Mapping Debris Flow Susceptibility Using Frequency Ratio Model in Kundasang, Sabah, Malaysia

Debris flows, characterized by their rapid movement and potential for extensive destruction, are among the most severe geological hazards, impacting both human settlements and natural environments [1-3]. Globally documented events have highlighted their capacity to disrupt infrastructure, livelihoods, and lives [4, 5]. The threat is particularly significant in regions prone to seismic activity, where the interaction between debris flows and other geomorphic processes increases the risk and severity of these events [6, 7]. Effective mitigation and land-use planning require a thorough understanding of the factors driving debris flow occurrence, behavior, and runout patterns [8].

In the context of the Crocker Range in Sabah, Malaysia, the interplay between geological hazards and human activities is notably complex [2, 9]. This area has been subject to numerous debris flow incidents, resulting in road closures, economic losses, and distress within affected communities [8, 10]. Inadequate land-use practices, including traditional farming and deforestation, have significantly heightened the susceptibility to debris flows, emphasizing the urgent need for a detailed vulnerability assessment of the region's terrain.

Despite existing research on debris flows in the Crocker Range, there is a noticeable gap in studies that integrate susceptibility mapping with practical implications for community planning and policy-making. Previous studies have largely focused on isolated factors without providing a comprehensive model that could guide land-use and infrastructure development in high-risk areas. Our study addresses this gap by applying the bivariate Frequency Ratio Model to assess debris flow susceptibility in the Kundasang area of Ranau, Sabah. By integrating terrain parameters with Geographic Information System (GIS) tools, this study not only contributes to the academic discourse but also offers actionable insights for local authorities and policymakers. The findings will support the development of targeted strategies to mitigate debris flow impacts, ensuring that infrastructure and community resilience are enhanced. In doing so, this research aims to bridge the divide between theoretical knowledge and practical application, providing a robust foundation for hazard mitigation and adaptation strategies in regions facing similar challenges.

The assessment of debris flow susceptibility involves utilising the Frequency Ratio (FR) model within a Geographic Information System (GIS) framework (Figure 3). The initiation of this analysis encompasses delineating the debris flow-prone regions using Google Earth Pro. The pivotal dataset for model construction is the Digital Elevation Model (DEM), with a consistent cell size of 30 meters adopted for evaluating all the parameters under consideration. To ensure comprehensive coverage of the primary debris flow region, Kundasang, the DEM is similarly employed to extract the basin.

3.png

Figure 3. Flowchart of analysing debris flow susceptibility map using frequency ratio model

An extensive review of the pertinent literature underpins the selection of parameters within the FR model. This proactive approach aids in refining the understanding of the contributing factors behind debris flow incidents. Employing ArcGIS version 10.5, diverse parameters are generated and integrated into the FR model. These encompass a spectrum of elements, including slope aspect, slope curvature, elevation, land use, soil association, slope angle (°), Steam Power Index (SPI), stream proximity, and Topography Wetness Index (TWI).

In the classification process, each of the causal factors within the FR model is meticulously treated to ensure an accurate representation of their contributions to debris flow susceptibility. Factors such as slope aspect, slope curvature, elevation, land use, soil association, SPI, and TWI are reclassified employing the Natural Break (Jenks) method. The Jenks method was selected because it optimally partitions data into classes by minimizing the variance within each class and maximizing the variance between classes. This is particularly effective for parameters with non-uniform distributions, such as elevation and slope curvature, where natural groupings or breaks are inherent. By applying this method, the classification reflects the natural structure of the data, ensuring that each class represents a distinct range of values with significant impact on debris flow susceptibility.

In contrast, slope angle (°) and stream proximity are classified through manual interval delineation. This robust methodology ensures the alignment of parameter classifications with the nuances of the landscape and facilitates the accurate interpretation of their roles in the debris flow susceptibility assessment.

3.1 Debris flow conditioning factors

Aspect, often called slope aspect, denotes the direction of the steepest slope on a terrain surface. Some researchers emphasise its significance as a pivotal factor in landslide susceptibility studies [17-19]. Aspect exerts influence over various elements, including the orientation of controlling discontinuities in landslides, precipitation patterns, wind impact, and solar exposure [3, 20]. The reclassification of aspect regions involves partitioning them into distinct classes: flat (-1), north (0°-22.5°), northeast (22.5°-67.5°), east (67.5°-112.5°), southeast (112.5°-157.5°), south (157.5°-202.5°), southwest (202.5°-247.5°), west (247.5°-292.5°), northwest (292.5°-337.5°), and north (337.5°-360°).

Curvature unveils the topographic morphology [21, 22]. Negative curvature signifies concave features, zero curvature corresponds to flat terrain, and positive curvature characterises convex topography [23]. These positive and negative values directly correlate with the range of flow speeds, representing slow to rapid movement. In particular, basin-shaped slopes, commonly observed in landslides, play a significant role [17]. Valleys, characterised by concave slopes, tend to retain groundwater, surface water, sediment, and organic matter, resulting in overhanging soils.

Elevation, a pivotal attribute shaped by diverse geologic and geomorphological processes, plays a crucial role [24]. Research has indicated that rainfall and debris flow exhibit heightened occurrence at higher elevations [3]. Utilising DEM data within ArcGIS, the elevation map was created and subsequently categorised into seven classes: <1000m, 1000m -1500m, 1500m-2000m, 2000m-2500m, 2500m-3000m, 3000m-3500m, and >3500m.

Land use significantly influences infiltration rates, evapotranspiration, and the formation of surface runoff, exerting both direct and indirect effects. The regional soil type heterogeneity influences infiltration, which, in turn, shapes overland flows [25]. The land use map, derived from the Sabah Forest Department (1988) and traced as a shapefile in ArcGIS, consists of four primary classes: settlement, scrub, dryland primary forest, and dryland secondary forest.

In landslide susceptibility modelling, slope angle is a pivotal parameter frequently associated with debris flow [26]. The slope angle governs surface runoff [10, 27]. Near the equator, Malaysia experiences a tropical climate with high temperatures and abundant yearly rainfall. Researchers studying slope failure have identified rainfall as one of the prominent triggers for slope failures [17]. Based on the slope classification established by the Department of Mineral and Geoscience Malaysia (JMG, 2006), the slope map is classified into five categories: Flat (<5°), Gentle (5°-15°), Moderate (15°-25°), Steep (25°-35°), and Very Steep (>35°).

Certain soil attributes, including particle size distribution, shape, and pore sizes, play a role in determining debris flow susceptibility. Soil quality impacts water infiltration, interflow and baseflow velocity, and water retention capacity [28]. With their larger surface area, fine-textured soils such as clay and silt tend to retain more water, particularly when unsaturated [29]. The soil association map, created based on the Agriculture Department of Sabah (JPNS, 1976) and traced as a shapefile in ArcGIS, comprises five soil association types: Trusmadi, Crocker, Pinousuk, Bidu-Bidu, and Kinabalu.

The Stream Power Index (SPI) serves to identify potential river erosion and landscape phenomena [17, 23]. As the catchment area and gradient increase, upslope water contribution and flow velocity also rise [17]. The SPI map is classified into five classes: <-8, -8 – -5, -5 – 0, 0 – 3, and 3 – 9.

Proximity to rivers reflects the degree of fluvial erosion and exacerbates slope instability [23]. Streams are crucial in conveying water during and after precipitation, often serving as a basis for debris flow mobility. Areas near rivers with steep slopes experience considerable erosion, amplifying the risk of landslides during intense rainfall and sloping topography [8]. The map incorporates five buffer classes: 100m, 200m, 300m, 400m, and above 500m.

The Topographic Wetness Index (TWI) is widely adopted to characterise the influence of topography on saturated runoff areas' location and size [17]. This index measures soil moisture content dictated by topography. The map categorises TWI into five classes: <5, 5 – 6, 6 – 8, 8 – 11, and 11 – 19.

3.2 Frequency ratio

The methodology employed for debris flow susceptibility mapping in this study relies on the established frequency ratio (Fr) model (Eq. (1)), a widely recognized statistical approach. This model facilitates the integration of various environmental factors as predictors, contributing to the dependent variable. In this context, Fr values above 1 signify a more robust correlation, while values below 1 indicate a weaker correlation [28, 30, 31]. The debris flow susceptibility index is derived by summing the frequency ratio values of individual factors for each pixel, quantifying the likelihood of debris flow occurrence [23].

Consequently, a higher index value corresponds to an elevated susceptibility to debris flow, whereas a lower index value signifies reduced susceptibility [30]. To categorize these indices, the calculated frequency ratio values undergo classification using both Jenks natural breaks and manual classification. This assigns five distinct classes: very low, low, moderate, high, and high susceptibility categories [32]. This refined classification enhances the accuracy of the susceptibility assessment and provides a more actionable basis for mitigation strategies.

$F_r=\frac{\frac{N_i}{N}}{\frac{S_i}{S}}$                    (1)

where,

N_i: No. of pixels in which the debris flow occurred in the i variable

N: No. of pixels with debris flow occurrence

S_i: No. of pixels of the i variable

S: Total number of pixels

The performance of the Frequency Ratio (FR) model was assessed using the Receiver Operating Characteristic (ROC) curve and the corresponding Area Under the Curve (AUC) metric. The ROC curve plots the true positive rate (sensitivity) against the false positive rate (1-specificity), offering a graphical representation of the model’s ability to discriminate between different susceptibility classes across various threshold values.

Debris flows pose significant threats to human settlements and the natural environment, particularly in regions prone to seismic activity and geomorphic processes. This study focused on the Kundasang area in Sabah, Malaysia, to comprehensively assess debris flow susceptibility and provide insights for effective hazard mitigation strategies. By integrating a bivariate Frequency Ratio (FR) model with Geographic Information System (GIS) techniques, we analyzed various terrain parameters, including slope aspect, curvature, elevation, land use, slope angle, soil association, Stream Power Index (SPI), stream proximity, and Topographic Wetness Index (TWI), to understand the factors influencing debris flow occurrences.

Our analysis revealed that parameters such as slope aspect, curvature, elevation, and land use significantly influence debris flow susceptibility. Specifically, concave terrains and north-facing slopes, elevations between 2500m and 3000m, and areas dominated by dryland primary forest exhibit higher susceptibility. Additionally, steep slope angles and specific soil associations, like the Kinabalu soil type, further increase susceptibility. The SPI and proximity to rivers highlighted areas at greater risk, particularly those within 100 meters of rivers, while higher TWI values indicated increased susceptibility due to elevated soil water content.

The comprehensive debris flow susceptibility map generated for the study area identifies varying levels of risk, with the southern region, including populated communities, particularly vulnerable to debris flow hazards. These findings highlight the complex interplay of geomorphic, climatic, and anthropogenic factors contributing to the heightened susceptibility observed in the Kundasang area.

The methodology and findings from this study can be applied to other regions with similar topographical and environmental conditions, such as areas with steep terrain and active seismic activity, to develop effective mitigation and land-use strategies tailored to local conditions. Based on these findings, we recommend that land-use planning in high-susceptibility areas incorporate zoning regulations that limit development on steep slopes and near riverbanks. Establishing buffer zones around high-risk areas and enforcing construction restrictions can reduce the risk of property damage and loss of life. Sustainable forest management practices, including reforestation and controlled land clearing, should be prioritized to stabilize slopes and reduce soil erosion. Infrastructure development should be carefully planned, considering the terrain’s vulnerabilities, to avoid exacerbating debris flow susceptibility. These insights will empower decision-makers, researchers, and local communities to implement proactive measures that safeguard lives, infrastructure, and ecosystems against the persistent threat of debris flows.