Landslide susceptibility assessment using machine learning models is a popular and consolidated approach worldwide. The main constraint of susceptibility maps is that they are not adequate for ...temporal assessments: they are generated from static predisposing factors, allowing only a spatial prediction of landslides. Recently, some methodologies have been proposed to provide spatiotemporal landslides prediction starting from machine learning algorithms (e.g., combining susceptibility maps with rainfall thresholds), but the attempt to obtain a dynamic landslide probability map directly by applying machine learning models is still in the preliminary phase. This work provides a contribution to fix this gap, combining in a Random Forest (RF) algorithm a static indicator of the spatial probability of landslide occurrence (i.e., a classical susceptibility index) and a number of dynamic variables (i.e., seasonality and the rainfall amount cumulated over different reference periods). The RF implementation used in this work allows the calculation of the Out-of-Bag Error and depicts Partial Dependence Plots, two indices that were used to quantify the variables’ importance and to comprehend if the model outcomes are consistent with the triggering mechanism observed in the case of study (Metropolitan City of Florence, Italy). The goal of this research is not to set up a landslide probability map, but to 1) understand how to populate training and test datasets with observations sampled over space and time, 2) assess which rainfall variables are statistically more relevant for the identification of the time and location of landslides, and 3) test the dynamic application of RF in a forecasting model for the spatiotemporal prediction of landslides. The proposed dynamic methodology shows encouraging results, consistent with the actual knowledge of the physical mechanism of the triggering of shallow landslides (mainly influenced by short and intense rainfalls) and identifies some benchmark configurations that represents a promising starting point for future regional-scale applications of machine learning models to dynamic landslide probability assessment and early warning.
Decision-making plays a key role in reducing landslide risk and preventing natural disasters. Land management, recovery of degraded lands, urban planning, and environmental protection in general are ...fundamental for mitigating landslide hazard and risk. Here, we present a GIS-based multi-scale approach to highlight where and when a country is affected by a high probability of landslide occurrence. In the first step, a landslide human exposure equation is developed considering the landslide susceptibility triggered by rain as hazard, and the population density as exposed factor. The output, from this overview analysis, is a global GIS layer expressing the number of potentially affected people by month, where the monthly rain is used to weight the landslide hazard. As following step, Logistic Regression (LR) analysis was implemented at a national and local level. The Receiver Operating Characteristic indicator is used to understand the goodness of a LR model. The LR models are defined by a dependent variable, presence–absence of landslide points, versus a set of independent environmental variables. The results demonstrate the relevance of a multi-scale approach, at national level the biophysical variables are able to detect landslide hotspot areas, while at sub-regional level geomorphological aspects, like land cover, topographic wetness, and local climatic condition have greater explanatory power.
Landslides are a major natural hazard that can significantly damage infrastructure and cause loss of life. In South Korea, the current landslide susceptibility mapping (LSM) approach is mainly based ...on statistical techniques (logistic regression (LR) analysis). According to previous studies, this method has achieved an accuracy of approximately 75.2%. In this paper, we expand upon this traditional approach by comparing the performance of six machine learning (ML) algorithms for LSM in Inje County, South Korea. The study employed a combination of geographical data gathered from 2005 to 2019 to train and evaluate six algorithms, including LR, Gaussian Naive Bayes (GNB), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Random Forest (RF), and Extreme Gradient Boosting (XGB). The effectiveness of these models was measured by various criteria, such as the percentage of correct classification (PCC) score, F1 score, and Kappa score. The results demonstrated that the PCC and F
1
scores of the six models fell between 0.869–0.941 and 0.857–0.940, respectively. RF and XGB had the highest PCC and F
1
scores of 0.939 and 0.941, respectively. This study indicates that ML can be a valuable technique for high-resolution LSM in South Korea instead of the current approach.
Landslide probability prediction plays an important role in understanding landslide information in advance and taking preventive measures. Many factors can influence the occurrence of landslides, ...which is easy to have a curse of dimensionality and thus lead to reduce prediction accuracy. Then the generalization ability of the model will also decline sharply when there are only small samples. To reduce the dimension of calculation and balance the model’s generalization and learning ability, this study proposed a landslide prediction method based on improved principal component analysis (PCA) and mixed kernel function least squares support vector regression (LSSVR) model. First, the traditional PCA was introduced with the idea of linear discrimination, and the dimensions of initial influencing factors were reduced from 8 to 3. The improved PCA can not only weight variables but also extract the original feature. Furthermore, combined with global and local kernel function, the mixed kernel function LSSVR model was framed to improve the generalization ability. Whale optimization algorithm (WOA) was used to optimize the parameters. Moreover, Root Mean Square Error (RMSE), the sum of squared errors (SSE), Mean Absolute Error (MAE), Mean Absolute Precentage Error (MAPE), and reliability were employed to verify the performance of the model. Compared with radial basis function (RBF) LSSVR model, Elman neural network model, and fuzzy decision model, the proposed method has a smaller deviation. Finally, the landslide warning level obtained from the landslide probability can also provide references for relevant decision-making departments in emergency response.
This study established a probability model based on the landslide spatial and size probabilities to predict the possible volume and locations of landslides in watershed scale under rainfall events. ...First, we assessed the landslide spatial probability using a random forest landslide susceptibility model including intrinsic causative factors and extrinsic rainfall factors. Second, we calculated the landslide volume probability using the Pearson type V distribution. Lastly, these probabilities were joined to predict possible landslide volume and locations in the study area, the Taipei Water Source Domain, under rainfall events. The possible total landslide volume in the watershed changed from 1.7 million cubic meter under the event with 2-year recurrence interval to 18.2 million cubic meter under the event with 20-year recurrence interval. Approximately 62% of the total landslide volume triggered by the rainfall events was concentrated in 20% of the slope units. As the recurrence interval of the events increased, the slope units with large landslide volume tended to concentrate in the midstream of Nanshi River subwatershed. The results indicated the probability model posited can be used not only to predict total landslide volume in watershed scale, but also to determine the possible locations of the slope units with large landslide volume.
The prediction and advanced warning of landslide hazards in large-scale areas must deal with a large amount of uncertainty, therefore a growing number of studies are using stochastic models to ...analyze the probability of landslide occurrences. In this study, we used a modified Thiessen’s polygon method to divide the research area into several rain gauge control areas, and divided the control areas into slope units reflecting the topographic characteristics to enhance the spatial resolution of a landslide probability model. We used a 2000–2015 long-term landslide inventory, daily rainfall, and effective accumulated rainfall to estimate the rainfall threshold that can trigger landslides. We then employed a Poisson probability model and historical rainfall data from 1987 to 2016 to calculate the exceedance probability that rainfall events will exceed the threshold value. We calculated the number of landslides occurring from the events when rainfall exceeds the threshold value in the slope units to estimate the probability that a landslide will occur in this situation. Lastly, we employed the concept of conditional probability by multiplying this probability with the exceedance probability of rainfall events exceeding the threshold value, which yielded the probability that a landslide will occur in each slope unit for one year. The results indicated the slope units with high probability that at least one rainfall event will exceed the threshold value at the same time that one landslide will occur within any one year are largely located in the southwestern part of the Taipei Water Source Domain, and the highest probability is 0.26. These slope units are located in parts of the study area with relatively weak lithology, high elevations, and steep slopes. Compared with probability models based solely on landslide inventories, our proposed landslide probability model, combined with a long-term landslide inventory and rainfall factors, can avoid problems resulting from an incomplete landslide inventory, and can also be used to estimate landslide occurrence probability based on future potential changes in rainfall.
The predicted climate change is likely to cause extreme storm events and, subsequently, catastrophic disasters, including soil erosion, debris and landslide formation, loss of life, etc. In the ...decade from 1976, natural disasters affected less than a billion lives. These numbers have surged in the last decade alone. It is said that natural disasters have affected over 3 billion lives, killed on average 750,000 people, and cost more than 600 billion US dollars. Of these numbers, a greater proportion are due to sediment-related disasters, and these numbers are an indication of the amount of work still to be done in the field of soil erosion, conservation, and landslides. Scientists, engineers, and planners are all under immense pressure to develop and improve existing scientific tools to model erosion and landslides and, in the process, better conserve the soil. Therefore, the purpose of this Special Issue is to improve our knowledge on the processes and mechanics of soil erosion and landslides. In turn, these will be crucial in developing the right tools and models for soil and water conservation, disaster mitigation, and early warning systems.
Landslides are a widespread, frequent, and costly hazard in Seattle and the Puget Sound area of Washington State, USA. Shallow earth slides triggered by heavy rainfall are the most common type of ...landslide in the area; many transform into debris flows and cause significant property damage or disrupt transportation. Large rotational and translational slides, though less common, also cause serious property damage. The hundreds of landslides that occurred during the winters of 1995-96 and 1996-97 stimulated renewed interest by Puget Sound communities in identifying landslide-prone areas and taking actions to reduce future landslide losses. Informal partnerships between the U.S. Geological Survey (USGS), the City of Seattle, and private consultants are focusing on the problem of identifying and mapping areas of landslide hazard as well as characterizing temporal aspects of the hazard. We have developed GIS-based methods to map the probability of landslide occurrence as well as empirical rainfall thresholds and physically based methods to forecast times of landslide occurrence. Our methods for mapping landslide hazard zones began with field studies and physically based models to assess relative slope stability, including the effects of material properties, seasonal groundwater levels, and rainfall infiltration. We have analyzed the correlation between historic landslide occurrence and relative slope stability to map the degree of landslide hazard. The City of Seattle is using results of the USGS studies in storm preparedness planning for emergency access and response, planning for development or redevelopment of hillsides, and municipal facility planning and prioritization. Methods we have developed could be applied elsewhere to suit local needs and available data.
PDF
