Spatial Distribution Of Landslides Research Articles

Landslide Distribution and Development Characteristics in the Beiluo River Basin

The Beiluo River Basin, situated in the central region of the Loess Plateau, frequently experiences landslide geological disasters, posing a severe threat to local lives and property. Thus, establishing a detailed database of historical landslides and analyzing and revealing their development characteristics are of paramount importance for providing a foundation for geological hazard risk assessment. First, in this study, landslides in the Beiluo River Basin are interpreted using Google Earth and ZY-3 high-resolution satellite imagery. Combined with a historical landslide inventory and field investigations, a landslide database for the Beiluo River Basin is compiled, containing a total of 1781 landslides. Based on this, the geometric and spatial characteristics of the landslides are analyzed, and the relationships between the different types of landslides and landslide scale, stream order, and geomorphological types are further explored. The results show that 50.05% of the landslides have a slope aspect between 225° and 360°, 68.78% have a slope gradient of 16–25°, and 38.97% are primarily linear in profile morphology. Areas with a high landslide density within a 10 km radius are mainly concentrated in the loess ridge and hillock landform region between Wuqi and Zhidan Counties and in the loess tableland region between Fu and Luochuan Counties, with a significant clustering effect observed in the Fu County area. Loess–bedrock interface landslides are relatively numerous in the northern loess ridge and hillock landform region due to riverbed incision and the smaller thickness of loess in this area. Intra-loess landslides are primarily found in the southern loess tableland region due to headward erosion and the greater thickness of loess in this area. Loess–clay interface landslides, influenced by riverbed incision and the limited exposure of red clay, are mainly distributed in the northern part of the southern loess tableland region and on both sides of the Beiluo River Valley in Ganquan County. These results will aid in further understanding the development and spatial distribution of landslides in the Beiluo River Basin and provide crucial support for subsequent landslide susceptibility mapping and geological hazard assessment in the region.

Human activities are intensifying the spatial variation of landslides in the Yellow River Basin

Human activities are a triggering factor for landslides in the Yellow River Basin (YRB, China). However, the extent to which the spatial distribution of landslides is affected by human activities is unclear. We constructed a human activity intensity index (HAII) based on nighttime light data and land cover data. Regression and dominance analyses were used to compare the effects of the HAII, precipitation, distance to river, distance to fault, topographic relief and slope on the landslides spatial density (LSD). The results showed that in the YRB, the HAII, as a dominance influencing factor, had a significant positive influence on the LSD. Moreover, regional differences in the human disturbance of nature intensify the spatial variation of LSD. To quantify the intensity of human disturbance to nature, a human-nature conflict index (HNCI) is constructed by quantifying the difference between the slope distributions of artificial and natural landscapes. The results show that in the middle section of the YRB, humans are developing more steep mountainous areas, leading to more dense landslides. This study provides a reference for landslide risk management and land use planning in the YRB.

A SHAP-enhanced XGBoost model for interpretable prediction of coseismic landslides

In the coseismic landslide hazard assessment (CLHA), advanced machine learning (ML) models have garnered significant attention due to their effectiveness in handling the complex relationships between landslides and various influencing factors. However, explaining the decision-making mechanisms of machine learning models that predict landslide spatial distribution based on these influencing factors remains challenging. This study compares the predictive performance of four models—XGBoost, RF, LR, and SVM—optimized using the TPE algorithm, and introduces the SHAP algorithm into the XGBoost model to achieve both global and local interpretations of CLHA. In various tests, the optimized XGBoost model demonstrated the best predictive performance, achieving an accuracy of 0.864 and an AUC value of 0.886. Global interpretations indicate that the occurrence of coseismic landslides is primarily influenced by triggering factors such as hypocentral distance and distance from the seismogenic fault. Terrain roughness and elevation, on the other hand, make significant contributions among the conditioning factors. Single-factor dependence plots indicate that the contribution of individual factors to landslides varies across different ranges of their feature values. Analysis of two-factor dependence plots reveals that interactions between factors are also crucial in influencing the occurrence of landslides. Combining field surveys with local interpretations confirms significant variations in the contributions of influencing factors within local ranges. The main innovation of this study lies in the integration of the SHAP algorithm into the CLHA model, revealing the decision-making mechanisms of the model for spatial prediction of coseismic landslides.

Study on Temporal and Spatial Distribution of Landslides in the Upper Reaches of the Yellow River

The geological structure of the upper reaches of the Yellow River is complex, especially in the Sigouxia-Laganxia section. It has always been a high-incidence area of landslide disasters, which poses a threat to the safe operation of the upper reaches of the Yellow River. In this study, based on the high-precision remote sensing image data, the spatial distribution of each landslide was obtained by superimposing the remote sensing image and the 1:50,000 digital elevation model (DEM). Some typical landslides were selected for detailed field investigation and field verification. The results show that the remote sensing image characteristics of landslides in the upper reaches of the Yellow River are obvious. Through remote sensing interpretation and field investigation, a total of 508 landslides of various types were found, including 24 giant landslides. The spatial spreading patterns of landslides mainly include six types: dumb-bell shape, rectangle, saddle type, long arc shape, triangle, and side-by-side shape. The length and width of the landslide deposit are mainly concentrated at 550–1500 m and 600–1500 m, and the average elevation of the sliding body is mainly concentrated between 2000 and 2800 m. The average slope of the landslide is mainly distributed between 15–20°. Giant landslides are mainly distributed in the Jianzha basin area, surrounded by the Jishishan fault and the Lajishan fault in the West Qinling Mountains. The spatial distribution characteristics of giant landslides have obvious regional differences due to different factors such as lithologic differences and riverside erosion. The research results are of great significance for the early identification, prevention, and mitigation of landslide disasters in the upper reaches of the Yellow River.

The Spatial Distribution Characteristics and Possible Influencing Factors of Landslide Disasters in the Zhaotong Area, Yunnan Province of China

The Zhaotong area in Yunnan Province stands out as one of the most susceptible areas to landslide disasters. The landslide susceptibility of the Zhaotong area can be attributed to its steep terrain, fractured rock formations and strong rainfall, compounded by its frequent seismic activity. This study utilized landslide data provided by the Zhaotong City Natural Resources and Planning Bureau and visually interpreted from high-resolution satellite images of Google Earth to establish the landslide database of the Zhaotong area, including 161 landslides and 3646 potential geological disasters. The distribution characteristics and possible influencing factors of landslides within the Zhaotong area were analyzed using the aforementioned data. The results show that the spatial distribution of landslides and potential geological disasters is roughly consistent; the most concentrated landslides occurred at the junction of Yiliang County, Zhaotong City, and Daguan County, indicating the necessity to enhance surveillance of these landslide-prone areas. The relationship of landslide locations and different influencing factors suggests that elevation, slope angle, and distance to rivers are closely related to landslide occurrence. Landslides are more likely to occur in areas with lower elevations with slope angles ranging from 10° to 40° and near river channels.

Investigating the reactivation of historical landslides during the 2022 Luding MS6.8 earthquake

On September 5, 2022, a strong earthquake with a magnitude of MS6.8 struck Luding County in Sichuan Province, China, triggering thousands of landslides along the Dadu River in the northwest-southeast (NW-SE) direction. We investigated the reactivation characteristics of historical landslides within the epicentral area of the Luding earthquake to identify the initiation mechanism of earthquake-induced landslides. Records of the two newly triggered and historical landslides were analyzed using manual and threshold methods; the spatial distribution of landslides was assessed in relation to topographical and geological factors using remote sensing images. This study sheds light on the spatial distribution patterns of landslides, especially those that occur above historical landslide areas. Our results revealed a similarity in the spatial distribution trends between historical landslides and new ones induced by earthquakes. These landslides tend to be concentrated within a range of 0.2 km from the river and 2 km from the fault. Notably, both rivers and faults predominantly influenced the reactivation of historical landslides. Remarkably, the reactivated landslides are characterized by their small to medium size and are predominantly situated in historical landslide zones. The number of reactivated landslides surpassed that of previously documented historical landslides within the study area. We provide insights into the critical factors responsible for historical landslides during the 2022 Luding earthquake, thereby enhancing our understanding of the potential implications for future co-seismic hazard assessments and mitigation strategies.

High-resolution landslide mapping and susceptibility assessment: Landslide temporal variations and vegetation recovery

In mountainous terrains, the frequent landslides and their associated impacts on human lives and the economy is increasing globally. Development of landslide inventory and afterward landslide susceptibility mapping are the main prerequisites for implementing landslide mitigation measures and protection in mountainous regions. The 2005 Kashmir earthquake induced different small and large landslides and some were active for the long term. So far many studies have used medium and high-resolution data to develop landslide inventory. This study aims to develop a 1st detailed, comprehensive and accurate landslide inventory using a very high-resolution image using a semi-automatic technique. The precise landslide inventory is employed to develop an accurate and comprehensive landslide susceptibility map considering the landslide inventory data logistic regression training model. Furthermore, the landslide’s temporal recovery from the earthquake and its reactivation due to rainfall in spare vegetation areas have been evaluated. Fine-resolution satellite images of Worldview-2 are applied to develop a detailed landslide inventory using a Support Vector Machine (SVM) classifier. A total of 63,630 landslides were identified using a semi-automatic technique within a study area of 265 km2. From regression modeling, the results show that geology, topography, and road networks have a significant impact on the spatial distribution of landslides. Model performance was evaluated based on the testing data, the model gives an AUC of 0.93 and the kappa value of 0.9353. The spatiotemporal NDVI has been and assessed to identify the landslide recovery and its reactivation due to extreme rainfall. The results show that 72.1 % of the landslides occurred in Muzaffarabad formation in the study area. The developed landslide susceptibility map can be further used for land-use planning and implementing mitigation measures for the safety of roads and other infrastructure in the area.

Early Identification and Dynamic Stability Evaluation of High-Locality Landslides in Yezhi Site Area, China by the InSAR Method

In mountainous regions, high-locality landslides have the characteristics of a latent disaster process with a wide disaster range, which can easily cause large casualties. Therefore, early landslide identification and dynamic stability evaluation are significant. We first used multi-temporal synthetic aperture radar data to detect potential landslides at Yezhi Site Area during the 2015–2020 period, identifying and mapping a total of 18 active landslides. The study area was found to have an average deformation rate between −15 and 10 mm/y during the period. Then, time series and spatiotemporal deformation characteristics of landslides were examined using interferogram stacking and small baseline interferometry techniques. The results show that the majority of the landslide deformations detected exhibit a periodic variation trend, and the study area was in a slow deformation state before 2017. Finally, combined with detection results, Google Earth optical images, and field investigations, it is concluded that the main factors affecting the time series deformation and spatial distribution of landslides in the study area are rainfall, geological factors, and engineering activities. The results of this study provide valuable technical references and support for early identification and dynamic stability evaluation of regional active landslides in complex terrain, especially for high-locality landslides.

Spatial distributions and multi-factor driving mechanism of landslide in southern Liaodong Peninsula

Coastal areas are characterized by high levels of urbanization and also the most active and highly sensitive to geological activities. Due to the impact of global climate change, landslides occur frequently in coastal areas, which have significantly impacted the sustainable socio-economic development of the region. Therefore, studying the spatial distribution of landslides and the factors that drive them holds immense practical significance for the geohazards prevention and mitigation work and optimization of ecological security patterns in coastal areas. This study takes the southern Liaodong Peninsula as the study area and analyzes the spatial distribution characteristics and influence mechanism of landslides. We first applied the Getis-Ord Gi*, kernel density analysis techniques, and Ripley’s K function to explore the spatial distribution characteristics of landslides in the study area at different scales. Then, we employed the information value method to test for the linkage relationships between the driving factors and landslides. Finally, we utilized the geographical detector to reveal the potential impacts and interaction of the driving factors on landslides. The results showed that: (1) landslides in the Southern Liaodong Peninsula exhibit strong spatial clustering characteristics, and have a spatial scale effect; (2) the high susceptibility areas are mainly concentrated in the southern hilly regions and the mountainous regions in the northern parts of the study area; (3) The information value method reveals that there is an optimal combination of factors driving landslides. (4) The factor detector analysis reveals that the primary driver of landslide spatial distribution is the geomorphological types (q-value of 39.10%). (5) The interaction detector indicates that the interplay of all driving factor pairs exhibited an enhanced effect, which is not a simple superposition effect but bivariate and non-linear. The collective influence of natural factors and human activities holds a more substantial influence over landslide development and distribution. The research offers guidance for the execution of landslide prevention initiatives and the establishment of preservation policies for the ecological environment in the southern Liaodong Peninsula region.

Investigation on the spatial distribution of landslides in Sichuan Province, southwest China

In this study, a total of 34,893 landslides collected from the Sichuan Province of China are used to reveal their spatial distribution. Correlations between the spatial distribution of landslides and five main factors [i.e. topography, rainfall, lithology, peak ground acceleration (PGA), and human activity] are analyzed based on the Geographical Information System (GIS) platform. Topographic factors exert a significant influence on landslide occurrence, and other factors also sensitive to landslides include lithology (Classes 3 and 4), PGA (≤0.15 g), rainfall (90–140 mm), and road density (≤0.6 km/km2). The geographical detector model is used to detect the relative importance of individual influencing factors and the factor interaction effects on the landslide distribution. The results show that the interacting factors between rainfall and PGA, lithology, river density, and road density are more closely related to the spatial distribution of landslides than every single factor. The results of this study can provide a foundation and insights for hazard reduction schemes in the future.

Revealing the relation between spatial patterns of rainfall return levels and landslide density

Abstract. It is known that the spatial rainfall pattern can mark landslide distribution across the landscape during extreme triggering events. However, the current knowledge of rainfall controls on this distribution remains limited. Here, to reveal what rainfall characteristics control landslide spatial distribution, we explore the spatiotemporal pattern of a rainfall event that triggered over 7500 landslides (area ≈ 100–104 m2) at a regional scale with an area of ≈400 km2 in Japan. Using a 5 km resolution radar-driven and gauge-adjusted hourly precipitation dataset with 32 years of records, we compared rainfall return levels for various time ranges from 1 to 72 h and landslide density in each grid cell of the precipitation dataset (≈25 km2). The results show that, even if local slope distributions within the grid cells are comparable, the number of landslides in a ≈25 km2 grid cell was substantially high when rainfall return levels exceeded the 100-year return period in all examined timespans (i.e., 1–72 h). In contrast, when only specific-duration rainfall intensities (e.g., 6–48 h) exceeded the 100-year return level, the landslide density in corresponding grid cells tended to be low. Consequently, the landslide density increased with the increase in rainfall return levels of various timespans rather than a specific rainfall intensity, such as downpours for a few hours or long-term cumulative rainfall for several days. Moreover, with the increase in the landslide density, the number of relatively large landslides exceeding ≈400 m2 increased. Therefore, the spatial differences in rainfall return levels potentially constrain the density of total landsliding and relatively large landslides. In this sense, whether rainfall intensities reach high return levels rarely experienced in a wide timespan ranging from a few hours to several days is one of the key determinants of the spatial distribution of landslides and the extent of related hazards.

Insights into geospatial heterogeneity of landslide susceptibility based on the SHAP-XGBoost model

The spatial heterogeneity of landslide influencing factors is the main reason for the poor generalizability of the susceptibility evaluation model. This study aimed to construct a comprehensive explanatory framework for landslide susceptibility evaluation models based on the SHAP (SHapley Additive explanation)-XGBoost (eXtreme Gradient Boosting) algorithm, analyze the regional characteristics and spatial heterogeneity of landslide influencing factors, and discuss the heterogeneity of the generalizability of the models under different landscapes. Firstly, we selected different regions in typical mountainous hilly region and constructed a geospatial database containing 12 landslide influencing factors such as elevation, annual average rainfall, slope, lithology, and NDVI through field surveys, satellite images, and a literature review. Subsequently, the landslide susceptibility evaluation model was constructed based on the XGBoost algorithm and spatial database, and the prediction results of the landslide susceptibility evaluation model were explained based on regional topography, geology, and hydrology using the SHAP algorithm. Finally, the model was generalized and applied to regions with both similar and very different topography, geology, meteorology, and vegetation, to explore the spatial heterogeneity of the generalizability of the model. The following conclusions were drawn: the spatial distribution of landslides is heterogeneous and complex, and the contribution of each influencing factor on the occurrence of landslides has obvious regional characteristics and spatial heterogeneity. The generalizability of the landslide susceptibility evaluation model is spatially heterogeneous and has better generalizability to regions with similar regional characteristics. Further explanation of the XGBoost landslide susceptibility evaluation model using the SHAP method allows quantitative analysis of the differences in how much various factors contribute to disasters due to spatial heterogeneity, from the perspective of global and local evaluation units. In summary, the integrated explanatory framework based on the SHAP-XGBoost model can quantify the contribution of influencing factors on landslide occurrence at both global and local levels, which is conducive to the construction and improvement of the influencing factor system of landslide susceptibility in different regions. It can also provide a reference for predicting potential landslide hazard-prone areas and for Explainable Artificial Intelligence (XAI) research.

The spatial relationship between landslide and land use activities in Langat River Basin: A case study

Landslide is a natural hazard that occurs all over the world and can cause enormous damage to the environment and human life. The developing countries are not spared from dealing with the issue of landslides where land use activities are recognised as one of the most triggering factors of landslide occurrences. In this study, investigation on spatial relationship was performed between landslide occurrences and land use activities at Langat River Basin. Landslide spatial distribution was determined using land use activities for the year 2021. The land use map was generated using supervised classification by Artificial Neural Network (ANN) approach. In addition, the accuracy assessment was applied using Kappa coefficient to validate the land use classification. A total of 70 historical landslide points was used to analyse the spatial relationship between landslide and land use activities. Based on the results, land use category at Langat River Basin was dominated by agriculture with 59.65%. The overall accuracy for land use classification was 90.81% and kappa coefficient was 0.88. Meanwhile, the spatial relationship between landslide occurrence and land use activities was higher in the built-up area with 48.57%. The growing population and economic sector in Langat River Basin were found to be contributing to the increase in built-up areas. Thus, these results are particularly useful in understanding the spatial relationship between landslide occurrence and land use activities where the development area is more vulnerable to facing landslide disasters. These findings will also help the government to achieve the agenda for Sustainable Development Goal (SDG) 2030 as highlighted to mitigate the risk of natural disasters.

Inventory and Spatial Distribution of Ancient Landslides in Hualong County, China

The establishment of a regional historical landslide inventory plays an indispensable role in landslide assessment and prevention. In this study, based on the Google Earth platform, an inventory of ancient landslides in Hualong County, Qinghai Province was established. The inventory includes 3517 ancient landslides with individual areas ranging from 2354.6 m2 to 12.44 km2. The dominant characteristics include an elevation of 2600–2800 m, slope of 10–20°, aspects SW, W, and NW, mudstone and sandstone of Paleoproterozoic, Neoproterozoic and Quaternary loess, 8–10 km from faults, 0–1 km from rivers, cultivated and grassland types, NDVI of 0.25–0.3, and an average precipitation in the range of 480–500 mm. In addition, the geometric analysis of landslides shows that the average height and length of ancient landslides in the study area are 151.92 m and 429.52 m, respectively. The power law relationship between the two is L = 0.41 × H1.37. The ancient landslide inventory of this study exhibits an integrated pattern of the development characteristics and spatial distribution of landslides in the Tibetan Plateau and the upper Yellow River basin, as well as providing a significant reference for subsequent landslide susceptibility mapping in the area.

Two public inventories of landslides induced by the 10 June 2022 Maerkang Earthquake swarm, China and ancient landslides in the affected area

The June 10, 2022 Ms5.8, Ms6.0 and Ms5.2 earthquake swarm of Maerkang, Sichuan, China has triggered a lot of landslides. The primary purpose of this paper is to prepare a detailed coseismic landslides inventory and ancient landslide inventory in the earthquake swarm-affected area. Based on pre- and post-quake high-resolution optical satellite images (planet), we delineated 650 individual coseismic landslides in the VI-intensity area and above, occupying an area of about 1.2 ​km2. The largest landslide covers about 70217 ​m2 in horizontal projection, while the smallest one is just 81 ​m2. In addition, based on the series of high-precise images on the Google Earth Platform, 759 ancient landslides were also delineated with an area of 117 ​km2. The maximum area reaches about 2 ​km2 and the minimum is 2580 ​m2. The two inventories provide a basis for landslide spatial distribution and hazard assessment, disaster prevention and mitigation of earthquake-triggered landslides in similar regions. We opened these two landslide inventories to facilitate researchers to carry out more in-depth work.

National and regional-scale landslide indicators and indexes: Applications in Italy

Abstract Indicators and indexes are quantifiable parameters used in scientific disciplines to summarize and communicate complex data in a simple and explanatory manner. In the field of natural hazards, indicators and indexes have been used to characterize natural processes, and the associated risk conditions in terms of impact, vulnerability, exposure, and resilience. In this paper, we formalize indicators at the municipal level to differentiate the Italian territory based on the spatial distribution of landslides. The indicators were combined with other information to define indexes able to better characterize the stability conditions of the municipalities and quantify the possible impact of slope movements on the road network. Indexes were defined only for the Umbria Region (Central Italy), which was chosen as an example. The proposed indicators and indexes show, in a simple way, the severity of the instability on the territory and can be used to support decision-makers to assess, evaluate, and manage landslide mitigation activities and civil protection actions.

Influence of DEM resolution on landslide simulation performance based on the Scoops3D model

Using the physical deterministic model to analyze landslide stability has become a hotspot of landslide disasters research all over the world. The Digital Elevation Model (DEM) resolution has a great influence on the simulation effect of 3 D physical models. However, few researchers have studied the prediction performance of the 3 D models under different DEM resolutions. Therefore, based on the 3 D model Scoops3D, the spatial distribution of landslides was simulated and predicted under five different DEMs resolutions (2.5 m, 5 m, 10 m, 20 m, and 30 m). The optimal parameters of the model were obtained through field investigations and laboratory experiments, and then, the simulation results were compared with the actual landslides distribution. Receiver operating characteristic (ROC) analysis and %LRclass index were used to quantitatively evaluate the prediction performance of the 3 D model under five different DEMs resolutions. The results show that Scoops3D has good performance in landslide spatial distribution prediction. In addition, we also found that the simulation results of high-resolution DEM were not ideal, while the prediction results of medium resolution DEMs (i.e., 5 m and 10 m) were more accurate. Therefore, this study provide a reference to select the most suitable DEM resolution for landslide stability analysis.

Modelling the systematics of cosmogenic nuclide signals in fluvial sediments following extreme events

AbstractThe effect of punctuated mass‐wasting events on longer‐term erosion rates is not fully understood, and yet it is key to quantifying sediment generation, source‐to‐sink dynamics, and landscape evolution in active orogens. The measurement of terrestrial cosmogenic nuclides (TCNs) in river sediments is a common method for determining basin‐averaged erosion rates over centennial to millennial timescales and is often used to compare erosional processes between catchments. However, these comparisons often overlook the role of landsliding rates and their spatial distribution in the measurement and potential variability of TCN signals. While it is widely accepted that basin‐scale perturbations should temporarily dilute TCN concentrations as landsliding mobilizes new, low‐concentration material, the impact of the catastrophic release of hillslope sediment caused by a single event on TCN signatures has not yet been systematically investigated. In this modelling study, we use a catchment in central Nepal to build upon previous modelling efforts to consider how TCNs are recorded in landscapes with varying erosion rates, landsliding rates, and spatial distribution of landslides. We then use the 25 April 2015 Mw 7.8 Gorkha earthquake, Nepal, as a case study to investigate how perturbations like earthquakes are recorded in TCN time series and transferred to and ultimately archived in the sedimentary record. We find that the likelihood of a perturbation being measured by TCN dilution is based on a multitude of factors, including background erosion rates, long‐term landsliding rates, and the connectivity of newly released material to the fluvial system. Especially in landscapes like the central Himalaya with high background erosion and landsliding rates, changes in detrital TCN concentrations are not a reliable indicator of an upstream perturbation, nor should we expect a clear dilution signal following a major event. Our modelling results emphasize that TCN dilution is not a universal characteristic of high‐magnitude landslide‐triggering events.

Spatial distribution and geometric characteristics of landslides with special reference to geological units in the area of Slavonski Brod, Croatia

A preliminary analysis of landslide spatial distribution and their geometric characteristics is presented for the area of Slavonski Brod, located in the northeastern part of Croatia and belonging to the Pannonian Basin System. A landslide inventory for the study area of 55.1 km2 is accomplished for the first time, based on the visual interpretation of a high resolution LiDAR digital terrain model. In total, 854 landslide polygons are delineated, corresponding to an average density of 15.5 landslides per square kilometre. The average landslide area is 839 m2, and most of the landslides can be classified as small landslides (76 %). The spatial relationship between landslides and geological units is analysed and expressed as a landslide index. The Late Pannonian sands with silts and gravel interlayers and Pliocene clay, sands, gravels, and coal are determined as the units that are most susceptible to landslide processes. The majority of landslides (85 %) are concentrated within these two units, for which a detailed analysis is performed, determining the morphometric parameters (slope and relief) and drainage network. The parameters’ classes that create favourable preconditions to slope instabilities are defined, based on the landslide density within individual classes. Besides, the geometric characteristics of landslides (size and shape) within these two units are compared. The results serve as the basis for further investigations. They help to foresee the area of future landslides through landslide susceptibility maps, and offer a better understanding of the influence of fluvial-denudation and slope processes on recent landscape evolution and form.

Predicting spatio-temporal man-made slope failures induced by rainfall in Hong Kong using machine learning techniques

Rain-induced man-made slope failures pose great threats to public safety as most man-made slopes are formed in densely populated areas. A critical step in managing landslide risks is to predict the time, locations and consequences of slope failures in future rainstorms. Based on comprehensive databases of in-service man-made slopes, rainstorms and landslides in Hong Kong during the past 35 years, a spatio-temporal landslide forecasting model for man-made slopes is developed in this study within a unified machine learning framework. With a storm-based data integration strategy and multiclass classification on landslide scales, the framework incorporates landslide time and consequences in landslide susceptibility mapping to successfully achieve spatio-temporal landslide forecasting. The machine learning-based landslide forecasting model is validated against historical landslide incidents both temporally and spatially and through a case study of the June 2008 storm; the model significantly outperforms the prevailing statistical rainfall–landslide correlations in terms of prediction accuracy. The model can predict the real-time evolution of probabilities, scales and spatial distribution of landslides during the progression of a rainstorm, which can never be achieved by statistical methods. It can serve as an essential module for state-of-the-art landslide risk assessment and early warning.