Development and Application of a Real-Time GIS Digital Platform for Landslide Risk Analysis

Climate change is rapidly altering our planet, leading to an alarming increase in the occurrence and severity of various disasters [1-4]. One of the most concerning consequences is the escalating frequency and intensity of landslides, which inflict significant damage on lives and properties worldwide [5-9]. In Thailand, over the past decade, heavy rains have become more frequent, resulting in devastating landslides. These destructive events involve the forceful descent of rocks, soil, and debris from higher elevations, causing severe damage to house, buildings, and public infrastructure, such as roads and bridges [10-12]. As a result, local communities endure immense suffering and bear the brunt of the aftermath [10-12]. Each landslide leads to human casualties, destroys homes and structures under the weight of debris, and devastates agricultural crops. Furthermore, the psychological impact on the affected individuals cannot be overlooked. The psychological consequences of landslides exhibit varying durations, ranging from short-term to long-term, contingent upon the scale and intensity of the impact, notably its influence on property. Recovery of mental well-being tends to occur relatively swiftly once the situation is addressed, in contrast to the prolonged recovery periods associated with physical health issues or loss of life. Consequently, the presence of a digital platform capable of offering supplementary information equips government agencies to enhance their problem-solving efficiency [13-15]. Currently, the monitoring and surveillance of landslides in Thailand rely on data provided by the Department of Mineral Resources. These data are typically presented in the form of static maps, illustrating the level of landslide risk in different areas [16-18]. While these maps serve the purpose of monitoring susceptible regions, they suffer from limitations. They lack real-time data analysis and presentation capabilities, and there is a lack of applications to promptly notify individuals of landslide-prone areas. This deficiency hinders the utilization of such information as input data for enhancing the accuracy of risk analysis models.

The analysis of landslide risk serves a critical function in proactive monitoring and early warning systems designed to mitigate the impact of landslides. In the absence of such analysis or in the event of inaccuracies within the analysis results, individuals and communities may find themselves ill-prepared to respond to unfolding situations. For instance, in the southernmost region of Thailand, a dearth of timely landslide information has led to fatalities, tens of thousands of missing and affected individuals, hundreds of damaged homes, and nearly a hundred compromised roadways. Implementation of landslide occurrence analysis coupled with a digital platform for public information dissemination has the potential to significantly reduce or alleviate the distress and losses incurred in such situations.

A thorough examination of existing literature and relevant research on landslide risk analysis utilizing Geographic Information System (GIS) reveals that various factors influencing landslides have been considered in the analysis [19-25]. These factors encompass rainfall, slope, aspect, curvature, soil composition, Land Use and Land Cover (LULC), distance from drainage systems, lithology, and proximity to roads, among others. Researchers have employed diverse methods and techniques, including statistical techniques and machine learning algorithms like Frequency Ratio (FR), Weights of Evidence (WOF), Analytical Hierarchy Process (AHP), Weight Linear Combination (WLC), Fuzzy logic, Random Forest (RF), and Artificial Neural Network (ANN) [19-25]. The analysis results have been presented through risk maps, illustrating the level of landslide risk. Remarkably, previous research highlights the high accuracy of these analysis results, facilitating effective landslide monitoring in specific study areas. However, there are limitations to the current methods and research. One limitation is the static nature of data analysis and presentation, which makes information accessibility challenging. Additionally, visually interpreting static maps makes it difficult to determine the proximity of risk areas to residential locations. To overcome these limitations, this research proposes a novel approach: integrating GIS, Global Positioning System (GPS), and Remote Sensing (RS). Furthermore, a digital platform will be developed to notify and display landslide data in each area, providing real-time data presentation. This integrated solution aims to address existing limitations, facilitating efficient landslide monitoring, and enabling comprehensive situational awareness for effective management across all sectors.

The research area of this study focuses on Chumphon Province, situated in the upper southern region of Thailand. This region frequently experiences landslides due to the influence of the southwest monsoon and northeast monsoon, resulting in heavy rainfall and flash floods almost every year. Consequently, landslides occur in various areas of the province.

To address this issue, the research presents a comprehensive landslide risk analysis methodology that integrates Geographic Information System (GIS), Global Positioning System (GPS), and Remote Sensing (RS). Spatial data, encompassing factors that contribute to landslide occurrences, are utilized. Additionally, a digital platform is developed to provide real-time notifications and spatial visualization of landslide-prone areas. The methodology consists of seven distinct steps, as depicted in Figure 1, with detailed explanations provided for each step in subsequent sections of the research.

In the context of this research, the criteria employed for categorizing each factor entail a systematic methodological approach. These factors are categorized into discrete levels using two distinct formats: firstly, the utilization of the Natural Breaks classification method is applied when dealing with quantitative data, exemplified by variables such as Rainfall. Secondly, for factors necessitating classification into categorical groups, such as Land Use/Land Cover (LULC), Geology, and Slope, a categorization scheme is employed.

3.1 Data preparation

This step involves the collection of data from relevant agencies, comprising both attribute data and spatial data. The data are obtained from various sources, some stored on the cloud while others are not. The research incorporates multiple datasets, including rainfall, slope, elevation, land use and land cover (LULC), normalized difference vegetation index (NDVI), distance from rivers, soil depth, and rock formation data. Further details and the sources of these datasets are provided in Table 1 for reference.

Table 1. Data and data sources

3.2 Data and image correction

The spatial data used in this research underwent a process of geometric correction to ensure accurate alignment with the correct coordinate system. Additionally, a method of geometric correction involving interpolation or resampling techniques was applied to adjust the resolution of each data type before conducting the comprehensive analysis. This approach aimed to enhance the consistency and reliability of the spatial data, enabling more accurate and meaningful results in the subsequent analysis.

3.3 Data query

Data querying involves the utilization of modified data, which includes retrieving data from cloud storage and storing it in a spatial database for the purpose of conducting landslide risk analysis. Furthermore, this data serves the purpose of managing and displaying spatial data via the developed applications.

3.4 Factor analysis and landslide risk analysis

In this research, the landslide risk analysis utilizes the Analytical Hierarchy Process (AHP) method to prioritize and compare the factors influencing landslide occurrence. The AHP method determines the order of importance for each factor, which is subsequently used in the Weighted Linear Combination (WLC) method for analysis. The analysis involves assigning weights and assessing the significance of each relevant factor. After obtaining the weights or importance values for each factor, the risk is evaluated. Eq. (1) illustrates the equation utilized in the analysis.

$\mathrm{W}_{\mathrm{t}}=\left(\mathrm{M}_1 \mathrm{~W}_1\right)+\left(\mathrm{M}_2 \mathrm{~W}_2\right)+\left(\mathrm{M}_3 \mathrm{~W}_3\right)+\ldots+\left(\mathrm{M}_n \mathrm{~W}_{\mathrm{n}}\right)$             （1）

where, W_t is the level of risk of landslides.

M₁, M₁, M₃, …. M_n is the score of factors 1, 2, 3 … to n.

W₁, W₂, W₃, … W_n is the weight of factors 1, 2, 3 … to n.

1.png

Figure 1. Purposed methodology

In this research, the risk analysis of landslides was conducted using eight factors, which are as follows:

Precipitation is a key contributing factor to landslide occurrences. When rainfall exceeds normal levels, it saturates the soil in the affected region. The increased water content in the soil adds weight and reduces its stability, leading to potential landslides. In this research, rainfall data were collected from eight monitoring stations operated by the Meteorological Department in Chumphon Province. To analyze the spatial distribution of precipitation, the Inverse Distance Weighting (IDW) technique was applied to interpolate the data. The interpolated rainfall data were then categorized into five distinct levels. Figure 2 illustrates the precipitation data and their distribution across the study area.

2.png

Figure 2. Precipitation map

3.png

Figure 3. Elevation map

Elevation represents the vertical distance measured relative to the mean sea level. Regions characterized by substantial height variations compared to their neighboring areas during periods of increased rainfall or flash floods are at a heightened risk of experiencing landslides. The elevation map, depicted in Figure 3, provides a visual representation of the stratification of elevation data into four distinct levels in the context of this research.

Slope constitutes a critical factor due to the inherent hazard in regions characterized by steep gradients, as there exists a potential for the displacement of high water-holding soils towards flatter or lower-lying terrain. Moreover, a correlation is expected to manifest between slope characteristics and LULC patterns within the respective area. Figure 4 presents the slope map, which is obtained through the computation of elevation values and subsequent stratification of the data layer into eight distinct levels.

LULC in areas characterized by natural vegetation exhibits a lower susceptibility to landslides due to the presence of enhanced soil cover and a well-developed root system compared to other land types. Conversely, agricultural practices on hillside slopes, as well as deforestation and vegetation removal, have led to rapid alterations in the hydrological flow patterns across the soil. The LULC map, depicted in Figure 5, illustrates the classification of LULC into four distinct classes in this research: Agriculture (A), Forest (F), Urban (U), and Water (W).

4.png

Figure 4. Slope

Soil depth plays a significant role in influencing water drainage and subsequently affects landslide dynamics, as different soil types exhibit varying characteristics and drainage capabilities. In this study, it was observed that the research area can be classified into four distinct groups based on soil depth: shallow soil, relatively shallow soil, intermediate soil, and deep soil. An illustration of the soil depth distribution, exemplified by shallow outcrops and deep soils, is presented in Figure 6.

5.png

Figure 5. LULC map

6.png

Figure 6. Soil depth map

The distance from rivers plays a crucial role in determining the susceptibility of an area to flash floods and landslides. Regions with waterways running through steep slopes are more prone to these hazards compared to areas located farther away from rivers. Figure 7 displays an example of the distance from rivers map, which visualizes the spatial distribution of distances between the study area and nearby river networks.

7.png

Figure 7. River distance map

8.png

Figure 8. Geology map

Geological formations exhibit distinct characteristics that influence their decay rates, with certain species demonstrating a higher propensity for decay. These variations play a crucial role in shaping landslide occurrences. Figure 8 provides concrete examples of stone characteristics, thereby reinforcing the correlation between geological attributes and landslide susceptibility.

In this research, the risk of landslides was assessed utilizing the Weighted Linear Combination (WLC) technique. The analysis incorporated a scoring criterion derived from the grading of each factor data (Rating), in conjunction with the factor weights obtained from experts using the Analytic Hierarchy Process (AHP) technique. The specific criteria and their details are provided in Table 1. The landslides were classified into three risk classes: low-risk, moderate-risk, and high-risk.

3.5 Accuracy assessment

The evaluation of the accuracy of the landslide risk analysis conducted in this research involved assessing the calculation accuracy. The obtained data were utilized to compare the results with actual landslide occurrences in the research area, sourced from the Department of Mineral Resources.

The evaluation of accuracy was conducted through the computation of pertinent metrics including Accuracy, Precision, Recall, and F-measure values. The corresponding equations for these metrics are detailed in Table 2.

In instances where the outcomes of the analysis are found to be inaccurate, it is imperative to scrutinize the input data specific to the affected region. This examination aims to identify and analyze the underlying factors contributing to the inaccuracies, thereby facilitating the refinement of the model.

Table 2. Accuracy assessment equations

3.6 Data visualization

The utilization of spatial data analysis generates map-based outputs that facilitate the monitoring of landslides within the study area. These maps serve as essential tools for visualizing and tracking landslide occurrences. Moreover, the integration of these maps into a developed digital platform enables the effective dissemination of information to the public.

3.7 Digital platform development

This research involved the development of a digital platform dedicated to the collection and visualization of landslide information within each area. The platform was implemented using PHP and JavaScript programming languages, with the utilization of Leaflet for map display and GPS location retrieval from users' mobile devices. To ensure efficient data management, a MySQL database was employed to store the collected information.

This study encompasses an in-depth analysis of landslide risk, complemented by the development of a digital platform that leverages GIS, RS, and GPS technologies to facilitate the real-time processing and visualization of data. The integration of these technologies offers the potential to mitigate the constraints encountered in previous research endeavours [19-25].

When comparing the results of this research with previous studies in the field, both similarities and differences were observed. Similarities were found in certain aspects, indicating consistency in the findings. For instance, the use of spatial data factors in landslide risk analysis demonstrated an accuracy of more than 80% in identifying and classifying the risk levels of landslides [19-25]. However, previous studies [19-25] were limited in their practical application due to the reliance on static maps without a digital platform for managing spatial data. Consequently, the information presented in those studies lacked real-time updates and failed to reflect the dynamic nature of landslide risks. In contrast, this research addressed these limitations by integrating landslide risk analysis with the development of a digital platform. This integration enabled the notification and display of landslide-related information in a comprehensive manner. The digital platform provided real-time updates, facilitating effective management and surveillance of landslides. By visualizing and managing the data, the platform enhanced the accuracy and usability of the information for decision-making processes. Overall, the findings of this research not only align with previous studies in terms of the effectiveness of spatial data factors in landslide risk analysis but also contribute significantly by introducing a digital platform that overcomes the limitations of static maps. The integration of analysis and visualization within the platform enhances the applicability and usefulness of the results for comprehensive landslide management.

Considering disparities in research methodologies and their application compared to previous investigations [19-25], this study is distinguished by its integration of landslide data analysis with the development of a digital platform. This platform serves the dual purpose of monitoring landslide occurrences and contributing to mitigation, and restoration efforts in landslide-affected areas. Moreover, it holds the potential to facilitate data management for government agencies, aiding in the planning and management of landslide disasters.

The real-time data tracking system employed in this study does not divulge personal information to the general users. Rather, it exclusively visualized the geographic coordinates of landslide occurrences, facilitating users or the public in monitoring specific risk points. Furthermore, the digital platform developed in this research possesses transferable applicability to other geographic areas due to the programmed coordinate retrieval mechanism. This functionality is achieved through an Application Programming Interface (API), which extracts coordinate values from the GPS module within the user's smartphone and subsequently processes and displays the data utilizing the same methodology.

The findings of this research indicate several advantages for the study area across various domains. The development of a digital platform has conferred the capability for diverse user groups, including citizens, local government entities, and disaster prevention and relief authorities, to engage in real-time monitoring, and reporting of landslide incidents. Furthermore, this platform empowers stakeholders to harness the acquired information for purposes such as strategic planning, budget allocation, and resource procurement aimed at mitigating the ramifications of landslide events. This innovation has effectively ameliorated the limitations of previous methodologies, which necessitated labour-intensive field surveys, resulting in a reduction of both time and budgetary requirements for operational activities.