This paper proposes a statistical model for mapping global landslide susceptibility based on logistic regression. After investigating explanatory factors for landslides in the existing literature, five factors were selected for model landslide susceptibility: relative relief, extreme precipitation, lithology, ground motion and soil moisture. When building the model, 70 % of landslide and nonlandslide points were randomly selected for logistic regression, and the others were used for model validation. To evaluate the accuracy of predictive models, this paper adopts several criteria including a receiver operating characteristic (ROC) curve method. Logistic regression experiments found all five factors to be significant in explaining landslide occurrence on a global scale. During the modeling process, percentage correct in confusion matrix of landslide classification was approximately 80 % and the area under the curve (AUC) was nearly 0.87. During the validation process, the above statistics were about 81 % and 0.88, respectively. Such a result indicates that the model has strong robustness and stable performance. This model found that at a global scale, soil moisture can be dominant in the occurrence of landslides and topographic factor may be secondary.

Extreme precipitation-induced landslide events are projected to increase under climate change, which poses a serious threat to human lives and property. In this study, a global-scale landslide risk assessment model was established using global landslide data, by considering landslide hazard, exposure, and vulnerability. The global climate model data were then employed to drive the established global landslide risk model to explore the spatial and temporal variations in future landslide risk across the globe as a result of extreme precipitation changes. The results show that compared to the 30-year period from 1971 to 2000, the average annual frequency of landslides triggered by extreme precipitation is projected to increase by 7% and 10%, respectively, in the future 30-year periods of 2031–2060 and 2066–2095. The global average annual casualty risk of landslides is projected to increase from about 3240 to 7670 and 8380, respectively (with growth rates of 140% and 160%), during the 2031–2060 and 2066–2095 periods under the SSP2-4.5 scenario. The top 10 countries with the highest casualty risk of landslides are China, Afghanistan, India, the Philippines, Indonesia, Rwanda, Turkey, Nepal, Guatemala, and Brazil, 60% of which are located in Asia. The frequency and intensity of extreme precipitation will increase under climate change, which will lead to an increase in casualties from landslides in mountainous areas globally, and this risk should be taken seriously. The present study was an attempt to investigate and quantify the impact of global landslide casualty risk under climate change, which still has uncertainty in terms of outcomes, and there remains a need for further understanding in the future of the propagation of uncertainty between the factors that affect the risk.

We propose a new convolutional neural networks method in combination with ordinal regression aiming at assessing the degree of building damage caused by earthquakes with aerial imagery. The ordinal regression model and a deep learning algorithm are incorporated to make full use of the information to improve the accuracy of the assessment. A new loss function was introduced in this paper to combine convolutional neural networks and ordinal regression. Assessing the level of damage to buildings can be considered as equivalent to predicting the ordered labels of buildings to be assessed. In the existing research, the problem has usually been simplified as a problem of pure classification to be further studied and discussed, which ignores the ordinal relationship between different levels of damage, resulting in a waste of information. Data accumulated throughout history are used to build network models for assessing the level of damage, and models for assessing levels of damage to buildings based on deep learning are described in detail, including model construction, implementation methods, and the selection of hyperparameters, and verification is conducted by experiments. When categorizing the damage to buildings into four types, we apply the method proposed in this paper to aerial images acquired from the 2014 Ludian earthquake and achieve an overall accuracy of 77.39%; when categorizing damage to buildings into two types, the overall accuracy of the model is 93.95%, exceeding such values in similar types of theories and methods.