A Comparative Analysis of Binary Logistic Regression and Analytical Hierarchy Process for Landslide Susceptibility Assessment in the Dobrov River Basin, Romania
A correct assessment of the landslide susceptibility component is extremely useful for the diminution of associated potential risks to local economic development, particularly in regard to land use planning and soil conservation. The purpose of the present study was to compare the usefulness of two methods, i.e., binary logistic regression (BLR) and analytical hierarchy process (AHP), for the assessment of landslide susceptibility over a 130-km2 area in the Moldavian Plateau (eastern Romania) region, where landslides affect large areas and render them unsuitable for agriculture. A large scale inventory mapping of all types of landslides (covering 13.7% of the total area) was performed using orthophoto images, topographical maps, and field surveys. A geographic information system database was created, comprising the nine potential factors considered as most relevant for the landsliding process. Five factors (altitude, slope angle, slope aspect, surface lithology, and land use) were further selected for analysis through the application of a tolerance test and the stepwise filtering procedure of BLR. For each predictor, a corresponding raster layer was built and a dense grid of equally spaced points was generated, with an approximately equal number of points inside and outside the landslide area, in order to extract the values of the predictors from raster layers. Approximately half of the total number of points was used for model computation, while the other half was used for validation. Analytical hierarchy process was employed to derive factor weights, with several pair-wise comparison matrices being tested for this purpose. The class weights, on a scale of 0 to 1, were taken as normalized landslide densities. A comparison of results achieved through these two approaches showed that BLR was better suited for mapping landslide susceptibility, with 82.8% of the landslide area falling into the high and very high susceptibility classes. The susceptibility class separation using standard deviation was superior to either the equal interval or the natural break method. Results from the study area suggest that the statistical model achieved by BLR could be successfully extrapolated to the entire area of the Moldavian Plateau.
A comparison of landslide susceptibility maps produced by logistic regression, multi-criteria decision, and likelihood ratio methods: a case study at İzmir, Turkey
Landslides
(2012)
P Aleotti et al.
Landslide hazard assessment: summary review and new perspectives
Bull Eng Geol Env
(1999)
I Armas
An analytic multicriteria hierarchical approach to assess landslide vulnerability. Case study: Cornu village, Subcarpathian Prahova Valley/Romania
Z Geomorphol
(2011)
I Armas
Weights of evidence method for landslide susceptibility mapping. Prahova Subcarpathians, Romania
Nat Hazards
(2012)
S B Bai et al.
GIS-based susceptibility mapping with comparisons of results from analytical hierarchy process versus logistic regression in the Three Gorges Area, China
S B Bai et al.
GIS-based rare events logistic regression for landslide-susceptibility mapping of Lianyungang, China
Environ Earth Sci
(2011)
G D Bathrellos et al.
Potential suitability for urban planning and industry development using natural hazard maps and geological-geomorphological parameters
Environ Earth Sci
(2012)
G D Bathrellos et al.
Assessment of rural community and agricultural development using geomorphological-geological factors and GIS in the Trikala prefecture (Central Greece)
Stoch Environ Res Risk A
(2013)
G D Bathrellos et al.
GIS-based landslide susceptibility mapping models applied to natural and urban planning in Trikala, Central Greece
Estud Geol
(2009)
D A Belsley et al.
Regression Diagnostics, Identifying Influential Data and Sources of Collinearity
(1980)
J Chacón et al.
Engineering geology maps: landslides and geographical information systems
Bull Eng Geol Environ
(2006)
C Chalkias et al.
GIS-based landslide susceptibility mapping on the Peloponnese Peninsula, Greece
Geosciences
(2014)
P K Champati Ray et al.
Fuzzy-based method for landslide hazard assessment in active seismic zone of Himalaya
Landslides
(2007)
S Chauhan et al.
Landslide susceptibility zonation of the Chamoli region, Garhwal Himalayas, using logistic regression model
Landslides
(2010)
T Chen et al.
Regional soil erosion risk mapping using RUSLE, GIS, and remote sensing: a case study in Miyun Watershed, North China
Environ Earth Sci
(2011)
M Constantin et al.
Landslide susceptibility assessment using the bivariate statistical analysis and the index of entropy in the Sibiciu Basin (Romania)
Environ Earth Sci
(2011)
D R Cox et al.
Analysis of Binary Data
(1989)
F C Dai et al.
Assessment of landslide susceptibility on the natural terrain of Lantau Island, Hong Kong
Env Geol
(2001)
S Dragićević et al.
Landslide susceptibility zonation in the Kolubara river basin (western Serbia) – analysis of input data
The choice of partition methods determines the correctness of LSMs, and multiple relevant approaches such as Jenks natural breaks, standard deviation, equal intervals and quantile have been applied (Baeza et al., 2016; Patriche et al., 2016). Among these approaches, Jenks natural breaks method deserves to be considered as the most prevailing (Aditian et al., 2018; Baeza et al., 2016; Patriche et al., 2016). Using Jenks natural breaks method, several classes can be formed based on the intrinsic features of dataset without any subjective thought.
Landslides, one of the most common hazards around the world, have brought about severe damage to life and property of human. To prevent and mitigate landslides, various models have been introduced to assess landslide susceptibility. In this paper, Hoeffding Tree (HT), a prevailing data stream mining algorithm, was employed to predict landslide susceptibility in Muchuan County, China for the first time. Meanwhile, Logistic Model Tree (LMT) and Bayes Network (BN) were applied to produce landslide susceptibility maps for comparison. The model performances were evaluated by Receiver Operating Characteristic (ROC) curves and areas under the curves (AUC). To obtain landslide inventory map, 279 landslides data was collected, and training and validation datasets were randomly divided with a proportion of 70% to 30%. Furthermore, twelve conditioning factors (altitude, slope angle, profile curvature, plan curvature, slope aspect, distance to roads, distance to rivers, TWI, NDVI, soil, land use and lithology) were selected to construct landslide susceptibility models. Moreover, correlations between conditioning factors and landslides were analyzed using Frequency Ratio (FR). The results showed landslides are prone to occur in areas where human activities concentrate, and all three models exhibited satisfying performances. Concretely, for training dataset, LMT model showed the highest AUC (0.854), followed by HT (0.726) and BN (0.709). However, for validation dataset, LMT and BN models generated similar AUC values (0.761 and 0.764 respectively), and the highest AUC value belonged to HT (0.802). The distributions of landslide susceptibility zones revealed that the interior of county town is mainly seated in low and very low susceptibility zones, whereas regions close to the border suffer high and very high landslide risk. The results acquired in this paper are significant to landslide prevention and urban planning in Muchuan, China. Additionally, this study proved that HT model is a promising classifier for landslide susceptibility modeling.
Multicollinearity exists when significant correlations are found among the predictors. Their presence may negatively affect the interpretation of regression coefficients (Patriche et al., 2016; Lin et al., 2017). In order to avoid multicollinearity, a tolerance test can be applied in terms of the coefficient of determination R2, where the values are computed as 1 − R2; the values less than 0.2 indicate multicollinearity (Bai et al., 2010).
In this paper, we present a suitable integration of discrete and continuous data in a unique methodology based on systematically collected landslide inventory data. Eleven landslide conditioning factors were analyzed and used, where eight correspond to DEM–derived variables, and three to thematic polygon–type variables (shallow geology, geomorphology and soil land–use). Principal Component Analysis (PCA) was used to avoid the effect of multicollinearity. Additionally, 3 proposals were developed using Logistic Regression (LR) and Weights of Evidence (WoE) methods that use the continuous and discrete variables efficiently, respectively. The performance of each proposal was evaluated by the Area Under the Curve (AUC) of the Receiver Operating Characteristic (ROC) curves. The analysis indicated that Proposal 1 with AUC = 0.8578 and Proposal 2 with AUC = 0.8459 have the best LSI assessment, while the performance of Proposal 3 with AUC = 0.8054 shows the lowest prediction approaches. In comparison with the WoE method, our proposal shows an increase in high and very high susceptibility in areas with complex topography, which is consistent with the reported landslides.
