Spatial Prediction of Landslide Susceptibility Using Random Forest Algorithm

To predict the potential geographic distribution of Oncomelania hupensis in Yunnan Province using random forest (RF) and maximum entropy (MaxEnt) models, so as to provide insights into O. hupensis surveillance and control in Yunnan Province. The O. hupensis snail survey data in Yunnan Province from 2015 to 2016 were collected and converted into O. hupensis snail distribution site data. Data of 22 environmental variables in Yunnan Province were collected, including twelve climate variables (annual potential evapotranspiration, annual mean ground surface temperature, annual precipitation, annual mean air pressure, annual mean relative humidity, annual sunshine duration, annual mean air temperature, annual mean wind speed, ≥ 0 ℃ annual accumulated temperature, ≥ 10 ℃ annual accumulated temperature, aridity and index of moisture), eight geographical variables (normalized difference vegetation index, landform type, land use type, altitude, soil type, soil textureclay content, soil texture-sand content and soil texture-silt content) and two population and economic variables (gross domestic product and population). Variables were screened with Pearson correlation test and variance inflation factor (VIF) test. The RF and MaxEnt models and the ensemble model were created using the biomod2 package of the software R 4.2.1, and the potential distribution of O. hupensis snails after 2016 was predicted in Yunnan Province. The predictive effects of models were evaluated through cross-validation and independent tests, and the area under the receiver operating characteristic curve (AUC), true skill statistics (TSS) and Kappa statistics were used for model evaluation. In addition, the importance of environmental variables was analyzed, the contribution of environmental variables output by the models with AUC values of > 0.950 and TSS values of > 0.850 were selected for normalization processing, and the importance percentage of environmental variables was obtained to analyze the importance of environmental variables. Data of 148 O. hupensis snail distribution sites and 15 environmental variables were included in training sets of RF and MaxEnt models, and both RF and MaxEnt models had high predictive performance, with both mean AUC values of > 0.900 and all mean TSS values and Kappa values of > 0.800, and significant differences in the AUC (t = 19.862, P < 0.05), TSS (t = 10.140, P < 0.05) and Kappa values (t = 10.237, P < 0.05) between two models. The AUC, TSS and Kappa values of the ensemble model were 0.996, 0.954 and 0.920, respectively. Independent data verification showed that the AUC, TSS and Kappa values of the RF model and the ensemble model were all 1, which still showed high performance in unknown data modeling, and the MaxEnt model showed poor performance, with TSS and Kappa values of 0 for 24%(24/100) of the modeling results. The modeling results of 79 RF models, 38 MaxEnt models and their ensemble models with AUC values of > 0.950 and TSS values of > 0.850 were included in the evaluation of importance of environmental variables. The importance of annual sunshine duration (SSD) was 32.989%, 37.847% and 46.315% in the RF model, the MaxEnt model and their ensemble model, while the importance of annual mean relative humidity (RHU) was 30.947%, 15.921% and 28.121%, respectively. Important environment variables were concentrated in modeling results of the RF model, dispersed in modeling results of the MaxEnt model, and most concentrated in modeling results of the ensemble model. The potential distribution of O. hupensis snails after 2016 was predicted to be relatively concentrated in Yunnan Province by the RF model and relatively large by the MaxEnt model, and the distribution of O. hupensis snails predicted by the ensemble model was mostly the joint distribution of O. hupensis snails predicted by RF and MaxEnt models. Both RF and MaxEnt models are effective to predict the potential distribution of O. hupensis snails in Yunnan Province, which facilitates targeted O. hupensis snail control.

The main objective of this study is to investigate potential application of frequency ratio (FR), weights of evidence (WoE), and statistical index (SI) models for landslide susceptibility mapping in a part of Mazandaran Province, Iran. First, a landslide inventory map was constructed from various sources. The landslide inventory map was then randomly divided in a ratio of 70/30 for training and validation of the models, respectively. Second, 13 landslide conditioning factors including slope degree, slope aspect, altitude, plan curvature, stream power index, topographic wetness index, sediment transport index, topographic roughness index, lithology, distance from streams, faults, roads, and land use type were prepared, and the relationships between these factors and the landslide inventory map were extracted by using the mentioned models. Subsequently, the multi-class weighted factors were used to generate landslide susceptibility maps. Finally, the susceptibility maps were verified and compared using several methods including receiver operating characteristic curve with the areas under the curve (AUC), landslide density, and spatially agreed area analyses. The success rate curve showed that the AUC for FR, WoE, and SI models was 81.51, 79.43, and 81.27, respectively. The prediction rate curve demonstrated that the AUC achieved by the three models was 80.44, 77.94, and 79.55, respectively. Although the sensitivity analysis using the FR model revealed that the modeling process was sensitive to input factors, the accuracy results suggest that the three models used in this study can be effective approaches for landslide susceptibility mapping in Mazandaran Province, and the resultant susceptibility maps are trustworthy for hazard mitigation strategies.

GIS-based evaluation of landslide susceptibility using hybrid computational intelligence models

Landslide susceptibility evaluation is critical for landslide prevention and risk management. Based on the slope unit, this study uses the information value method- random forest (IV-RF) model to evaluate the landslide susceptibility in the deep valley area. First, based on the historical landslide data, a landslide inventory was developed by using remote sensing technology (InSAR and optical remote sensing) and field investigation methods. Twelve factors were then selected as the input data for a landslide susceptibility model. Second, slope units with different scales were obtained by the r.slopeunits method and the information value method- random forest (IV-RF) model is used to evaluate the landslide susceptibility. Finally, the spatial distribution characteristics of landslide susceptibility grade under the optimal scale are analyzed. The results showed that under the slope unit obtained when c = 0.1 and a = 3 × 105 m2, the internal homogeneity/external heterogeneity of 8425 slope units extracted by the r.slopeunits method is the best, with an AUC of 0.905 and an F1 of 0.908. In this case, the accuracy of landslide susceptibility evaluation is the highest as well; it is shown that the finer slope units would not always lead to the higher accuracy of landslide susceptibility evaluation results; it is necessary to comprehensively consider the internal homogeneity and external heterogeneity of the slope units. Under the optimal slope unit scale, the number of landslides in the highly and extremely highly susceptible areas in the landslide susceptibility map accounted for 82.60% of the total number of landslides, which was consistent with the actual distribution of landslides; this study shows that the method, combining the slope unit and the information value method- random forest (IV-RF) model, for landslide susceptibility evaluation can obtain high accuracy.

This study aimed to produce landslide susceptibility maps using frequency ratio (FR) and evidential belief function (EBF) models based on GIS for Gongliu County, China. For this purpose, a detailed landslide inventory map was prepared and 12 landslide conditioning factors were considered; these factors were as follows: slope angle, slope aspect, curvature, plan curvature, profile curvature, altitude, the normalized difference vegetation index, rainfall, the topographic wetness index (STI), distance from roads, distance from rivers, and lithology. The landslide inventory map was prepared from published sources and aerial photographs, supported by field work. A total of 233 landslides were identified and mapped in the study area. Out of which, 70 % landslides were applied for establishing the model and 30 % landslides were used to validate the model. GIS software was used to analyze landslide conditioning factors and to map landslide susceptibility. Landslide susceptibility maps were prepared using FR and EBF models based on ArcGIS 10.0 and classified into five susceptibility zones: very low, low, moderate, high, and very high. Finally, in order to validate the accuracy of the landslide susceptibility maps produced from two models, the area under curve approach was applied. The validation results showed that the maps using FR and EBF models have a success rate of 82.41 and 79.69 %, respectively. Similarly, the maps using FR and EBF models have a predictive rate of 77.26 and 68.79 %, respectively. Based on the results, both landslide susceptibility maps produced by the two models have a high accuracy and will be helpful for land-use planning in the study area.

This study attempted to predict ground subsidence occurrence using statistical and machine learning models, specifically the evidential belief function (EBF), index of entropy (IoE), support vector machine (SVM), and random forest (RF) models in the Rafsanjan Plain in southern Iran to investigate 11 possible causative factors: slope percent, aspect, topographic wetness index (TWI), plan and profile curvatures, normalized difference vegetation index (NDVI), land use, lithology, distance to river, groundwater drawdown, and elevation. The Boruta algorithm was applied to determine the importance of the possible causative factors. NDVI, groundwater drawdown, land use, and lithology had the strongest relationships with land subsidence. Finally, we generated land subsidence maps using different machine learning and statistical models. The accuracy of these models was assessed using the AUC value and the true skill statistic (TSS) metrics. The SVM model had the highest prediction accuracy (AUC = 0.967, TSS = 0.91), followed by RF (AUC = 0.936, TSS = 0.87), EBF (AUC = 0.907, TSS = 0.83), and IoE (AUC= 0.88, TSS = 0.8).

Chapter 9 - Landslide susceptibility modeling using a generalized linear model in a tropical river basin of the Southern Western Ghats, India

Landslide susceptibility mapping using frequency ratio, logistic regression, artificial neural networks and their comparison: A case study from Kat landslides (Tokat—Turkey)

Correb area in Mekdela Woreda, South-Wollo Zone, North-Western Ethiopia is one of the most landslide affected region. The primary aim of this study was to assess the landslide susceptibility of the area and create a landslide susceptibility map. A bivariate statistical frequency ratio model was used to accomplish this goal. Using fieldwork and Google Earth picture interpretation, a map of landslides was made. A total of 524 landslides were identified and then classified in to training landslide (75%) for model development and validation landslide (25%) for model validation. Eight landslide causative factors slope, aspect, elevation, distance to drainage, distance to spring, slope material, distance to lineament, and vegetation cover were integrated with training landslides in order to determine frequency ratio value(weight) for each class of landslide causative factors. Furthermore, each causative factor’s prediction rate or weight was established. Relative frequency values were allocated to the appropriate factor classes and prediction rate values were allocated to the appropriate causative factors and summed using a raster calculator algorithm in order to produce the landslide susceptibility map. The final landslide susceptibility map revealed that 27.54% (31.52 km2) of the study area falls under very low susceptible, 30.35% (34.74 km2) low susceptible, 21.36% (24.44 km2) moderate susceptible, 14.74% (16.88 km2) high susceptible, and the rest 6.01% (6.88 km2) as very high susceptible. This shows that more than 20% of the area has probability and risk of landslide while about 80% has relatively low probability and risk of this natural hazard. The performance of modified frequency ratio model for landslide susceptibility mapping was validated using simply overlay, relative landslide density index, and receiver operating characteristics curve. Both training and validation landslides were rare in the very low and low susceptibility classes and strongly concentrated in the high and very high susceptibility classes, according to validation using simple overlay and the relative landslide density index. The implemented model performed exceptionally well, as seen by the receiver operating characteristics curve, which showed 82.6% success rate and an 83.1% prediction rate. In general, the model produced reasonable accuracy. The resultant map would be useful for general land use planning, site selection, and landslide prevention and mitigation programs.

Potential impacts of future climate on the spatio-temporal variability of landslide susceptibility in Iran using machine learning algorithms and CMIP6 climate-change scenarios

The present study is dealt with the application of weighted overlay analysis (WOA) model, certainty factor (CF) model, analytical hierarchy process (AHP) model for the preparation of landslide susceptibility zonation map of Darjeeling Himalaya. To perform three models, various data layers with regard to elevation, slope aspect, slope angle, slope curvature, geology, soil, lineament density, distance to lineament, drainage density, distance to drainage, stream power index (SPI), topographic wetted index (TWI), rainfall, normalized differential vegetation index (NDVI) and land use and land cover (LULC) were taken into account. For the preparation of various data layers, topographical maps, Google earth images, SRTM DEM, http://www.worldclim.org, satellite image (Landsat TM) and some authorized data were being processed on GIS environment (ArcMap 10.1). The prepared landslide susceptibility maps of Darjeeling Himalaya were classified into five, i.e. very low, low, moderate, high, and very high landslide susceptibility. To validate three landslide susceptibility zonation maps derived from WOA, CF, and AHP models, ROC Curve and frequency ratio plot methods were incorporated. ROC curve showed the level of accuracy of each landslide susceptibility map. The study revealed that WOA, CF, and AHP were with the accuracy level of 65.4%, 81.2%, and 67.5%. Frequency ratio plots sugessted that moderate, high, and very high landslide susceptibility zones in Darjeeling Himalaya are experienced with greater probability landslide phenomena.

Landslides are a form of soil erosion threatening the sustainability of some areas of the world. There is, therefore, a need to investigate landslide rates and behaviour. In this research, we introduced a novel hybrid artificial intelligence approach of rotation forest (RF) as a meta classifier based on reduced error pruning tree (REPTree) as a base classifier called RF-REPTree, for landslide susceptibility mapping (LSM) in the Kalaleh watershed, Golestan Province, Iran. Some benchmark models, including the open-source Java decision tree (J48), naive Bayes tree (NBTree), and REPTree were used to compare the designed model. A total of 249 landslide locations were identified and mapped. The group was split into training (70%) and testing (30%) data for modelling and reliability analysis. Based on a literature review and multi-collinearity tests, 16 landslide conditioning factors (LCFs) were selected. Of the LCFs, the topographical position index (TPI) had the highest correlation with landslide occurrence. The LSM produced by RF-REPTree revealed that nearly 29% of the study areas have high to very high landslide susceptibility (LS). Statistical analysis of the model results included the receiver operating characteristic curve (ROC), the efficiency test, the true skill statistic (TSS), and the kappa index. ROC demonstrated that the AUC values of RF-REPTree, REPTree, J48, and NBTree models were 0.832, 0.700, 0.695, and 0.759 for succession rate curves and 0.794, 0.740, 0.788, and 0.728 for prediction rate curves, respectively. Therefore, all models were judged to be acceptably accurate for LSM. Among the LS models, the RF-REPTree model achieved the highest accuracy, followed by REPTree, J48, and NBTree. The results of LSM can be used to target the mitigation of landslide hazards and provide a foundation for sustainable environmental planning.

Landslide susceptibility maps are widely used in landslide hazard management. Although many models have been proposed, mapping unit definition is a matter that still needs to be fully examined. In the literature, the most reported mapping units are pixels and slope units, while in this work, developed in the Rio de Janeiro region (Brazil), the use of drainage basins as a mapping unit is examined; even if their use leads to the definition of maps with a coarser spatial resolution than pixels-based maps, they convey information that can be easily and rapidly handled by civil defense organizations. However, for the morphometrical characterization of entire basins, a standardized procedure does not exist, and the susceptibility results may be sensitive to the approach used. To investigate this issue, a random forest model was used to assess landslide susceptibility, using 12 independent variables: four categorical (land use, soil type, lithology and slope orientation) and eight numerical variables (slope gradient, elevation, slope curvature, profile curvature, planar curvature, flow accumulation, topographic wetness index, stream power index). For each basin, the numerical variables were aggregated according to different approaches, which, in turn, were used to set up four different model configurations: i) maximum values, ii) mean values, iii) standard deviation values, iv) joint use of all the above. The resulting maps showed noticeable differences and a quantitative validation procedure showed that the best configurations were the ones based on mean values of independent variables, and the one based on the combination of all the values of the numerical variables. The main outcomes of this work consist of a landslide susceptibility map of the study area, to be used in operational procedures of risk management and in some insights on the best approaches to aggregate raster cell data into wider spatial units.

This study presented herein compares the effect of the sampling strategies by means of landslide inventory on the landslide susceptibility mapping. The conditional probability (CP) and artificial neural networks (ANN) models were applied in Sebinkarahisar (Giresun–Turkey). Digital elevation model was first constructed using a geographical information system software and parameter maps affecting the slope stability such as geology, faults, drainage system, topographical elevation, slope angle, slope aspect, topographic wetness index, stream power index and normalized difference vegetation index were considered. In the last stage of the analyses, landslide susceptibility maps were produced applying different sampling strategies such as; scarp, seed cell and point. The maps elaborated were then compared by means of their validations. Scarp sampling strategy gave the best results than the point, whereas the scarp and seed cell methods can be evaluated relatively similar. Comparison of the landslide susceptibility maps with known landslide locations indicated that the higher accuracy was obtained for ANN model using the scarp sampling strategy. The results obtained in this study also showed that the CP model can be used as a simple tool in assessment of the landslide susceptibility, because input process, calculations and output process are very simple and can be readily understood.

Landslide susceptibility mapping is vital for disaster management and sustainable land-use planning. This research was conducted in Wayanad, Kerala, India, to identify landslide susceptible zones. The study used large geospatial datasets, such as elevation, slope, aspect, curvature, stream power index, topographic wetness index, land use and land cover, rainfall, flow accumulation, geology, and geomorphology. It is followed by the application of various machine learning and deep learning models such as the support vector machine, artificial neural networks, logistic regression, random forest, gradient boosting machine, recurrent neural networks long short-term memory, and deep neural network models to map the landslide susceptible zones. The model was trained and validated using the landslide inventory map, which contains 298 sites of landslides. The random forest model, with 97 % accuracy, performed the best. It is possible to effectively mitigate landslides and plan long-term land use by identifying hazardous zones within the study region.