A New Parameter-Free Slope Unit Division Method That Integrates Terrain Factors

Abstract

With increasing research on geological hazards and the development of geographic information technology, slope units play an increasingly important role in landslide susceptibility assessment and prevention work. The scientific and reasonable division of slope units directly impacts the accuracy and practicality of analysis results. Despite the significant progress in slope unit division techniques, most existing methods still have certain limitations, such as a strong dependence on manually set thresholds during the division process, resulting in low levels of automation and efficiency. To address this issue, a new parameter-free slope unit extraction algorithm that integrates terrain factors, called Terrain Factor Parameter-Free Slope Unit Division (TFPF-SU), is introduced. This eliminates the issue of manually setting parameter thresholds during the slope unit division process. This algorithm fully utilizes the terrain information provided by digital elevation models (DEMs) to accurately calculate the curvature, slope, and aspect data for each point. On the basis of the inherent consistency principles among slope, aspect, and curvature, object-oriented image segmentation technology is used to achieve slope unit division. We select Dongchuan District in Yunnan Province, China, as a test area to verify the TFPF-SU algorithm and conduct a detailed comparative analysis and validation of the results with those obtained via traditional hydrological analysis methods from both qualitative and quantitative perspectives. In the quantitative analysis, we utilize the size and shape of the slope units. The results indicate the following: ① the slope units obtained with the TFPF-SU method are more uniform in size, avoiding issues with oversized or irregularly shaped units; ② the slope unit shapes obtained with the TFPF-SU method are more reasonable, with about 70% of the units falling within a reasonable shape index range, compared to only about 32% with the hydrological method; and ③ the slope units produced by the TFPF-SU method align more closely with terrain authenticity, exhibiting a higher degree of topographical conformity.

1. Introduction

Landslides, as major geological disasters, not only cause significant economic losses due to their suddenness and severity but also often pose immeasurable threats to human lives [1,2,3,4]. To mitigate the adverse impacts of landslides, conducting systematic regional landslide susceptibility analysis is crucial. This is a key step in identifying potential landslide risk areas in advance and formulating effective disaster reduction measures [5,6,7]. In landslide susceptibility analysis, the choice and delineation of evaluation units are core elements that influence the accuracy and practical value of the analysis results. Therefore, the selection and delineation of evaluation units are of paramount importance. Currently, the most used evaluation units include grid units [8], regional units [9], units with homogeneous conditions, geomorphological units, and slope units [10,11,12,13,14,15,16]. However, slope units have gradually gained favor in landslide hazard prediction and assessment because of their unique characteristics [17]. The delineation of slope units directly corresponds to actual topographic structures, and, thus, slope units intuitively reflect surface slope changes and their close association with landslide formation mechanisms. Therefore, in recent years, many scholars have used slope units for landslide hazard risk assessment, aiming to obtain more realistic and targeted risk distribution maps and susceptibility zonation results. This information can be used to effectively guide disaster prevention and mitigation efforts, thereby minimizing the negative impacts of landslides on society, the economy and the environment [18,19].

Early slope unit subdivision relied on manual operations, making the process cumbersome, error-prone, and highly subjective; thus, it is suitable only for small-scale applications. Currently, a widely used method for extracting slope units in geological hazard assessment research is the hydrological analysis-based [20,21] framework first proposed by Carrara et al. [14]. The core of this method is to treat slope units as unique topographic units defined by the interweaving of ridgelines and valley lines [22] based on a digital elevation model (DEM). In this process, DEM preprocessing, including depression filling, terrain reversal, and other operations, is typically performed first to accurately identify watersheds and drainage paths. Although this method is theoretically universal and practical, in actual applications, especially under complex topographic conditions, it may produce some small cross-sectional units that do not conform to actual geomorphological features, as well as overly elongated units in the flow direction [15,23,24,25]. These require extensive manual intervention and correction to improve accuracy and practicality. To avoid these issues, many scholars have made improvements to this method. For example, Lucian Drăguţ [26] divided units by classifying landform elements through image analysis, combining layers of elevation, profile curvature, plan curvature, and slope data. Yan et al. [15] proposed the curvature watershed method [27], in which the boundaries of convex and concave geomorphological elements are extracted using positive and negative curvatures to obtain slope units. However, this method is essentially a hydrological analysis method. Alvioli et al. [16] developed a professional software tool called r.slopeunits [28,29], integrated into the GRASS 6.4 GIS software, for delineating slope units. It uses advanced algorithms to automatically identify and extract drainage lines and ridgelines in the terrain and construct precise slope units based on these key topographic features. The method incorporates an adaptive algorithm that allows for the flexible and accurate delineation of slope units through an iterative process [30]. However, the effective application of this software relies on the user’s reasonable setting of internal parameters, which requires certain levels of geological and geomorphological knowledge and a geographic information system (GIS) technical background. Wang et al. [24] used morphological image analysis methods combined with logical algorithms to process DEM data, enabling the precise extraction of ridgelines and valley lines as a basis for slope unit delineation. Although this image analysis technique performs well in specific scenarios, especially in evaluating the stability of slopes formed by infinite boundaries or artificial excavations, its applicability is limited in cases with highly complex and variable mountainous terrains. Additionally, Huang et al. [25] proposed a multiscale segmentation-based slope unit delineation strategy. This method can be adapted to assess geomorphological changes at multiple scales by subdividing terrain data at different scales, promoting detailed and comprehensive slope unit delineation. However, the implementation of this method involves cumbersome input parameter calculations and requires high computational power and in-depth theoretical support, thus posing a challenge in practical cases.

In summary, diverse and distinct methods have been developed for the extraction and analysis of slope units. However, in the existing slope unit division techniques, some common problems and challenges are inevitably faced. A common challenge is that many methods require manual threshold setting during execution [10,31], where the choice of parameters is crucial in slope unit division. Even the slightest error in these parameters can lead to overly dense or sparse slope units, necessitating frequent and lengthy trials to determine suitable parameters. This greatly increases the subjectivity of the division process, potentially leading to unstable extraction results due to differences in individual experience. To address this issue, we use information obtained from a DEM (i.e., homogeneity constraints in terms of terrain). By avoiding manual parameter settings, the slope unit subdivision is automated. This new method eliminates the complex parameter setting process, significantly reducing subjectivity and uncertainty in operation while improving the stability and consistency of the subdivision results.

A novel and efficient slope unit extraction method that does not require manual parameter setting is proposed. We selected the complex terrain of Dongchuan, Yunnan, as the study area and used DEM data to perform slope unit division. To verify the effectiveness of the method, a comparison was made with the commonly used hydrological analysis approach. This new method aims to eliminate the need for manually setting parameter thresholds during slope unit division. It reduces the influence of subjective factors on the final results and improves the reliability and efficiency. This approach of slope unit division is particularly beneficial under different terrain conditions, especially in applications involving risk assessment and management in landslide-prone areas. By eliminating the cumbersome parameter setting process, the new method is expected to significantly improve the limitations of current slope unit extraction techniques. This advancement enhances geological disaster prediction, ecological protection, land use planning, and tasks in other fields.

2. Methods and Study Area

2.1. Study Area

Dongchuan District, under the jurisdiction of Kunming city, Yunnan Province, is located on the northern edge of the Yunnan–Guizhou Plateau (China). Its longitude ranges from 102°47′ E to 103°18′ E, and its latitude ranges from 25°57′ N to 26°32′ N, covering a total area of approximately 1858.79 km². Its specific location is shown in Figure 1. The region features high mountains and deep valleys, with dramatic differences in elevation. Influenced by the Xiaojiang fault zone, it is a seismic zone with fractured surface rocks and poor stability. Climatically, this area has a subtropical monsoon climate, contributing to a fragile geological environment. The area is rich in mineral resources, but excessive mining and cultivation have severely depleted forest resources. Owing to its unique geological and topographical features, coupled with long-term human activities, Dongchuan District frequently experiences natural disasters such as landslides, mudslides, and mountain collapses, which pose significant threats to ecological security and socioeconomic development.

2.2. Principle of the Method

In this paper, a DEM-based slope unit division scheme aimed at accurately dividing the surface into contiguous areas with similar slope, aspect, and curvature characteristics is proposed. This method achieves slope unit division on the basis of the principle of minimum regional heterogeneity. Unlike traditional hydrology-based subdivision methods, this approach relies more on terrain geometry to reflect natural slope variations more accurately. By combining a region-growing algorithm, starting from DEM grid pixels, it recursively expanded the neighborhood through region growing. This process assigned adjacent pixels with the same slope, aspect, and curvature categories into the same slope unit. Thus, it divided the entire study area into continuous, uniform, enclosed regions.

The region-growing algorithm [32,33] was first proposed by R. Adams and L. Bischof. Its basic theoretical framework lies in dividing an image by identifying and aggregating pixel units with similar properties. The algorithm requires one or more seed points to be specified as starting points. It then traverses the neighboring pixels of the seed points and compares them with the seed points, incorporating the pixels that exhibit high similarity among predetermined attributes into the current growing region. This process continues in the outward direction until no accessible pixels meet the established similarity criteria, thus completing region growth.

In this study, we used a dynamic seed-point selection mechanism. We adopted a pixel-by-pixel traversal method to identify and select unclassified pixels, also known as seed points, as the starting points for region growth. Therefore, every unassigned pixel has the potential to become a new seed point. This dynamic seed-point selection mechanism was implicitly implemented during the traversal process. A stacked structure was used to recursively label the eligible pixels and indicate that they are part of the same slope unit until all the eligible adjacent pixels were assigned, thus completing the slope unit division process. By dynamically selecting seed points and expanding neighborhoods based on terrain similarity, the slope unit subdivision becomes more flexible and detailed. This approach, combined with a recursive marking method using a stack structure, ensures unit continuity and completeness. This approach avoids both over-segmentation and under-segmentation in complex terrains. Ultimately, the entire study area was divided into relatively uniform slope units, each exhibiting high internal consistency in terms of slope, aspect, and curvature, providing a solid foundation for further geological and terrain analysis.

By fully taking into account the influence of terrain factors and incorporating a dynamic seed-point selection mechanism, this subdivision method demonstrates strong adaptability in handling complex terrains, especially in scenarios requiring a fine-grained analysis of topographic features. The slope units generated by this method more accurately reflect natural slope forms, eliminating the artificial transition zones and irregular units commonly seen in traditional methods, offering a more natural and authentic terrain unit subdivision solution.

2.3. Specific Process

The process of extracting slope units via the Terrain Factor Parameter-Free Slope Unit Division (TFPF-SU) method is shown in Figure 2.

First, to ensure the accuracy and reliability of the slope unit division, it is necessary to check for NoData values in the DEM data. Furthermore, we implemented rigorous data preprocessing on the original DEM data by applying Gaussian filtering to effectively reduce noise and smooth out anomalies. The goal was to eliminate nontopographic factors that might affect subsequent analysis results and to enhance data quality.

Second, using the processed DEM data with ArcGIS 10.7 software, we systematically calculated a series of key topographic parameters, such as slope, aspect, and curvature. These parameters comprehensively reflect the terrain undulations, surface directional characteristics, and local terrain curvature, providing a solid foundation for the precise division of slope units.

The calculated slope, aspect, and curvature were then classified to obtain discrete label matrices. Next, an array of the same size as the DEM was initialized to record the unit affiliation ID of each pixel. With this information, unclassified pixels were identified through pixel-by-pixel traversal and selected as potential region-growing starting points, or seed points. Seed point selection is dynamic, with every unassigned pixel having the potential to become a new seed point, ensuring coverage of the entire DEM area. Starting from a selected seed point, slope units were gradually constructed through neighborhood expansion. Neighborhood expansion follows the principle of topographic similarity, ensuring that pixels newly added to the slope unit share similar topographic features with the seed point on the basis of the classified label matrices of slope, aspect, and curvature. Using a stacked structure, eligible pixels were recursively labeled to indicate that they are associated with the same slope unit. A seed point was first added to the stack, and its neighboring pixels were checked. If a neighboring pixel meets the topographic similarity criteria and is unassigned, it is labeled as part of the current slope unit and added to the stack. This process was repeated until all eligible neighboring pixels were assigned, completing the division of one slope unit. Then, the process returned to the previous step to select the next unclassified pixel as a new seed point, repeating the region-growing and recursive labeling processes until all the pixels in the DEM area were assigned to the corresponding slope units. Through the above steps, the scientific division of slope units was achieved. Each slope unit exhibits a high degree of consistency in terms of slope trends, aspect distributions, and topographic curvature characteristics, ensuring the accuracy and reliability of the division results.

Finally, the entire result was optimized through postprocessing steps, such as splitting and merging, to generate the final slope units.

The implementation of this process relies on the Python 3.9 programming language and ArcGIS software. Furthermore, we validated the effectiveness of this method by conducting a comparative analysis with traditional hydrological analysis methods.

Aspect [34] is a fundamental topographic factor that describes the morphology of the Earth’s surface. It refers to the orientation of a slope. The calculation formula is Equation (1):

$$\theta =a\tan 2\left({\displaystyle \frac{\partial z}{\partial y}},-{\displaystyle \frac{\partial z}{\partial x}}\right)\times {\displaystyle \frac{180}{\pi }}$$

(1)

where ${\displaystyle \frac{\partial z}{\partial x}}$ represents the slope variation in the x-direction, ${\displaystyle \frac{\partial z}{\partial y}}$ represents the slope variation in the y-direction, and $\theta$ denotes the aspect.

The aspect significantly impacts the subdivision of slope units, primarily because it is related to variations in environmental factors such as solar radiation, the rainfall distribution, and temperature conditions. These differences directly affect the physical properties, hydrological processes, vegetation cover, and the geological stability of slopes, thereby influencing the risk level of landslides. Therefore, when subdividing slope units, we applied the eight-direction aspect classification [35], dividing the slopes into subunits oriented towards different directions. The aspect division diagram is shown in Figure 3.

The slope [36] gradient primarily affects the subdivision of slope units by directly influencing the processes of surface energy and material transfer, the soil mechanical state, the vegetation coverage capacity, and geological stability. The calculation formula is Equation (2):

$$\delta =\arctan \left(\sqrt{{\left({\displaystyle \frac{\partial z}{\partial x}}\right)}^2+{\left({\displaystyle \frac{\partial z}{\partial y}}\right)}^2}\right)\times {\displaystyle \frac{180}{\pi }}$$

(2)

where ${\displaystyle \frac{\partial z}{\partial x}}$ represents the rate of elevation change in the x-direction, ${\displaystyle \frac{\partial z}{\partial y}}$ represents the rate of elevation change in the y-direction, and $\delta$ denotes the slope.

Therefore, incorporating the slope gradient into the subdivision of slope units helps reveal the differential impact of variations in slope steepness on both the intrinsic properties and the external characteristics of slope units. This provides critical foundational information for refined geographic analysis and decision support. According to the “Operating Guidelines for Forest Resource Planning and Design Survey in Yunnan Province” [37], the slope gradients were classified as shown in

Table 1. Framework for slope division.

Slope Division	Code
0~5°	I
5~15°	II
15~25°	III
25~35°	IV
35~45°	V
45~90°	VI

.

Curvature [38] plays a crucial role in slope unit division, accurately reflecting the terrain’s curvature and local morphology, thereby affecting the accuracy of terrain representation. The calculation formula is Equation (3):

$$k=\left(k_{plan}+k_{pro}\right)/2$$

(3)

where $k$ denotes the average curvature; $k_{plan}$ denotes the plane curvature; and $k_{pro}$ denotes the section curvature.

In combination with Pike’s [39] and Wang et al.’s [22] introduction to curvature, we can determine the convexity or concavity of each grid cell through the mean curvature (k). When k > 0, the grid cell is identified as convex, and, when k < 0, the grid cell is identified as concave. Convex grid cells are classified as ridge areas, whereas concave cells are classified as valley areas. The boundary between these two regions, where the mean curvature changes signs, is a flat area (k = 0). Thus, the terrain is divided into three categories: ridges, valleys, and flat areas. Considering curvature allows for the scientific delineation of the convex, concave, and flat regions of slopes.

3. Experimental Results

3.1. Slope Units Extracted via the TFPF-SU Method

In this study, on the basis of a DEM with a spatial resolution of 8.5 m, three key terrain parameters were precisely extracted: curvature, slope, and aspect using ArcGIS software. The data were provided by the China Aero Geophysical Survey and Remote Sensing Center for Natural Resources and downloaded from Bigemap. To ensure the comparability of results, all relevant data were unified to use the 2000 National Geodetic Coordinate System, with a projected system based on the Gauss–Krüger projection in 3° zones, and the elevation reference adopted the 1985 National Elevation Datum. Using these parameters, we conducted an in-depth integration and classification analysis aimed at achieving a refined and scientific division of slope units. During the classification process, we strictly adhered to the principle of terrain similarity, ensuring that each divided slope unit exhibited high consistency in terms of the continuity of the slope gradient (a slope should change smoothly between adjacent units, without sudden changes), the regularity of the aspect distribution (the direction of the slopes within a unit should be relatively concentrated, avoiding mixed and inconsistent directions), and the uniformity of curvature characteristics (the curvature features within a unit should remain consistent, reflecting a homogeneous terrain form). Through this comprehensive consideration and rigorous classification of slope, aspect, and curvature parameters, we ensured the precision and rationality of slope unit division, effectively revealing the subtle features and spatial distribution patterns of the terrain structure in the study area. The TFPF-SU method extracted a total of 4407 slope units within the study area.

3.2. Slope Units Extracted via a Hydrological Method

Currently, a widely used method for slope unit extraction in the field of geological hazard assessments is the hydrological analysis method proposed by Carrara et al. [13]. We will validate the effectiveness of our method by comparing it with the hydrological analysis method.

Surface hydrological analysis was used to divide slope units on the basis of the DEM, as shown in Figure 4. The steps mainly include DEM inversion, depression filling, flow direction extraction, confluence accumulation calculation, river network generation, catchment basin generation, and raster-to-vector conversion. The core of this process involves identifying valley lines and ridge lines via positive and negative DEM datasets, respectively. By integrating the watershed areas from both datasets and making necessary manual adjustments to ensure reasonable division, the resulting regions separated by watershed divide lines are classified as slope units.

In this study, the flow accumulation threshold for both the positive and negative DEM datasets was set to 100 for slope unit division. When determining the flow accumulation threshold, we referenced the method proposed by Huang et al. [23], which identifies the appropriate threshold based on the relationship between flow accumulation and stream density. As the flow accumulation threshold increases, stream density initially decreases rapidly and then stabilizes. Generally, the inflection point is chosen as the appropriate threshold for slope unit extraction. However, due to errors in the hydrological method, such as the erroneous classification of parallel pseudovalleys, standard slope units could not be formed, thus limiting further quantitative assessments. Therefore, manual adjustments were made. Following the adjustments, the study area was divided into 1184 slope units.

4. Results, Analysis, and Discussion

4.1. Results and Analysis

When applying the TFPF-SU method and the hydrological analysis method for slope unit subdivision, both methods were based on 8.5 m of DEM data to ensure the reliability of the comparison. We compared the two results from both qualitative and quantitative perspectives, analyzing the practicality and rationality of the TFPF-SU method. For a more intuitive comparison of the division results between the two methods during the qualitative analysis, we selected two regions from Figure 5: Case 1 is a relatively flat area, and Case 2 is a mountainous area with significant terrain fluctuations.

4.1.1. Qualitative Analysis

As shown in Figure 6, the unit boundaries obtained via the hydrological method do not completely match the actual ridge and valley lines, with the distinction between horizontal planes and inclined planes often unclear. Moreover, different terrain types are merged into the same unit, and the hydrological method relies on terrain inversion to define subwatershed boundaries. These terrain depressions generated on the basis of the inversion height can be corrected by filling depressions but can lead to the division of artificially large flat areas, leading to the appearance of unrealistic parallel flow sections or pseudochannels in the river network (Figure 6b,d). The watershed boundaries defined on the basis of such river networks also tend to exhibit inaccurate parallelism among channels. Additionally, many nonflat areas are improperly smoothed during processing, and their original terrain characteristics are lost. In contrast, the slope units generated by the TFPF-SU method exhibit greater uniformity and conformity to the terrain, effectively avoiding problems associated with overly large unit areas or the improper integration of multiple independent slopes. This method not only finely distinguishes between inclined and flat areas but also eliminates the phenomenon of elongated and irregular slope units through a standardized unit shape design, providing an accurate and intuitive depiction of slope terrain (Figure 6a,c). Overall, the TFPF-SU method provides significant advantages in maintaining the rationality of the slope unit structure and the authenticity of the terrain.

This method not only overcomes the limitations of traditional hydrological subdivision but also provides a more refined analytical tool for the study of complex terrains, making it highly valuable in practical applications.

4.1.2. Quantitative Analysis

The size and shape of slope units can be used to quantitatively evaluate the extracted results.

The unit size is crucial for mapping landslide susceptibility, as various indicators describing geological and environmental factors are extracted on the basis of slope units. To make the environmental factors of each unit comparable, the area covered by each unit should be similar. In this study, size uniformity can be reflected by the distribution of the unit area. Figure 7a,b shows the area distributions of the slope units extracted via the two methods. The TFPF-SU method divides the region into slope units with areas generally between 100,000 and 500,000 m², with 69.93% of the units falling within this range. The hydrological method divides the region into slope units with areas generally between 20,000 and 1,000,000 m², with 38.78% of the units falling within this range. In addition, the average, standard deviation, minimum, and maximum values of the slope unit area obtained with the TFPF-SU method are 422,774, 286,190, 1819, and 4,802,357, respectively, and those for the hydrological method are 1,600,211, 1,381,013, 21,829, and 13,206,480, respectively. Overall, the TFPF-SU method more uniformly extracts slope elements.

The shape of a slope unit can be described by the ratio of the square of the perimeter to the area (R). The minimum ratio of circles is 12, whereas the ratios of squares and equilateral triangles are moderately higher, with values of 16 and 20, respectively. If a unit is too flat or elongated, the ratio greatly increases. When the value of R reaches 30, the aspect ratio of the slope unit exceeds 5:1. Among all slope units extracted via the TFPF-SU method, 70% of the units have R index values ranging from 16 to 28, and 16.79% of the R index values exceed 30. Among all the slope units extracted via the hydrological method, approximately 32.03% of the units have R index values between 16 and 28, and 57.94% have R index values exceeding 30. It can be seen that the slope units extracted by the TFPF-SU method are more uniform than those extracted by the hydrological method.

Therefore, the TFPF-SU method is more suitable than the hydrological method for dividing landslide susceptibility mapping units.

4.2. Discussion

4.2.1. Advantages of the TFPF-SU Method in Extracting Slope Units

Another advantage of the TFPF-SU method is its ability to better reflect and preserve terrain characteristics. This method divides the surface into regions with similar topographic attributes based on slope, aspect, and curvature. This means that each slope unit is uniform in these geometric characteristics, reflecting natural terrain variations and avoiding artificial boundaries or anomalous shapes that may be introduced by hydrological methods. The region-growing algorithm recursively expands adjacent pixels that conform to terrain features until the similarity conditions are no longer met. This recursive process captures the continuous changes in local terrain and ensures that the intrinsic terrain characteristics of each slope unit are consistent, thereby better preserving the complexity of natural landforms. Therefore, this method has significant advantages in preserving natural terrain features, especially for areas with significant terrain variations, where it can better capture and reflect terrain complexity.

4.2.2. Future Improvements of the TFPF-SU Method

Slope units hold great potential for applications in landslide susceptibility analysis. However, in landslide susceptibility mapping, it is important not only to emphasize the impact of topographic factors but also to comprehensively consider various environmental factors such as geology, hydrology, and vegetation. For example, the strength, degree of weathering, and permeability of soils and rocks directly affect slope stability. Heavy rainfall can increase soil weight, reduce friction between rock and soil, weaken their resistance to sliding, and ultimately trigger landslides. However, our current slope unit subdivision method mainly relies on terrain features and overlooks other key environmental variables. For instance, on low-elevation yet uniform surfaces, it is difficult to accurately identify and delineate boundaries between different landform types. As a result, the extracted slope units may span multiple types of rock and soil or have different surface runoff characteristics, and these factors are not fully incorporated into the analysis, affecting the accuracy and reliability of landslide susceptibility predictions. Therefore, future improvements to the TFPF-SU method should not only optimize the extraction of terrain features. They should also introduce various environmental factors such as geological structures, groundwater dynamics, and vegetation cover. This will enhance the comprehensiveness of the slope unit subdivision and the accuracy of landslide susceptibility mapping. Only by integrating these multidimensional factors can we more comprehensively reflect slope stability and provide a more scientific basis for landslide prediction and prevention.

5. Conclusions

A new slope unit definition and extraction method (TFPF-SU) for regional landslide analysis is proposed. On the basis of the principle of consistency between the slope direction and curvature, a slope unit is defined as a continuous and uniform closed area with no abrupt change in slope direction. To assess the accuracy and efficiency of the developed TFPF-SU method, Dongchuan District, Yunnan Province, China, was selected as the research area, and the proposed method was compared with a commonly used hydrological analysis method. The following conclusions can be drawn from this study:

(1) A new method for extracting slope units (TFPF-SU) was proposed by integrating terrain parameters such as slope, aspect, and curvature. This method dynamically selects seed points and uses region growing based on terrain similarity, combined with a stack-structured recursive marking process. This approach makes the slope unit subdivision more flexible and detailed, ensuring unit continuity and completeness.

(2) Through qualitative and quantitative analyses, the effectiveness of the TFPF-SU method was validated, and a detailed comparative study with the hydrological analysis method was conducted. The results show that the TFPF-SU method outperforms the hydrological method in terms of structural rationality and terrain authenticity of the slope units. Compared to the hydrological method, the slope units generated by TFPF-SU are more uniform and better suited for landslide analysis scenarios.