 0.98 0.12
NW 13.20 15.14 1.15 0.14
Curvature Concave 43.99 43.03 0.98 0.34 1.58 1.59 0.00 0.01
Flat 9.34 7.97 0.85 0.30
Convex 46.67 49.00 1.05 0.36
Landform 1 2.28 0.20 0.09 0.02 2.45 3.00 0.18 1.07
2 2.39 0 0 0.00
3 4.25 5.78 1.36 0.23
4 7.85 1.00 0.13 0.02
5 15.23 15.34 1.01 0.17
6 23.75 32.27 1.36 0.23
7 28.01 35.86 1.28 0.22
8 16.25 9.56 0.59 0.10
Distance to faults (m) <400 10.41 21.51 2.07 0.32 2.21 2.32 0.05 0.32
400–800 10.26 15.34 1.50 0.23
800–1,200 10.01 9.56 0.96 0.15
1,200–2,000 18.62 24.90 1.34 0.21
>2,000 50.71 28.69 0.57 0.09
Lithology 1 54.47 70.32 1.29 0.25 2.53 2.59 0.02 0.11
2 25.40 14.94 0.59 0.12
3 1.05 1.00 0.95 0.19
4 3.65 3.39 0.93 0.18
5 5.44 3.78 0.70 0.14
6 9.98 6.57 0.66 0.13
Distance of roads (m) <400 18.63 30.88 1.66 0.39 1.93 2.00 0.04 0.15
400–1,200 27.62 25.30 0.92 0.22
1,200–2,000 20.79 17.13 0.82 0.20
>2,000 32.96 26.69 0.81 0.19
Distance of Rivers (m) <400 22.69 39.84 1.76 0.34 2.10 2.32 0.09 0.49
400–800 19.67 18.53 0.94 0.18
800–1,200 17.24 11.35 0.66 0.13
1,200–2,000 14.41 22.71 1.58 0.30
>2,000 25.99 7.57 0.29 0.06
Rainfall (mm/year) <1,530 50.69 65.74 1.30 0.39 1.93 2.00 0.03 0.12
1,530–1,580 29.41 21.12 0.72 0.22
1,580–1,650 7.86 4.98 0.63 0.19
>1,650 12.03 8.17 0.68 0.20
NDVI <0.2 2.89 6.37 2.20 0.34 2.20 2.32 0.05 0.35
0.2–0.3 9.78 14.14 1.45 0.23
0.3–0.35 24.41 29.68 1.22 0.19
0.35–0.4 35.52 33.47 0.94 0.15
>0.4 27.40 16.33 0.60 0.09
GIS-based landslide susceptibility mapping using FR and IOE models  29

Figure 4: Landslide susceptibility map derived by the FR model. Figure 5: Landslide susceptibility map derived by the IOE model.

landslide occurrence density, and the landform, slope 20.59. Obviously, the larger LSI value indicates a higher
angle, distance of rivers, and NDVI are the most impor- susceptibility to land sliding. After which the LSI map is
tant factors that influence landslide distribution. Before classified into four sensitivity classes (low, moderate,
LSM development, LSI is calculated by using the fol- high, and very high) by using the quantile classifier.
lowing equation: The corresponding area proportions for low, moderate,
high, and very high susceptibility classes are 4.18, 22.71,
⎛ (Slope angle × 0.55) + (Slope aspect ⎞ 41.63, and 31.47%, respectively.
⎜ × 0.46) + (Curvature × 0.01) ⎟
⎜ ⎟
⎜ + (Landform × 1.07) + (Distance to faults ⎟
YIOE = ⎜ × 0.32) + (Lithology × 0.11) ⎟. (11)
⎜ ⎟ 5 Validation of the landslide
+ (Distance to roads × 0.15)
⎜ ⎟
⎜ + (Distance to rivers × 0.49) ⎟
susceptibility maps
⎜ ⎟
⎝ + ( Rainfall × 0.12) + (NDVI × 0.35) . ⎠ Without validation, any try or effort to ascertain the land-
As a result of the IOE model, an LSI map (Figure 5) is slide sensitivity in an area is meaningless [12]. In this
constructed, in which the LSI value varies from 4.14 to study, the raster distributions (RD) histogram (Figure 6)

Figure 6: Raster distributions of landslides within different classes. (a) FR, and (b) IOE.
30  Yu Liu et al.

Figure 7: Success and prediction rate curves for the LSMs produced in this study. (a) Success; (b) prediction.

and the ROC curve (Figure 7) have been drawn to validate The AUC plot assessment results of prediction rate curves
and compare the performance of the two models. (Figure 7(b)) found that the AUC values for FR and IOE
In Figure 6, the blue and green bars represent the models are 0.786 and 0.802, respectively. It can be con-
proportions of the landslide-prone areas in the total cluded that both the success rate and the prediction rate
area and known landslide points in each classed area, curve show a similar result, and the models used in this
respectively. Generally, a higher green bar in the higher study exhibited reasonably good accuracy in landslide
class with a lower blue bar in the corresponding class susceptibility prediction. The IOE model is more suitable
indicates more accuracy and better fitting of the model for landslide susceptibility mapping in the study area.
is validated, which can provide a scientific basis for the In this paper, FR and IOE models were used to develop
prediction and management of landslide hazards [2]. the landslide vulnerability map of Shexian County. It can
Here, the proportions of landslide points in the high class provide a comprehensive and inexpensive assessment of
are divided by the corresponding proportions of area for the study area and its ability to support individual or com-
two models, the results are 1.49 (FR) and 1.52 (IOE), prehensive uses, such as road construction and logging.
respectively. It indicates that both models can be used Managers and foresters can then make decisions and
for assessing landslide susceptibility, while the IOE model develop prescriptions that will produce highly predictable
is more accurate and better fitting in our study area. results to produce sustainable products, maintain site
The ROC curve is also adopted to check the quality of quality, and significantly reduce the risk of any adverse
deterministic and probabilistic detection and forecast effects.
systems. In this study, 351 landslide locations (70%) are
randomly selected from the observed dataset as the training
data, and the remaining 151 landslide locations (30%) are
used for validation. Then, the success rate curves were 6 Conclusion
plotted by Python using training data (Figure 7(a)). In
Figure 7(a), the x-axis is 1-specificity indicating the prob- In this study, two widely accepted models – frequency
ability of predicted disaster points, and the y-axis is suscept- ratio (FR) and index of entropy (IOE) – are used to pro-
ibility, which represents the probability of disaster points duce landslide susceptibility mappings for She County in
correctly predicted in the total area. The AUC is used to the mountainous region of South Anhui, China. A total of
qualitatively analyze the prediction accuracy of the land- 502 landslide locations has been identified and randomly
slide susceptibility map. When the AUC value close to 1.0 divided into two groups as training (70%) and validation
indicates fitting better, the value below 0.5 indicates fitting (30%) datasets. To perform sensitivity mapping, 10 land-
randomly [50]. The analysis results of the success rate curve slide-influencing factors, including slope angle, slope
indicated that the IOE model has a higher AUC value aspect, curvature, landform, lithology, distance to faults,
(0.816), whereas the FR model has 0.808. Because the distance to roads, distance to rivers, rainfall, and normal-
success rate method used the data in the previous mod- ized difference vegetation index (NDVI), are selected.
eling, its results are less meaning for the prediction cap- Four susceptibility classes – low, moderate, high, and
ability evaluation of the models. Therefore, the prediction very high – are derived with the quantile data classifica-
rate is adopted to measure the prediction performance. tion method in the susceptibility maps produced by FR
GIS-based landslide susceptibility mapping using FR and IOE models  31

and IOE models. The success rate and prediction rate Reference
methods are adopted to validate and compare the perfor-
mance of the two models. The results indicate that the [1] Chen W, Chen X, Peng JB, Panahi M, Lee S. Landslide sus-
IOE model has a success rate of 0.816 and a predictive ceptibility modeling based on ANFIS with teaching-learning-
based optimization and Satin bowerbird optimizer. Geosci
accuracy of 0.802, which is higher than that of the FR
Front. 2021;12(1):93–107.
model, with AUC values of 0.808 and 0.786, respectively.
[2] Sharma S, Mahajan AK. A comparative assessment of infor-
The validation process also indicates that the models uti- mation value, frequency ratio and analytical hierarchy process
lized in this study exhibit reasonably good accuracy in models for landslide susceptibility mapping of a Himalayan
predicting landslide susceptibility. The interpretation of watershed, India. Bull Eng Geol Env. 2018;78(4):2431–48.
the susceptibility map indicates that landform, slope [3] Zhou S, Zhang Y, Tan X, Abbas SM. A comparative study of the
bivariate, multivariate and machine-learning-based statistical
degree, and distance to rivers play a major role in land-
models for landslide susceptibility mapping in a seismic-
slide occurrence and distribution. The research results prone region in China. Arab J Geosci. 2021;14:440.
are worthy of preliminary land use planning and hazard [4] Froude MJ, Petley DN. Global fatal landslide occurrence from
mitigation in the next future. 2004 to 2016. Nat Hazard Earth Sys. 2018;18(8):2161–81.
[5] Wang ZT, Liu QM, Liu Y. Mapping landslide susceptibility using
Acknowledgments: The authors are grateful for the sup- machine learning algorithms and GIS: A Case Study in Shexian
County, Anhui Province, China. Symmetry. 2020;12:1954.
port of the Natural Science Foundation of Anhui Province
[6] Pourghasemi HR, Rahmati O. Prediction of the landslide sus-
(Nos. 2008085QD191 and 1908085ME144), the National ceptibility: which algorithm, which precision? Catena.
Natural Science Foundation of China (No. 52104073), 2018;162:177–92.
the Independent Research Fund of the State Key Laboratory [7] Pourghasemi HR, Teimoori YZ, Panagos P, Pradhan B. Analysis
of Mining Response and Disaster Prevention and Control and evaluation of landslide susceptibility: a review on articles
published during 2005–2016 (periods of 2005–2012 and
in Deep Coal Mines (Anhui University of Science and
2013–2016). Arab J Geosci. 2018;11:193.
Technology) (No. SKLMRDPC19ZZ06), and the University- [8] Zou Z, Xiong C, Tang H, Criss RE, Su A, Liu X. Prediction of
level Key Projects of Anhui University of Science and landslide runout based on influencing factor analysis. Env
Technology (No. QN2019110). Wei Chen, Yanli Wu, and Earth Sci. 2017;76:723.
Liyong Cai are thanked for their assistance in data collec- [9] Pradhan B, Seeni MI, Kalantar B. Performance evaluation and
sensitivity analysis of expert-based, statistical, machine
tion and analysis.
learning, and hybrid models for producing landslide suscept-
ibility maps. Springer Int Publ. 2017;193–232.
Funding information: The Natural Science Foundation of [10] Chen W, Fan L, Li C, Pham BT. Spatial prediction of landslides
Anhui Province (Nos. 2008085QD191 and 1908085ME144), using hybrid integration of artificial intelligence algorithms
the National Natural Science Foundation of China (No. with frequency ratio and index of entropy in Nanzheng County,
52104073), the Independent Research Fund of the State China. Appl Sci. 2019;10(1):29.
[11] Reichenbach P, Rossi M, Malamud BD, Mihir M, Guzzetti F.
Key Laboratory of Mining Response and Disaster Prevention
A review of statistically-based landslide susceptibility models.
and Control in Deep Coal Mines (Anhui University of Science Earth-Sci Rev. 2018;180:60–91.
and Technology) (No. SKLMRDPC19ZZ06), and the University- [12] Jaafari A, Najafi A, Pourghasemi HR, Rezaeian J, Sattarian A.
level Key Projects of Anhui University of Science and GIS-based frequency ratio and index of entropy models for
Technology (No. QN2019110). landslide susceptibility assessment in the Caspian forest,
northern Iran. Int J Env Sci Technol. 2014;11:909–26.
[13] Soyoung P, Jinsoo K. A comparative assessment of the efficacy
Author contributions: Yu Liu – investigation and writing-
of frequency ratio, statistical index, weight of evidence, cer-
original draft; Anying Yuan – writing-review and editing; tainty factor, and index of entropy in landslide susceptibility
Zhigang Bai – methodology; Jingzhong Zhu – data curation. mapping. Korean J Remote Sens. 2020;36(1):67–81.
[14] Wubalem A, Meten M. Landslide susceptibility mapping using
Conflict of interest: Authors state no conflict of interest. information value and logistic regression models in Goncha
Siso Eneses area, northwestern Ethiopia. SN Appl Sci.
2020;2:807.
Ethical approval: The conducted research is not related to [15] Nsengiyumva JB, Luo G, Amanambu AC, Mind’je R, Habiyaremye G,
either human or animal use. Karamage F, et al. Comparing probabilistic and statistical methods
in landslide susceptibility modeling in Rwanda/Centre-Eastern
Data availability statement: All data generated or ana- Africa. Sci Total Env. 2019;659(1):1457–72.
lysed during this study are included in this published [16] Aghdam IN, Varzandeh MHM, Pradhan B. Landslide suscept-
ibility mapping using an ensemble statistical index (Wi) and
article.
32  Yu Liu et al.

adaptive neuro-fuzzy inference system (ANFIS) model at based Naïve Bayes, RBF Classifier, and RBF Network machine
Alborz Mountains (Iran). Env Earth Sci. 2016;75(7):553. learning algorithms. Sci Total Env. 2019;663:1–15.
[17] Pham BT, Shirzadi A, Shahabi H, Omidvar E, Singh SK, [32] Dao DV, Jaafari A, Bayat M, Mafi-Gholami D, Qi C, Moayedi H,
Sahana M, et al. Landslide susceptibility assessment by novel et al. A spatially explicit deep learning neural network model
hybrid machine learning algorithms. Sustainability. for the prediction of landslide susceptibility. Catena.
2019;11(16):4386. 2020;188:104451.
[18] Arabameri A, Pal SC, Rezaie F, Chakrabortty R, Saha A, [33] Zhao Z, Liu ZY, Xu C. Slope unit-based landslide susceptibility
Blaschke T, et al. Decision tree based ensemble machine mapping using certainty factor, support vector machine,
learning approaches for landslide susceptibility mapping. random forest, cf-svm and cf-rf models. Front Earth Sci.
Geocarto Int. 2021;3:1–28. 2021;9:589630.
[19] Arabameri A, Pradhan B, Rezaei K, Sohrabi M, Kalantari Z. GIS- [34] Xiao L, Zhang Y, Peng G. Landslide susceptibility assessment
based landslide susceptibility mapping using numerical risk using integrated deep learning algorithm along the China-
factor bivariate model and its ensemble with linear multi- Nepal highway. Sensors. 2018;18(12):4436.
variate regression and boosted regression tree algorithms. [35] Sun X, Chen J, Han X, Bao Y, Zhou X, Peng W. Landslide sus-
J Mt Sci. 2019;16(3):595–618. ceptibility mapping along the upper Jinsha River, south-wes-
[20] Steger S, Brenning A, Bell R, Glade T. The influence of sys- tern China: a comparison of hydrological and curvature
tematically incomplete shallow landslide inventories on sta- watershed methods for slope unit classification. Bull Eng Geol
tistical susceptibility models and suggestions for improve- Env. 2020;79:4657–70.
ments. Landslides. 2017;14(5):1767–81. [36] Mind’je R, Li L, Nsengiyumva JB, Mupenzi C, Nyesheja EM,
[21] Umar Z, Pradhan B, Ahmad A, Jebur MN, Tehrany MS. Kayumba PM, et al. Landslide susceptibility and influencing
Earthquake induced landslide susceptibility mapping using an factors analysis in Rwanda. Env Dev Sustain.
integrated ensemble frequency ratio and logistic regression 2019;22:7985–8012.
models in West Sumatera Province, Indonesia. Catena. [37] Erener A, Mutlu A, Sebnem, Düzgün H. A comparative study
2014;118:124–35. for landslide susceptibility mapping using GIS-based multi-
[22] Kursah MB, Wang Y, Bayoh HD, Tarawally M. A comparative criteria decision analysis (MCDA), logistic regression (LR)
study on the predictive ability of archived and SBAS-InSAR and association rule mining (ARM). Eng Geol. 2016;203:45–55.
inventories for landslide susceptibility using frequency ratio [38] Nhu VH, Mohammadi A, Shahabi H, Ahmad BB, Al-Ansari N,
model in Western Area, Sierra Leone. Env Earth Sci. Shirzadi A, et al. Landslide detection and susceptibility mod-
2021;80:387. eling on cameron highlands (Malaysia): a comparison between
[23] Arca D, Citiroglu HK, Tasoglu IK. A comparison of GIS-based random forest, logistic regression and logistic model tree
landslide susceptibility assessment of the Satuk village algorithms. Forests. 2020;11(8):830.
(Yenice, NW Turkey) by frequency ratio and multi-criteria [39] Acharya TD, Lee DH. Landslide susceptibility mapping using
decision methods. Env Earth Sci. 2019;78:81. relative frequency and predictor rate along Araniko highway.
[24] Zhang TY, Zhang L, Zhao H, Zhao X. GIS-based landslide sus- KSCE J Civ Eng. 2019;23(2):763–76.
ceptibility mapping using hybrid integration approaches of [40] Paryani S, Neshat A, Javadi S, Pradhan B. Comparative per-
fractal dimension with index of entropy and support vector formance of new hybrid ANFIS models in landslide suscept-
machine. J Mt Sci. 2019;16(6):72–90. ibility mapping. Nat Hazards. 2020;103:1961–88.
[25] Youssef AM, Pourghasemi HR. Landslide susceptibility map- [41] Juliev M, Mergili M, Mondal I, Nurtaev B, Pulatov A, Hübl J.
ping using machine learning algorithms and comparison of Comparative analysis of statistical methods for landslide
their performance at Abha Basin, Asir Region, Saudi Arabia. susceptibility mapping in the Bostanlik District, Uzbekistan.
Geosci Front. 2021;12:639–55. Sci Total Env. 2019;653:801–14.
[26] He K, Ma G, Hu X, Liu B. Failure mechanism and stability [42] Raja NB, Çiçek I, Türkolu N, Aydin O, Kawasaki A. Landslide
analysis of a reactivated landslide occurrence in Yanyuan City, susceptibility mapping of the Sera River Basin using logistic
China. Landslides. 2021;18:1097–114. regression model. Nat Hazards. 2017;85(3):1323–46.
[27] Nadim F, Kjekstad O, Peduzzi P, Herold C, Jaedicke C. Global [43] Weidner L, DePrekel K, Oommen T, Vitton S. Investigating
landslide and avalanche hotspots. Landslides. large landslides along a river valley using combined physical,
2006;3(2):159–73. statistical, and hydrologic modeling. Eng Geol.
[28] Abuzied SM, Alrefaee HA. Spatial prediction of landslide-sus- 2019;259:105169.
ceptible zones in El-Qaá area, Egypt, using an integrated [44] Zhang S, Li Z, Hou X, Yi Y. Impacts on watershed-scale runoff
approach based on GIS statistical analysis. Bull Eng Geol Env. and sediment yield resulting from synergetic changes in cli-
2019;78(4):2169–95. mate and vegetation. Catena. 2019;179:129–38.
[29] Arabameri A, Pradhan B, Rezaei K, Lee CW. Assessment of [45] Gigovic L, Drobnjak S, Pamuar D. The application of the hybrid
landslide susceptibility using statistical- and artificial intelli- gis spatial multi-criteria decision analysis best-worst metho-
gence-based FR–RF integrated model and multiresolution dology for landslide susceptibility mapping. Int J Geo-Inf.
DEMs. Rem Sens. 2019;11(9):999. 2019;8(2):1–25.
[30] Van Westen CJ, Castellanos E, Kuriakose SL. Spatial data for [46] Liu H, Li X, Meng T, Liu Y. Susceptibility mapping of damming
landslide susceptibility, hazard, and vulnerability assess- landslide based on slope unit using frequency ratio model.
ment: An overview. Eng Geol. 2008;102(3–4):112–31. Arab J Geosci. 2020;13:790.
[31] He Q, Shahabi H, Shirzadi A, Li S, Chen W, Wang N, et al. [47] Shahabi H, Khezri S, Ahmad BB, Hashim M. Landslide sus-
Landslide spatial modelling using novel bivariate statistical ceptibility mapping at central Zab basin, Iran: A comparison
GIS-based landslide susceptibility mapping using FR and IOE models  33

between analytical hierarchy process, frequency ratio and value and index of entropy model in Bhanupali-Beri
logistic regression models. Catena. 2014;115:55–70. region, Himachal Pradesh, India. Env Dev Sustain.
[48] Wang QQ, Li WP, Wu YL, Pei YB, Xie P. Application of statistical index 2021;23:5233–50.
and index of entropy methods to landslide susceptibility assess- [50] Xiao T, Segoni S, Chen L, Yin K, Casagli N. A step beyond
ment in gongliu (Xinjiang, China). Env Earth Sci. 2016;75:599. landslide susceptibility maps: A simple method to investigate
[49] Singh P, Sharma A, Sur U, Rai PK. Comparative landslide and explain the different outcomes obtained by different
susceptibility assessment using statistical information approaches. Landslides. 2020;17(3):627–40.

You might also like