1. Introduction
Geological disasters refer to adverse geological processes or phenomena caused by natural or human factors. Geological disasters can be divided into more than 30 types. Natural geological disasters are induced by rainfall, snowmelt, and earthquakes, while artificial geological disasters are caused by engineering excavation, loading, and blasting. Common geological disasters mainly include six types of disasters related to geological processes, including collapse, landslide, mud-rock flow, subsidence, ground fissures, and land subsidence (
Figure 1). Geological disasters have the characteristics of being sudden, uncontrollable, and highly destructive, causing harm to people’s lives and property safety [
1]. Especially in recent years, extreme weather has increased, and excessive rainfall is more likely to trigger geological disasters [
2,
3]. In 2022, there were 5659 geological disasters in China, mainly landslides and collapses (
Figure 2 and
Figure 3). Conducting geological hazard risk assessment, analyzing the occurrence patterns of regional geological hazards, mastering the risk and hidden dangers of critical geological hazards, and predicting and identifying the occurrence of geological hazards are necessary means of effectively resolving disaster risks [
4].
The research on geological hazards and risk assessment is in the development stage. It can be divided into three categories: qualitative evaluation, quantitative evaluation, and a combination of qualitative and quantitative reviews. Qualitative evaluation methods include the Analytic Hierarchy Process (AHP) [
5,
6] and the Comprehensive Indicator Method (CIM) [
7,
8]. The AHP is a multi-objective decision analysis method that quantifies the empirical judgments of decision-makers. It is widely used for analysis and decision-making when the target structure is complex and needs more necessary data. It can obtain a satisfactory decision structure, especially for complex problems that are difficult to quantify fully. The CIM refers to comprehensively evaluating benefits by weighting the average number of individual benefit indicators and calculating the comprehensive value based on a reasonable set of benefit indicator systems. Commonly, these methods heavily rely on the expertise of individuals, which is subjective and susceptible to human influence. On the other hand, quantitative evaluation methods are grounded in data and offer a more objective approach to inferring the likelihood of geological disasters. Some prominent examples include logistic regression analysis [
9,
10], neural network methods [
11,
12], random forest methods [
13,
14], and the information value model method [
15,
16]. Logistic regression analysis, while unaffected by subjective factors, can introduce uncertainty in areas with dense vegetation. The neural network method exhibits substantial capabilities in addressing complex issues involving incomplete or insufficient data but suffers from limitations in sample selection and iterative processes. The random forest method demonstrates high predictive ability but necessitates extensive geological disaster data for reliable results, making it less suitable for regions with fewer incidents. The information value model method calculates the contribution of each factor through the information value of known geological disaster points and factors, establishing a prediction model. Compared to other methods, it requires less data. Still, it can only reflect the likelihood of geological disasters occurring under specific combinations of influencing factors and does not account for variations in the impact levels of each factor [
17]. Hence, a combination of quantitative and qualitative evaluation proves more beneficial in enhancing the accuracy of geological disaster assessment results.
With the rapid development of GIS and machine learning technology (ML), the means and methods for risk assessment of geological disasters are also maturing. Based on remote sensing and GIS, Tan et al. [
18] introduced a hierarchical entropy variable weight method using an entropy algorithm to reduce personal impact and obtain more accurate evaluation results. Lyu and Yin [
19] merged AHP and the analytical network process (ANP) into a geographic information system and integrated interval numbers into the fuzzy hierarchical analysis process (FAHP) to evaluate the risks of various disasters in Hong Kong, improving the accuracy of the multi-risk assessment. Yang et al. [
20] proposed an improved coupling landslide susceptibility evaluation model by combining the theory of unascertained measures (UM), dynamic, comprehensive weighting (DCW) based on the AHP entropy weight method and set pair analysis (SPA) theory. Chen and Zhang [
21] conducted a comparative study on GIS-based Bayesian networks (BN), Hoeffding trees (HT), and logistic model trees for landslide susceptibility modeling, demonstrating that the HT model is a good classifier for landslide susceptibility modeling. Rong et al. [
22], based on the integrated machine learning model (MLM) and scenario simulation technology, calculate the precipitation in different extreme precipitation return periods and evaluate the landslide risk with the susceptibility results. It is found that the optimized Random Forest model has the best all-around performance in sensitivity evaluation.
In 2020, the State Council of China launched the first comprehensive survey of natural disaster risks (regarding geological disasters), which shifted its work philosophy from focusing on post-disaster relief to pre-disaster prevention and from reducing disaster losses to reducing disaster risks. After two years of effort, by the end of 2022, 2041 counties completed geological hazard risk surveys, and 1522 counties completed 1:50,000 geological hazard risk surveys. The first survey task has been fully completed as scheduled. The system has systematically conducted the comprehensive remote sensing identification of geological hazards in 713 counties (cities and districts) with an area of 4.07 million Square kilometers that is prone to geological disasters in China, completed the fine survey of 2161 cities and towns, 6615 important hazard surveys, 6250 engineering treatments, 2676 hazard elimination and removal, and relocated 125,000 people from 34,000 households threatened by geological disasters. In addition, more than 20,000 universal professional monitoring projects (rainfall monitoring, slope displacement and crack monitoring, groundwater level monitoring, video monitoring, etc.) were completed and operated before the flood season, and the inspection and monitoring system for more than 264,000 geological disaster group measurement and prevention personnel was improved. More than 200 ministerial-level experts are stationed in 30 provinces nationwide to strengthen risk prevention technical support. The integration of “civil air defense” and “technical prevention” was further enhanced, and geological disaster investigation and evaluation, monitoring and early warning, comprehensive management, emergency response, and grassroots disaster prevention capabilities were further improved. This article is a follow-up study based on the results of this work.
Laoshan District is the most economically developed area in Qingdao, Shandong Province, China. How to effectively avoid the harm caused by geological disasters in urban planning is an urgent problem for government departments to solve. At the same time, there are many mountains in Laoshan District that are prone to collapse disasters under the conditions of rainstorms. This article introduces extreme rainfall factors based on conventional geological hazard risk assessment in response to this issue. Four extreme rainfall conditions have been set: 10-year, 20-year, 50-year, and 100-year return periods. Through the combination of AHP and IM methods, a complete geological hazard assessment, including susceptibility, hazard, vulnerability, and risk, was carried out in Laoshan District. Based on the evaluation results, prevention and control zones were divided, and corresponding prevention measures were proposed. The research results can provide authoritative disaster risk information and a scientific basis for decision-making for effective local natural disaster prevention and control work, proper regional land planning, and sustainable economic and social development. Finally, an automatic geological hazard monitoring and warning system is introduced, which can provide a reference for geological hazard prevention in similar rainstorm areas.