Weights of evidence modeling for landslide susceptibility mapping of Kabi-Gebro locality, Gundomeskel area, Central Ethiopia

Kabi-Gebro locality of Gundomeskel area is located within the Abay Basin at Dera District of North Shewa Zone in the Central highland of Ethiopia and it is about 320Km from Addis Ababa. This is characterized by undulating topography, intense rainfall, active erosion and highly cultivated area. Geologically, it comprises weathered sedimentary and volcanic rocks. Active landslides damaged the gravel road, houses and agricultural land. The main objective of this research is to prepare the landslide susceptibility map using GIS-based Weights of Evidence model. Based on detailed field assessment and Google Earth image interpretation, 514 landslides were identified and classified randomly into training landslides (80%) and validation landslides (20%). The most common types of landslides in the study area include earth slide (rotational and translational slide), debris slide, debris flow, rock fall, topple, rock slide, creep and complex. Nine landslide causative factors such as lithology, slope, aspect, curvature, land use/land cover, distance to stream, distance to lineament, distance to spring and rainfall were used to prepare a landslide susceptibility map of the study area by adding the weights of contrast values of these causative factors using a rater calculator of the spatial analyst tool in ArcGIS. The final landslide susceptibility map was reclassified as very low, low, moderate, high and very high susceptibility classes. This susceptibility map was validated using landslide density index and area under the curve (AUC). The result from this model validation showed a success rate and a validation rate accuracy of 82.4% and 83.4% respectively. Finally, implementing afforestation strategies on bare land, constructing surface drainage channels & ditches, providing engineering reinforcements such as gabion walls, retaining walls, anchors and bolts whenever necessary and prohibiting hazardous zones can be recommended in order to lessen the impact of landslides in this area.

GIS-Based Weights-of-Evidence Model for Landslide Susceptibility Mapping and Assessment in the Hillslopes of Bujumbura, Burundi

The aim of this research is the modelling of landslide susceptibility in the hillslopes of Bujumbura using the Weights-of-Evidence model, a probabilistic data modelling approach relevant for predicting future landslides at a regional scale. Initially, characteristics and spatial mapping of different landslides type were identified (fall, flow, slide, complex) by thorough interpretation of high-resolution remote sensing data (mountainous areas with difficult access) and intensive fieldwork. Subsequently, the main landslides controlling factors were selected (lithology, fault density, land use, drainage density, slope aspect, curvature, slope angle, and elevation) using in-depth field knowledge and relevant literature. A landslide inventory map with a total of 569 landslide sites was constructed using the data from various sources. Out of those 569 landslide sites, 285 (50.1%) of the data taken before the 2000s was used for training and the remaining 284 (49.9%) sites (post-2000 events) were used for the accuracy assessment purpose. Thereafter, a prediction map of future landslides was generated with an accuracy of 73.7%. The main geo-environmental landslides factors retained are the high density of drainage networks, the lithology often made with weathered gneiss, the high fault density, the steep topography and the convex slope curvature. The landslide susceptibility map validated was reclassified into very high, high, moderate, low and very low zones. The established susceptibility map will allow with the interaction of the real terrain to locate roads, dwellings, urban extension areas, dams located in high landslides risk zones. These infrastructures will require intervention to address their vulnerability with new facilities, slope stabilization, creation of bypass roads, etc. The susceptibility map produced will be a powerful decision-making tool for drawing up appropriate development plans. Such an approach will make it possible to mitigate the socio-economic impacts due to slope instabilities.

Flash-Flood Potential Mapping Using Deep Learning, Alternating Decision Trees and Data Provided by Remote Sensing Sensors

There is an evident increase in the importance that remote sensing sensors play in the monitoring and evaluation of natural hazards susceptibility and risk. The present study aims to assess the flash-flood potential values, in a small catchment from Romania, using information provided remote sensing sensors and Geographic Informational Systems (GIS) databases which were involved as input data into a number of four ensemble models. In a first phase, with the help of high-resolution satellite images from the Google Earth application, 481 points affected by torrential processes were acquired, another 481 points being randomly positioned in areas without torrential processes. Seventy percent of the dataset was kept as training data, while the other 30% was assigned to validating sample. Further, in order to train the machine learning models, information regarding the 10 flash-flood predictors was extracted in the training sample locations. Finally, the following four ensembles were used to calculate the Flash-Flood Potential Index across the Bâsca Chiojdului river basin: Deep Learning Neural Network–Frequency Ratio (DLNN-FR), Deep Learning Neural Network–Weights of Evidence (DLNN-WOE), Alternating Decision Trees–Frequency Ratio (ADT-FR) and Alternating Decision Trees–Weights of Evidence (ADT-WOE). The model’s performances were assessed using several statistical metrics. Thus, in terms of Sensitivity, the highest value of 0.985 was achieved by the DLNN-FR model, meanwhile the lowest one (0.866) was assigned to ADT-FR ensemble. Moreover, the specificity analysis shows that the highest value (0.991) was attributed to DLNN-WOE algorithm, while the lowest value (0.892) was achieved by ADT-FR. During the training procedure, the models achieved overall accuracies between 0.878 (ADT-FR) and 0.985 (DLNN-WOE). K-index shows again that the most performant model was DLNN-WOE (0.97). The Flash-Flood Potential Index (FFPI) values revealed that the surfaces with high and very high flash-flood susceptibility cover between 46.57% (DLNN-FR) and 59.38% (ADT-FR) of the study zone. The use of the Receiver Operating Characteristic (ROC) curve for results validation highlights the fact that FFPIDLNN-WOE is characterized by the most precise results with an Area Under Curve of 0.96.

The Impact of Quality of Digital Elevation Models on the Result of Landslide Susceptibility Modeling Using the Method of Weights of Evidence

The paper discusses the impact that the quality of the digital elevation model (DEM) has on the final result of landslide susceptibility modeling (LSM). The landslide map was developed on the basis of the analysis of archival geological maps and the Light Detection and Ranging (LiDAR) digital elevation model. In addition, complementary field studies were conducted. In total, 92 landslides were inventoried and their degree of activity was assessed. An inventory of the landslides was prepared using a 1-m-LiDAR DEM and field research. Two digital photogrammetric elevation models with an elevation pixel resolution of 20 m were used for landslide susceptibility modeling. The first digital elevation model was obtained from a LiDAR point cloud (DEM–airborne laser scanning (ALS)), while the second model was developed based on archival digital stereo-pair aerial images (DEM–Land Parcel Identification System (LPIS)). Both models were subjected to filtration using a Gaussian low-pass filter to reduce errors in their elevation relief. Then, using ArcGIS software, a differential model was generated to illustrate the differences in morphology between the models. The maximum differences in topographic elevations between the DEM–ALS and DEM–LPIS models were calculated. The Weights-of-Evidence model is a geostatistical method used for the landslide susceptibility modeling. Six passive factors were employed in the process of susceptibility generation: elevation, slope gradient, exposure, topographic roughness index (TRI), distance from tectonic lines, and distance from streams. As a result, two landslide susceptibility maps (LSM) were obtained. The accuracy of the landslide susceptibility models was assessed based on the Receiver Operating Characteristic (ROC) curve index. The area under curve (AUC) values obtained from the ROC curve indicate that the accuracy of classification for the LSM–DEM–ALS model was 78%, and for the LSM–LPIS–DEM model was 73%.