Study on the Spatio-Temporal Distribution and Influencing Factors of Soil Erosion Gullies at the County Scale of Northeast China

Abstract

Gully erosion refers to the landform formed by soil and water loss through gully development, which is a critical manifestation of soil degradation. However, research on the spatio-temporal variations in erosion gullies at the county scale remains insufficient, particularly regarding changes in gully aggregation and their driving factors. This study utilized high-resolution remote sensing imagery, gully interpretation information, topographic data, meteorological records, vegetation coverage, soil texture, and land use datasets to analyze the spatio-temporal patterns and influencing factors of erosion gully evolution in Bin County, Heilongjiang Province of China, from 2012 to 2022. Kernel density evaluation (KDE) analysis was also employed to explore these dynamics. The results indicate that the gully number in Bin County has significantly increased over the past decade. Gully development involves not only headward erosion of gully heads but also lateral expansion of gully channels. Gully evolution is most pronounced in slope intervals. While gentle slopes and slope intervals host the highest density of gullies, the aspect does not significantly influence gully development. Vegetation coverage exhibits a clear threshold effect of 0.6 in inhibiting erosion gully formation. Additionally, cultivated areas contain the largest number of gullies and experience the most intense changes; gully aggregation in forested and grassland regions shows an upward trend; the central part of the black soil region has witnessed a marked decrease in gully aggregation; and meadow soil areas exhibit relatively stable spatio-temporal variations in gully distribution. These findings provide valuable data and decision-making support for soil erosion control and transformation efforts.

1. Introduction

Soil erosion refers to the process in which soil and its parent material are detached and transported by external forces acting on surface particles [1,2]. Depending on the type of external force, natural soil erosion can be categorized into water erosion, wind erosion, freeze–thaw erosion, and gravity erosion [3]. Among these types, gully erosion forms distinctive linear, trough-like depressions that represent one of the most severe manifestations of soil degradation [4]. The progression of erosion gully channels poses the greatest threat to soil resources, significantly reducing arable land area. Once erosion gullies reach a size where they cannot be leveled through conventional tillage practices, the resulting fragmentation of fields impedes the operation of large-scale agricultural machinery, thereby decreasing farming efficiency [5]. Furthermore, the development of erosion gullies adversely affects soil properties [6,7], contributing to a decline in overall soil quality. Additionally, the presence of erosion gullies leads to reductions in key indicators of soil fertility, increasing the likelihood of excessive or improper use of chemical fertilizers [8], which further exacerbates the degradation of cultivated land. Consequently, understanding the spatio-temporal dynamics of erosion gullies is crucial for assessing soil erosion conditions, ensuring national food security, and promoting sustainable regional development.

The full progression of gully erosion can be classified into four distinct phases: rill formation, shallow gully development, incised gully stage, and deep gully formation [9]. Rills are generally characterized by depths and widths no greater than 20 cm and can often be smoothed out through regular agricultural practices such as plowing [10]. Consequently, they are sometimes grouped under sheet erosion or viewed as an intermediate phase between sheet erosion and more advanced gully erosion in academic literature [11]. Shallow gullies, on the other hand, form temporary channels that may also be filled by farming activities, although their remnants remain visible. If left unchecked, these features gradually transition into incised gullies. Both incised and deep gullies exhibit well-defined gully structures that cut across agricultural land, leading to substantial land degradation [12]. Depending on the spatial scope of observation, gully erosion monitoring can generally be divided into two categories: on-site field measurements and large-area remote sensing analysis. Field-based monitoring techniques include the use of measuring tapes, pin meters, GPS devices, and 3D laser scanning systems [13]. Among these, GPS RTK (Real-Time Kinematic) technology is particularly valued for its high precision, fast data acquisition speed, and practical efficiency, making it one of the most widely used tools for on-ground gully erosion assessment [14]. When it comes to monitoring gully development over large regions, the scale and number of gullies involved significantly increase the complexity and workload of field surveys. In such cases, remote sensing techniques offer distinct advantages, including broad spatial coverage, all-weather operability, and the ability to provide continuous temporal data, making them highly effective for regional soil erosion assessment [15,16]. Current research in this area mainly focuses on employing UAV (Unmanned Aerial Vehicle) photogrammetry to generate high-resolution orthoimages for analyzing the spatial patterns and evolutionary trends of gullies. Alternatively, satellite-based remote sensing imagery is commonly used for visual identification and interpretation of prominent gully types, such as incised and deep gullies [17,18].

Prior studies on gully evolution have mainly concentrated on three aspects: erosive forces, soil resistance to erosion, and mitigation strategies. Erosive forces are primarily driven by factors such as rainfall, surface flow, gravitational effects, and other environmental influences that contribute to gully erosion [19]. The resistance of soil to gully erosion is largely affected by topographic features, vegetation density, and patterns of land use [20]. Mitigation strategies, on the other hand, refer to the impacts of human interventions aimed at controlling gully development [21]. Notably, increased surface runoff is widely acknowledged as the primary trigger for gully initiation, with variations in runoff typically linked to climate variability or transformations in land use [22]. Among climatic variables, rainfall has the most pronounced effect on runoff generation and is regarded as a dominant driver in gully formation [23,24]. During the process of runoff-induced soil detachment and transport, topography plays a decisive role. At a regional level, variations in terrain often lead to a non-uniform and clustered spatial distribution of gullies [25,26]. In particular, slope gradient and aspect significantly influence the spatial arrangement of gullies [27]. The relationship between slope steepness and water-induced erosion is nonlinear and characterized by a threshold behavior. Once a certain slope threshold is exceeded, erosion rates tend to decline due to decreased flow velocity and energy loss [28]. Furthermore, vegetation cover is widely recognized as a crucial element affecting both soil and water retention and gully development dynamics [29,30].

As a critical commodity grain production base, Northeast China has garnered significant attention from the state regarding erosion gully prevention and control. However, the task of gully management remains highly challenging. According to the results of China’s first national water resources census conducted between 2010 and 2012 [31], within the 9.45 × 10⁵ km² of the Northeast black soil region, the total number of gullies reached 2.96 × 10⁵, with a cumulative length of approximately 1.96 × 10⁵ km and a total area of about 3647 km² [32]. Furthermore, in 2020, the Songliao River Water Conservancy Commission surveyed a 1.09 × 10⁶ km² area of the Northeast black soil region, revealing that the total number of gullies was 6.67 × 10⁵, including 3.94 × 10⁵ main gullies and 2.73 × 10⁵ branch gullies, with a total length of 2.102 × 10⁵ km, a total area of 4029.57 km², and a gully density of 0.2 km/km². These findings clearly indicate that erosion gullies in Northeast China are extensively distributed. However, existing studies on the distribution and investigation of erosion gullies in Northeast China predominantly focus on large-scale regional analyses, such as the entire Northeast region [33,34], typical black soil zone such as the Songnen Plain [35,36], or provincial-level assessments [37,38]. Limited attention has been given to the local-level factors influencing gully development or to the formulation of targeted prevention and control policies. Although provincial-level surveys such as those conducted by the Songliao River Water Conservancy Commission have collected data on quantity, length, area, and density of erosion gullies, these aggregated statistics often obscure localized driving forces. For example, the 16.01% increase in gully numbers in Bin County reflects the direct impact of inappropriate agricultural practices and insufficient drainage infrastructure at the county level.

Given the vastness of Northeast China, different areas exhibit varying vegetation, topographic, and land use conditions, which in turn affect the spatio-temporal characteristics of gullies differently. Additionally, precipitation patterns and amounts influence slope runoff and sediment production, serving as a key driving force for gully formation and development. Moreover, the development of erosion gully is closely related to soil type [39]. Therefore, Bin County in Heilongjiang Province (an area severely affected by soil erosion) was selected in this study. Specifically, the spatial distributions and changes in erosion gullies from 2012 to 2022 were extracted through visual interpretation using high-resolution remote sensing images. Subsequently, kernel density evaluation was applied to quantify the spatial aggregation patterns of erosion gullies across Bin County, thereby clarifying their spatio-temporal evolution and the underlying influencing factors. This study investigates the spatial distribution characteristics and dynamic changes in erosion gullies in Bin County, aiming to quantify the influence of various factors on gully distribution patterns, enhance the understanding of the primary drivers of gully formation and development, and elucidate the associated developmental mechanisms. The findings contribute to a more comprehensive understanding of the processes governing gully initiation and expansion in the black soil region of Northeast China. Furthermore, this research also provides a solid theoretical foundation and scientific basis for zonal and graded management strategies for gully erosion control, thereby offering actionable insights for the sustainable protection of black soil resources and the safeguarding of regional food security.

2. Materials and Methods

2.1. Study Area

Bin County was selected as the study area in this study, located in Heilongjiang Province, China. The county has a perimeter of 377 km and a total area of 3843 km² (as shown in Figure 1). Bin County exhibits a cold temperate continental monsoon climate, characterized by significant temperature variations in spring and autumn, a short and warm summer, and an extremely cold winter influenced by Siberian cold air masses. The annual average temperature and precipitation are 3.9 °C and 681 mm, respectively. The predominant soil types include dark brown soil, albic dark brown soil, albic soil, meadow soil, and black soil. In addition, the total cultivated land spans approximately 1.78 × 10⁵ ha while the total forest area covers approximately 1.23 × 10⁵ ha. Among all counties in Heilongjiang Province, Bin County ranks first in terms of the gully number and total eroded area, second in terms of total gully length, and third in terms of gully density [40], indicating that it is severely impacted by soil erosion.

2.2. Data Source

The high-resolution remote sensing data used for extracting gully information consists of GF-1 satellite images with a resolution of 2 m. The digital elevation model (DEM) data with a spatial resolution of 30 m for obtaining topographic information was sourced from 91 Map Assistant. Annual average precipitation data from 2012 to 2022 were obtained from the National Tibetan Plateau Scientific Data Center (https://www.gscloud.cn/). Soil type data were retrieved from the National Science and Technology Infrastructure Platform-National Earth System Science Data Sharing Service—Northeast Black Soil Science Data Center (https://northeast.geodata.cn/). In addition, the Landsat 8-OLI remote sensing image (with a spatial resolution of 30 m) used to obtain information on vegetation coverage and land use types was sourced from the Geospatial Data Cloud Platform (https://www.gscloud.cn/).

2.3. Extraction of Erosion Gully

The visual interpretation method is a classical expert-driven approach widely used in soil erosion investigations [41,42], which involves delineating the extent of soil erosion through image analysis and classifying or grading erosion types by integrating relevant thematic information. Specifically, this study began with field investigations to collect detailed data on soil types, lithology, land use, as well as the vegetation growth within gullies. Following the fieldwork, interpretation markers for different gully types were established by combining the characteristics of remote sensing images with the results of the field surveys (

Table 1. Interpretation criteria for different types of erosion gullies.

Position	Category	Photo	Image	Feature
127°55′52″ E 45°53′49″ N	Developing gully			Narrowly striped, light green, with low vegetation coverage at the gully bottom
127°32′27″ E 45°48′37″ N	Stable gully			Blocky, light green, with high vegetation coverage at the gully bottom
127°32′38″ E 45°47′56″ N	Newly formed gully			Striped, bare soil color, no vegetation coverage at the gully bottom
127°53′24″ E 45°38′38″ N	Farmland gully			Narrowly striped, bare soil color, no vegetation coverage at the gully bottom
127°54′56″ E 45°54′24″ N	Forestland gully			Banded, dark brown, with moderate vegetation coverage at the gully bottom
127°25′02″ E 45°48′30″ N	Road gully			Narrowly striped, bare soil color, no vegetation coverage at the gully bottom
128°04′39″ E 45°51′27″ N	Grassland gully			Striped, bare soil color, no vegetation coverage at the gully bottom

). Thereafter, visual interpretation was performed using a human-computer interactive approach to accurately determine the head positions of individual erosion gullies.

To ensure the accuracy of the visual interpretation, a field survey was also conducted in this study (Figure 2). Specifically, the layer files of the suspected gullies were imported into Ovitalmap V10.3.0 software. Based on GPS-RTK technology, on-foot inspections and measurements were performed to assess and document the development status, surface coverage, terrain features, and other relevant information of the suspected gullies. For areas inaccessible by foot, the Phantom4 RTK drone was employed for photogrammetric surveys in order to obtain precise location information and spatial distribution for the gullies across the entire Bin County in both 2012 and 2022.

2.4. Topographic Information Extraction

The terrain information extracted in this study includes slope and aspect. Slope refers to the angle between the tangent plane at any point on the ground surface and the horizontal plane [43]. Specifically, slope analysis of digital elevation model (DEM) involves calculating the rate of change in elevation values between two adjacent pixels in the DEM to quantify the steepness of the terrain using the following formula [44]:

$$\mathrm{S}=\arctan \sqrt{{\mathrm{p}}^2+{\mathrm{q}}^2}\times 180/\pi$$

(1)

where S represents the slope, p denotes the rate of elevation change in the north–south direction, and q indicates the rate of elevation change in the east–west direction. In this study, the values of p and q were calculated within a 3 × 3 DEM window using the numerical differentiation method. The results were then classified according to the slope classification standard for topographic detail maps proposed by the International Geographical Union [45]: 0° to 0.5° is categorized as plain, 0.5° to 2° as gentle slope, 2° to 5° as moderate slope, 5° to 15° as slope, 15° to 35° as steep slope, 35° to 55° as very steep slope, and 55° to 90° as vertical cliff.

Aspect refers to the direction of the projection of the normal of the slope on the horizontal plane. In remote sensing images, the aspect is the direction of the maximum variable of the elevation value, with its equation as follows:

$$\mathrm{D}=\arctan ({\displaystyle \frac{\mathrm{p}}{\mathrm{q}}})$$

(2)

where D represents the aspect, p represents the elevation change rate in the north–south direction, and q represents the elevation change rate in the east–west direction.

2.5. Land Use Type Extraction

This study employed supervised classification to extract land use types, which involves selecting representative and typical training sample areas, extracting the spectral characteristics of various ground objects from these areas to train the remote sensing images, and subsequently deriving a discriminant function [46]. Based on this discriminant function, unclassified remote sensing data were classified. The study adopted the maximum likelihood classification method, which calculates the probability of each pixel belonging to each category and assigns the pixel to the category with the highest probability [47]. The discriminant rule is as follows:

$$\mathrm{D}=\mathrm{I}\mathrm{n}({\alpha }_{\mathrm{i}})-[0.5\mathrm{I}\mathrm{n}(\left|\mathrm{C}\mathrm{o}{\mathrm{v}}_{\mathrm{i}}\right|)]-[0.5(\mathrm{X}-{\mathrm{M}}_{\mathrm{i}})\mathrm{T}(\mathrm{C}\mathrm{o}{\mathrm{v}}_{\mathrm{i}}^{-1})(\mathrm{X}-{\mathrm{M}}_{\mathrm{i}})]$$

(3)

where D, i, and X represent the weighted distance, a certain feature type, and the measurement vector of the pixel, M_i, Cov_i, and α_i refer to the mean value of the pixel brightness in the training area of the category, the covariance matrix, and the probability of the pixel belonging to the category.

2.6. Extraction of Fractional Vegetation Coverage (FVC)

The Normalized Difference Vegetation Index (NDVI) is an important remote sensing indicator reflecting the growth condition of vegetation. After radiometric calibration and atmospheric correction of the remote sensing images, NDVI was extracted using the following formula to reflect the vegetation conditions on the ground [48]:

$$\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}=(\mathrm{N}\mathrm{I}\mathrm{R}-\mathrm{R})/(\mathrm{N}\mathrm{I}\mathrm{R}+\mathrm{R})$$

(4)

where NIR represents the reflectance of the near-infrared band and R represents the reflectance of the red band.

The pixel bisection model was carried out to extract FVC in Bin County based on the clear relationship between NDVI and FVC, which can effectively reduce the influence of atmospheric, soil background, and vegetation type on remote sensing information, and the calculation formula is as follows [49]:

$$\mathrm{F}\mathrm{V}\mathrm{C}=(\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}-\mathrm{N}\mathrm{D}\mathrm{V}{\mathrm{I}}_{\mathrm{s}\mathrm{o}\mathrm{i}\mathrm{l}})/(\mathrm{N}\mathrm{D}\mathrm{V}{\mathrm{I}}_{\mathrm{v}\mathrm{e}\mathrm{g}}-\mathrm{N}\mathrm{D}\mathrm{V}{\mathrm{I}}_{\mathrm{s}\mathrm{o}\mathrm{i}\mathrm{l}})$$

(5)

where NDVI_soil represents the NDVI value of completely bare soil or areas without vegetation cover, and NDVI_veg is the NDVI value of vegetation. It should be noted that the most crucial aspect of this method is to determine the values of NDVI_soil and NDVI_veg. Although variations exist in the selection of cumulative frequency thresholds due to the influence of atmospheric conditions, illumination, and humidity [50,51], the majority of studies adopted the NDVI values at cumulative frequencies of 5% and 95% as representative endmember values for pure soil and pure vegetation, respectively. This threshold selection criterion has been extensively validated and is broadly applied across a wide range of applications [52,53,54]. To further evaluate the impact of vegetation cover on soil erosion, vegetation coverage was further classified into five grades according to the SL190-2007 “Soil Erosion Classification Standard” [55]: low (0–0.3), medium-low (0.3–0.45), medium (0.45–0.6), medium-high (0.6–0.75), and high (0.75–1).

2.7. Kernel Density Evaluation

Kernel density evaluation (KDE) provides a robust tool for analyzing the spatial distributions of erosion gullies, enabling the description of their aggregation patterns and reflecting the continuous spatial distributions [56,57]. Specifically, KDE estimates point or line densities within the sliding window of remote sensing images, generating a continuous density surface that illustrates the aggregation status of point or line elements across the study area [58]. Specifically, the specific implementation process of KDE analysis in this study is as follows: (1) Define the search radius (also regarded as the bandwidth of the kernel function); (2) Draw a circle with the specified radius centered on each grid cell and identify all gullies within the circle; (3) Use the kernel function to calculate the density contribution of each gully; (4) Accumulate the density contribution values of all pixels containing gullies within the search radius, yielding the kernel density value.

$${\mathrm{f}}_{\mathrm{n}}(\mathrm{X})={\displaystyle \frac{1}{\mathrm{n}\mathrm{h}}}{\displaystyle {\sum }_{\mathrm{i}=1}^{\mathrm{n}}\mathrm{k}(\frac{\mathrm{x}-{\mathrm{x}}_{\mathrm{i}}}{\mathrm{n}}})$$

(6)

where f_n(x) represents the kernel density value. The larger this value is, the higher the probability of the event occurring and the denser the distribution of erosion gullies. k is the kernel function; x − x_i is the distance from the estimation point to the sample X_i; h is the bandwidth of the kernel function. An increase in bandwidth can result in a smoother density estimation of the points; however, it may obscure the underlying density structure. Conversely, a smaller bandwidth leads to more abrupt and uneven density variations. In this study, the bandwidth was determined using the default method in ArcGIS V10.5 software, which employs an improved Silverman’s empirical rule [59] with the corresponding formula listed as follows:

$$\mathrm{h}=0.9\times \min (\mathrm{S}\mathrm{D},\sqrt{{\displaystyle \frac{1}{\mathrm{l}\mathrm{n}2}}}\times \mathrm{D}\mathrm{m})\times {\mathrm{n}}^{-0.2}$$

(7)

where Dm represents the median distance from the average center of the input points to the average center of all points, and SD indicates the standard distance from the average center of the input points to the average center of all points. It should be noted that the above empirical rule for bandwidth has been widely applied [60,61]. Moreover, previous studies have also indicated that the selection of empirical parameters in the formula should be combined with the distribution characteristics of the samples. When the samples follow a normal distribution, 1.06 is usually adopted; regarding non-normal distributions of samples, the empirical coefficient is generally set at 0.9 [62]. After considering the spatial distribution characteristics of the erosion gullies in Bin County, different bandwidths were tested and compared with the calculation results of Formula (11) for the optimal bandwidth of KDE.

2.8. Geographic Detector Model

The Geographic detector is a widely used and effective spatial statistical method that enables the quantitative analysis of relationships between spatial heterogeneity and its potential driving factors. In this study, six environmental factors (including soil type, precipitation, slope, aspect, land use type, and vegetation coverage) were selected to identify the key driving factors influencing the distribution density of erosion gullies in Bin County in both 2012 and 2022. Additionally, the interactive effects between each pair of factors were examined to determine whether they exhibited synergistic enhancement or weakening effects. The methodology is structured as follows:

For each factor affecting the overall kernel density distribution of gullies in Bin County during these two years, the independent explanatory powers q(X_m) and q(X_n) of any two factors X_m and X_n are first calculated (the detailed calculation procedure of q is carried out based on Reference [63]). Subsequently, the combined explanatory power $\mathrm{q}({\mathrm{X}}_{\mathrm{n}}\cap {\mathrm{X}}_{\mathrm{m}})$ of the two-factor interaction is computed. The type of interaction is determined by comparing the values of q(X_m), q(X_n), and $\mathrm{q}({\mathrm{X}}_{\mathrm{n}}\cap {\mathrm{X}}_{\mathrm{m}})$, with the classification criteria presented in

Table 2. Types of factor interactions in geoprobes.

Interaction Criterion	Interaction
$q(X_n\cap X_m)<\min [q(X_n),q(X_m)]$	Nonlinear attenuation
$\min [q(X_n),q(X_m)]<q(X_n\cap X_m)<\max [q(X_n),q(X_m)]$	Single-factor nonlinear attenuation
$q(X_n\cap X_m)>\max [q(X_n),q(X_m)]$	Two-factor enhancement
$q(X_n\cap X_m)=q(X_n)+q(X_m)$	Independent
$q(X_n\cap X_m)>q(X_n)+q(X_m)$	Nonlinear enhancement

q(X_n) and q(X_m) represent the q-values of the dependent variables X_n and X_m, respectively; q(X_n∩X_m) is the interaction between q(X_n) and q(X_m); min[q(X_n), q(X_m)] refers to taking the minimum value between q(X_n) and q(X_m); max[q(X_n), q(X_m)] represents taking the maximum value between q(X_n) and q(X_m).

.

3. Results

3.1. Extraction Results of Erosion Gullies

The visual interpretation results of the head positions of erosion gullies in Bin County for 2012 and 2022 are presented in Figure 3, respectively. It should be noted that the number of gullies was statistically analyzed using the attribute table function. It can be seen from the figure that in 2012, the total number of erosion gullies in Bin County was 6056, while it increased to 7026 in 2022. Over the 10-year period, the number of erosion gullies in Bin County increased by 970, representing a growth rate of 16.01%.

3.2. Extraction Results of Influencing Factors

Figure 4 illustrates the spatial distribution of extracted influencing factors in the study area. As shown in the figure, Bin County encompasses five major soil types found in Heilongjiang Province, such as dark brown soil, albic dark brown soil, albic soil, meadow soil, and black soil. The average annual precipitation in the study area ranges from 588 to 723 mm, exhibiting a significantly increasing trend from northwest to southeast. Concurrently, when combined with the spatial distribution characteristics of slope, it is evident that plain areas are predominantly concentrated along the Songhua River, with slopes gradually increasing from northwest to southeast. Steep slopes exceeding 15° are primarily concentrated on the outer edges of the study area, while plain terrain is relatively scarce in local regions. The distribution of other slopes across different directions appears relatively uniform. Land use types in Bin County are mainly categorized into seven classes, including forest, grassland, cultivated land, transportation land, construction land, water bodies, and other human-disturbed land. Among these land uses, cultivated land and forest dominate as the primary land use types. Specifically, cultivated land is predominantly concentrated in the northwestern, northern, and central areas of Bin County, whereas forest land is largely distributed in the eastern and southern parts of the study area. Additionally, Figure 4 reveals that areas with low vegetation coverage are mainly located in the northwestern, northern, and central regions of Bin County, while high vegetation coverage areas are predominantly distributed in the southern, southeastern, and eastern parts of the county.

The area statistics under different influencing factors are presented in Figure 5. As shown in the figure, the largest area corresponds to slopes ranging from 5° to 15°, measuring 1300.64 km², which accounts for approximately 33.82% of the total study area. In contrast, the smallest area is associated with slopes ranging from 0° to 0.5° at 149.83 km², representing only 3.89% of the total area. Additionally, Figure 5 reveals that the north-facing slope has the largest area within the study region, with 662.39 km², accounting for approximately 17.24% of the total area. The smallest area corresponds to regions without a defined slope direction, comprising only 1.3% of the total area. Areas with other slope directions are concentrated between 400 and 500 km². Regarding vegetation coverage, the largest area is characterized by low vegetation coverage, spanning 2461.64 km², accounting for approximately 64.05% of the total area. Areas with medium-low, medium, and medium-high vegetation coverage are relatively small, measuring 1.53 km², 5.96 km², and 35.3 km^2, respectively, and collectively accounting for only 1.1% of the total area. The high vegetation coverage area measures 1339.24 km², representing approximately 34.85% of the total area. Land use in Bin County is predominantly arable land, covering 2182.95 km², approximately 56.80% of the total area. The second largest land use type is forest land, with an area of 1221.06 km², accounting for approximately 31.77% of the total area. Furthermore, Figure 5 indicates that dark brown soil constitutes the largest soil type in the study area, with a total area of 1897.33 km², primarily distributed in the eastern and southern parts of the region. Meadow soil is the second largest soil type, covering 994.91 km², which refers to 25.89% of the total area, mainly distributed in the northern part of the region and along the Songhua River. Albic soil has the smallest area, at only 2.59 km². In addition, Figure 5 shows that average annual precipitation in the study area between 2012 and 2022 is concentrated between 618 and 678 mm. The largest area corresponds to precipitation levels between 618 and 648 mm at 2118.58 km², accounting for approximately 55.12% of the total area. However, regions with precipitation exceeding 678 mm are quite small, comprising only 1.8% of the total area.

3.3. Results of KDE for Different Factors

The kernel density distributions of gully features in 2012 and 2022 were independently analyzed, with results shown in Figure 6. From the figure, it can be seen that the maximum kernel density of erosion gullies in Bin County decreased from 9.5 in 2012 to 6.7 in 2022. In previously highly aggregated regions, such as the northwest and southwest, a general downward trend in kernel density was observed, with variations in the magnitude of decline across different areas. Notably, a spatial diffusion pattern was identified, particularly in the southeast and eastern regions, where kernel densities exhibited an overall increasing trend. These areas were characterized by low or no significant gully aggregation in the earlier period.

3.3.1. Soil Types

Table 3. Variation in the gully number under different soil types.

Soil Type	Gully Number in 2012	Gully Number in 2022	Gully Change
Dark brown	2953	3887	934
Albic dark brown	3	25	22
Albic	0	1	1
Meadow	1480	1265	−215
Black	1620	1848	228

summarizes the statistical results of gully counts across different soil types. As shown in the table, the dark brown soil area has the highest number of erosion gullies, with 2953 in 2012 and 3887 in 2022, accounting for 48.76% and 55.32% of the total, respectively. The black soil area ranks second with 1620 erosion gullies in 2012 and 1848 in 2022, representing 16.99% and 27.84% of the total, respectively. The albic soil area exhibits the smallest number of gullies. The table also indicates that the number of gullies in the four soil types has increased over the study period, except for the meadow soil area, since the high organic content of meadow soils accounts for their stability with lower remediation priority than for erosion-prone black soils.

Figure 7 illustrates the variation in kernel density of erosion gullies across different soil types. It is worth noting that due to the limited area and extremely low number of gullies in the albic dark brown soil and albic soil regions, the figure only depicts KDE results for the dark brown soil, meadow soil, and black soil regions. As shown in the figure, the kernel density of gullies increases in the southeast while decreasing in the northeast in the dark brown soil region, exhibiting a relatively large variation range. In the meadow soil region, the overall variation in kernel density is relatively small, primarily characterized by a decrease in the northern and eastern parts. The changes in kernel density in the black soil region are more complex, with significant decreases observed in certain areas, reaching a minimum value of −7.574.

3.3.2. Precipitation

Table 4. Variation in the gully number under different precipitation regions.

Precipitation (mm)	Gully Number in 2012	Gully Number in 2022	Gully Change
588–618	672	824	152
618–648	4092	3974	−118
648–678	1292	2226	934
678–708	0	2	2
708–723	0	0	0

indicates the number and changes in erosion gullies across different precipitation intervals. As shown in the table, the highest density of gully distribution occurs in the precipitation interval of 619–648 mm with 4092 gullies in 2012 and 3974 gullies in 2022, accounting for approximately 67.57% and 56.56% of the total, respectively. However, the number of gullies in this interval decreased by 118 over the study period. Furthermore, Table 3 indicates that the greatest increase in gully count is observed in the precipitation interval of 648–678 mm, with a total increase of 934 gullies.

Figure 8 illustrates the variation in kernel density of gullies across different annual average precipitation intervals. Due to the limited area of regions with precipitation exceeding 678 mm and the predominance of forest cover, only two gullies were identified. Consequently, this study focuses solely on the kernel density variation within the precipitation interval of 588–678 mm. As shown in the figure, for the precipitation interval of 588–618 mm, the areas where gully kernel density increases are concentrated in the central region, with a maximum value reaching 3.987. In the 618–648 mm precipitation interval, the gully kernel density exhibits a significant decrease, predominantly in the southwestern part of the region, with a minimum value of −7.084. Within the 648–678 mm precipitation range, the areas where kernel density increases display a distinct strip-like distribution, with a maximum value of 5.574.

3.3.3. Slope Interval

The count and changes in gullies in different slope intervals were statistically analyzed, with results presented in

Table 5. Variation in the gully number under different slope intervals.

Slope Interval	Gully Number in 2012	Gully Number in 2022	Gully Change
Plain	45	58	13
Gentle	432	672	240
Moderate	1513	2035	522
Slope	3468	3503	35
Steep	598	758	160

. As shown in the table, the plain area exhibits the smallest number of distributed gullies. With increasing slope values, the number of gullies follows a trend of first rising and then falling, reaching its peak in the steeper slope interval. In 2012 and 2022, the number of gullies in this interval was 3468 and 3503, accounting for 57.27% and 49.85% of the total. The second most significant distribution occurs in the gentle slope interval, where the number of gullies was 1513 and 2035, representing 24.98% and 28.96% of the total, respectively. Notably, the gentle slope interval experienced the largest change in gully count, with an increase of 522.

Figure 9 presents the results of changes in the kernel density of gullies across different slope ranges. Within each slope range, the increase in kernel density is primarily concentrated in the southeastern and southern parts of Bin County. Notably, the maximum increase in kernel density occurs within the slope range, reaching a value of 2.498. Conversely, in the northern and central regions, the kernel density decreases with the minimum value dropping to −4.379.

3.3.4. Slope Aspect

Table 6. Variation in the gully number under different slope aspects.

Aspects	Gully Number in 2012	Gully Number in 2022	Gully Change
Flat	9	20	11
North	1104	1212	108
Northeast	834	1059	225
East	833	888	55
Southeast	741	706	−35
South	622	716	94
Southwest	559	752	193
west	634	822	188
Northwest	720	851	131

shows the distribution of gullies across different slope aspects in both 2012 and 2022. As shown in the table, the plain area without a defined slope aspect exhibits the smallest number of gullies, with only 9 gullies in 2012 and 20 gullies in 2022. Additionally, Table 6 indicates that there is no significant variation in gully counts among different slope aspects. While the number of gullies has decreased on southeast-facing slopes, it has increased on other slope aspects. However, these changes are relatively minor and not statistically pronounced.

Figure 10 illustrates the variation in kernel density of gullies across different slope aspects. Given the minimal number of gullies on flat land, the change in kernel density is relatively inconspicuous. Furthermore, as depicted in Figure 10, the kernel density exhibits more pronounced changes on north-facing, east-facing, and northeast-facing slopes. Specifically, the maximum reduction in kernel density on north-facing slopes reaches −1.387, while the maximum increase on northeast-facing slopes reaches 0.874. However, compared to other slope aspects, these differences remain relatively small.

3.3.5. Land Use Types

Table 7. Variation in the gully number under different land use types.

Land Use Types	Gully Number in 2012	Gully Number in 2022	Gully Change
Transportation land	51	118	67
Forest	732	941	207
Human-disturbed land	7	4	−3
Construction land	50	53	3
Cultivated land	4873	4730	−143
Grassland	343	1180	837

summarizes the number and changes in gullies across different land use types. As shown in the table, cultivated land exhibits the highest number of gullies with 4873 in 2012 and 4730 in 2022, accounting for 80.47% and 67.32%, respectively. Forest land ranks second, with 732 gullies in 2012 and 941 in 2022, representing 12.09% and 13.39% in total. Additionally, human-disturbed land shows the smallest number of gullies. Regarding the change in gully counts, transportation land, forest land, construction land, and grassland all exhibit an increasing trend, with grassland showing the most significant increase. In contrast, cultivated land and artificial land use demonstrate a decreasing trend in gully numbers.

For gullies distributed across different land use types, the changes in their kernel density were analyzed and displayed separately, with results presented in Figure 11. As shown in the figure, the changes in kernel density in the cultivated land area are the most pronounced, with increases observed in the southern and southeastern regions, reaching a maximum value of 2.542. Conversely, the kernel density in the northern cultivated land area decreased significantly, reaching a minimum value of −5.765. Additionally, Figure 11 indicates that the kernel density in forest areas increased by 0.542, concentrated in the eastern and southeastern parts of the study area. In grassland areas, the increase in kernel density is concentrated in the eastern and northwestern parts, with a maximum value of 1.470.

3.3.6. Vegetation Coverage

Table 8. Variation in the gully number under different vegetation coverage grades.

Vegetation Coverage	Gully Number in 2012	Gully Number in 2022	Gully Change
Low	5012	5036	24
Medium-low	0	1	1
Medium	1	9	8
Medium-high	14	24	10
High	1029	1956	927

presents the number and variation in gullies across different vegetation coverage grades. As shown in the table, gullies are predominantly distributed in the low vegetation coverage range, with counts of 5012 in 2012 and 5036 in 2022, accounting for 82.76% and 71.67% of the total. The second most significant distribution occurs in the high vegetation coverage range, with 1029 gullies in 2012 and 1956 in 2022, representing approximately 16.99% and 27.84%, respectively. Table 7 also indicates that the number of gullies in the medium-low, medium, and medium-high vegetation coverage ranges is relatively small. Furthermore, the number of gullies in all vegetation coverage grades shows an increasing trend, with the high vegetation coverage range experiencing the most pronounced change by 927 gullies.

Figure 12 illustrates the changes in the kernel density of gullies across different vegetation coverage grades. Due to the limited number of gullies distributed in the medium-low and medium vegetation coverage intervals, variations in their kernel density are minimal and exert a negligible influence on gully development. Consequently, the kernel density results for these intervals are ignored in Figure 12. As shown in the figure, changes in the kernel density of gullies within the low and high vegetation coverage grades are relatively pronounced, whereas changes in the medium-high vegetation coverage have a comparatively minor impact on the region. The increase in kernel density across all vegetation coverage grades is primarily concentrated in the southeastern and eastern parts of the region, while the kernel density exhibits a decreasing trend in the southwestern part.

3.4. Interaction Results of the Geographic Detector

The q-values of the geographical detector model for Bin County in 2012 and 2022 are summarized in

Table 9. The q value of the geographical detector model calculation results in 2012.

	Soil Type	Precipitation	Slope Interval	Slope Aspect	Land Use Type	Vegetation Coverage
soil type	0.16
precipitation	0.17	0.06
slope interval	0.22	0.12	0.05
slope aspect	0.23	0.11	0.19	0.03
land use type	0.24	0.14	0.18	0.17	0.09
vegetation coverage	0.21	0.11	0.11	0.11	0.15	0.06

and

Table 10. The q value of the geographical detector model calculation results in 2022.

	Soil Type	Precipitation	Slope Interval	Slope Aspect	Land Use Type	Vegetation Coverage
soil type	0.29
precipitation	0.38	0.28
slope interval	0.31	0.32	0.03
slope aspect	0.30	0.30	0.11	0.00
land use type	0.30	0.32	0.13	0.07	0.02
vegetation coverage	0.33	0.33	0.12	0.08	0.08	0.05

, respectively. The q-values along the diagonal of the tables represent the explanatory power of each individual factor on gully distribution, whereas the off-diagonal q-values reflect the explanatory power of interactions between any two factors. Overall, multi-factor combinations exhibited significant nonlinear enhancement or two-factor enhancement effects, indicating that the formation and spatial distribution of gullies were not dominated by a single factor but resulted from the combined and interactive influences of multiple natural and environmental variables. Furthermore, the explanatory power of multi-factor combinations on gully density distribution in 2022 is generally higher than that in 2012, suggesting that the synergistic effects of multiple factors on gully development have become more pronounced over time. Specifically, the single-factor q-values of soil type and precipitation were notably higher than those of other factors, highlighting their role as the primary driving forces behind gully development and distribution.

4. Discussion

The development process and distribution characteristics of gullies are closely related to various natural and environmental factors. Specifically, Figure 7 indicates that the meadow soil is predominantly located in the northern part of Bin County, along the Songhua River, where abundant water supply and lush vegetation contribute to a high organic residue content. After freezing, decomposition occurs slowly and incompletely, resulting in higher organic matter content compared to other soil types. This contributes to a relatively small number of gullies and stable kernel density variation in the meadow soil region. In contrast, the organic matter in dark brown soil is concentrated in the surface layer and sharply decreases in depth [64]. However, the gravel content increases from the surface to the bottom, making this soil type more susceptible to water erosion [65]. In addition, the black soil is mostly clay loam to light clay, exhibiting strong water retention but poor water permeability [66]. In the study area, the black soil region largely overlaps with cultivated land. Unreasonable reclamation activities can increase the kernel density of gullies. Nevertheless, the black soil cultivated area is also a key focus for gully control in the region, and artificial gully protection has also been effective in suppressing gully development.

As shown in Table 4, an increase in precipitation does not necessarily lead to more intense gully development; the influence of surface cover conditions, topographic features, and human intervention measures must also be considered. In the region with an average annual precipitation of 588–618 mm, the central area has steeper slopes compared to surrounding areas, and cultivated land is the dominant land use type. In the 618–648 mm precipitation zone, cultivated land remains the primary land use, and the area is heavily influenced by human activities, resulting in the highest number of gullies. However, this region largely overlaps with the black soil zone and represents a key area for local gully prevention and control, where gully development has been notably alleviated in certain sections. Table 4 also indicates that erosion gully expansion is most pronounced in the 648–678 mm precipitation zone, rather than in the area with the highest rainfall. This may be attributed to human intervention suppressing gullies in high-rainfall croplands, such as contour ridges, while forest-farmland edges in 648–678 mm zones had insufficient erosion control. Furthermore, the increase in gully density is primarily concentrated in the transitional zone between forestland and cropland, which is dominated by young forest stands. Due to the underdeveloped root systems of these young trees, soil retention capacity is limited, thereby reducing the effectiveness of soil and water conservation measures.

Slope is a crucial factor in determining the extent of soil erosion, as it impacts the infiltration, runoff, and flow velocity of water [67,68]. Due to the relatively small number of gullies distributed in the plain and steep slope ranges, changes in kernel density are less pronounced in these areas. The primary influence on kernel density variation is observed in the gentle slope and moderate slope ranges. This is because these slope ranges encompass the most favorable conditions for gully formation and development, leading to both the highest gully distribution counts and the most significant changes in kernel density. In addition, as the slope increases, rainfall energy and erosive force intensify, which can thus lead to increased soil moisture and subsequent destruction of the soil structure. Loose material on the soil surface can also reduce slope stability and accelerate erosion development. Once the slope reaches a certain threshold, the rain-bearing area and runoff duration increase with slope, resulting in a decrease in both erosion gully speed and quantity [26,69]. Previous studies have demonstrated that when comparing gully density across different slope aspects during the same period, a consistent pattern emerges: gully density is typically higher on sunny slopes than on shady slopes, and higher on windward slopes than on leeward slopes [70,71]. This pattern is primarily attributed to the combined effects of freeze–thaw processes and wind erosion. However, in the analysis of gully density changes across slope aspects in this study from 2010 to 2020, it was observed that the differences in gully number and density among various slope aspects were not statistically significant, which is consistent with Jiang et al. [72]. A likely explanation is that, over the long-term development and evolution of gullies, water erosion becomes the dominant driving force, while the influence of slope aspect on soil transport and erosion by surface runoff remains relatively limited.

Regarding land use, the distribution of gullies within areas of artificial land use, transportation land use, and construction land use is relatively sparse, and their influence on the overall change in regional gully density remains limited, with kernel density values primarily ranging between −0.05 and 0.1. Although the direct damage to roads caused by surface runoff may be relatively minor, its impact on the surrounding environment can be substantial. The presence of roads tends to concentrate surface runoff, which significantly increases the likelihood of gully erosion. Furthermore, road construction can redirect concentrated runoff to adjacent watersheds, thereby expanding watershed areas and accelerating gully development. Concurrently, alterations in drainage patterns associated with urbanization may also contribute to gully formation. This highlights the necessity of monitoring and mitigating erosion gully development induced by construction activities during the urbanization process. Despite this, the presence of roads still tends to concentrate surface runoff, thereby increasing the likelihood of gully erosion. Transportation infrastructure may not only redirect concentrated runoff to adjacent watersheds but also expand watershed areas, thus intensifying gully development [73]. Furthermore, alterations in drainage patterns associated with urbanization may also contribute to gully formation [74]. This highlights the importance of monitoring and implementing preventive measures against erosion gully development caused by construction activities during the urbanization process [75,76]. In contrast, cultivated land areas are characterized by relatively gentle slopes, longer slope lengths, and a generally loose surface soil structure. Furthermore, the study area predominantly employs slope-following or slope-tilted farming practices, which result in shorter runoff convergence times and increased runoff volumes. Combined with the lack of effective drainage and water interception infrastructure for runoff regulation, cultivated land exhibits the highest number of gullies and the most significant increase in gully density. Notably, the maximum decrease in kernel density also occurs in cultivated land, indicating that it is a key target area for human intervention in gully prevention and control. The number and density of gullies in forest and grassland areas are also increasing. This trend can be attributed to two factors: the relatively steep average slopes in these areas, and the degradation of soil erosion resistance caused by illegal deforestation and inappropriate planting practices. Furthermore, as shown in Table 7 and Figure 11, there has been a substantial increase in gullies on grasslands, with an additional 837 gullies recorded between 2012 and 2022. This trend corresponds to the expansion of the livestock industry in Bin County. According to Fan et al. [40], the number of livestock in Bin County increased by 23% from 2015 to 2022, suggesting that overgrazing (especially for the grazing quotas introduced in the eastern grasslands) is the primary driver of the increased gully formation on grasslands.

Table 8 and Figure 12 also indicate that vegetation coverage has a distinct threshold effect in inhibiting the development of gullies. When the vegetation coverage exceeds 0.6, the gully density on slopes can be reduced by 42%. The underlying reason lies in the fact that, regarding the above-ground component, the vegetation canopy effectively suppresses wind erosion. Furthermore, the aerial parts of plants can slow the development of gullies by altering the spatial redistribution of rainfall and modifying the kinetic energy of raindrops. Concurrently, the surface layer of dead branches and leaves helps reduce surface runoff formation, thereby decreasing sediment loss [77]. Numerous studies have demonstrated that vegetation significantly reduces soil erosion rates and runoff shear stress, which in turn enhances surface resistance to erosion and increases the dissipation of hydrodynamic energy [78,79]. Below ground, the root system plays an equally critical role in soil erosion prevention due to its well-developed structure that provides substantial soil stabilization. As vegetation coverage increases, root density rises, leading to greater soil cohesion and a reduced likelihood of surface runoff and erosion occurrence [80,81]. Moreover, root systems can influence the soil erosion process through mechanical reinforcement mechanisms, primarily including root cohesion, root tensile strength development, and plant anchoring effects [82].

From Table 9 and Table 10, it can be concluded that the q-values of any two-factor interactions consistently exceed those of individual factors, particularly for the combination of slope and aspect, and the combination of land use type and aspect, which demonstrate significantly higher q-values, indicating strong synergistic enhancement. From a mechanistic perspective, the enhanced interactions between soil type and precipitation, as well as between soil type and vegetation coverage, suggested that in the black soil region of Northeast China, changes in soil type, increased precipitation intensity, and variations in vegetation coverage were critical to gully formation and development. Different soil types exhibited varying resistance to wind and water erosion, and increased precipitation accelerated soil and water loss, thereby promoting gully development. In contrast, extensive vegetation coverage can effectively mitigate soil erosion and inhibit gully expansion. Furthermore, the reliability and robustness of the interaction between these two factors require further investigation and long-term validation. It should also be emphasized that while these results reflect the combined influence of only two factors, the development process of gullies and the interaction mechanisms among various influencing factors are highly complex and multifaceted. Therefore, systematically interpreting gully development and spatial distribution from a macro-level perspective, as well as exploring the comprehensive influence mechanisms of multiple interacting factors on gully formation, has become a critical focus for future research.

The influence of various factors on gully development and spatial distribution highlights the urgent need to formulate scientifically sound and effective policies aimed at reducing soil erosion and curbing gully expansion. Specifically, future gully prevention and control efforts should prioritize the conversion of steep slope farmland into forest, with appropriate models selected according to regional characteristics in order to enhance the overall effectiveness of soil and water conservation. In addition, land resource planning should be conducted in a scientific and systematic manner, with farmland reclamation areas rationally designated to ensure orderly land development and consolidation. Concurrently, suitable farming practices should be implemented to reduce the incidence of soil erosion. Key gully control and land creation measures (such as land leveling, irrigation and drainage infrastructure, and field road construction) should also be actively promoted. Moreover, particular emphasis should be placed on advancing artificial afforestation initiatives and enhancing the protection of existing natural forests to further increase vegetation cover. Finally, building upon strengthened soil and water conservation, the agricultural structure should be rationally optimized, with conservation efforts integrated into the broader framework of ecosystem governance and protection.

5. Conclusions

This study employed visual interpretation methods to extract the spatial distribution characteristics and dynamic changes in gullies across the entire Bin County region in 2012 and 2022. In conclusion, the overall number of gullies increased from 2012 to 2022. The changes in gullies located in the lower part of the slope interval were the most intense, and the number of gullies in the gentle slope was the largest. The gully number in the plain area was the smallest, and the spatial changes in different slope aspects were not significantly different. Vegetation coverage had a threshold effect in inhibiting the gully development. The gully number in the cultivated area was the largest, and the growth number of gullies in the forest and grassland area showed a significant upward trend. In addition, the gully density in the black soil area decreased significantly, and the spatial change in gullies in the meadow soil area was relatively gentle. Moreover, precipitation had a certain promoting effect on the development of gullies. However, the comprehensive influence of multiple factors on the distribution and development of erosion gullies still requires further in-depth research. In future research, it is essential to conduct comprehensive comparative analyses of the performance of various machine learning and deep learning approaches in identifying gully features and extracting morphological parameters, while striving to enhance model efficiency and automation to mitigate limitations such as low-resolution outputs, misclassification, omission resulting from, or socio-economic drivers of land use. Additionally, greater attention should be given to understanding the effects of region-specific environmental variables on gully formation. For instance, long-term monitoring of snowmelt and temperature variations can provide valuable insights into the impacts of freeze–thaw cycles and snowmelt runoff on gully initiation and evolution in Northeast China.