Soil Erosion Characteristics in Tropical Island Watersheds Based on CSLE Model: Discussion of Driving Mechanisms

Abstract

Previous research has primarily focused on soil erosion issues in arid and semi-arid regions, with a limited understanding of soil erosion mechanisms in tropical areas. Additionally, there is a lack of a holistic perspective to determine the spatial attribution of soil erosion. The conversion of tropical rainforests into economically driven plantations, like rubber and pulpwood, has resulted in distinct soil erosion characteristics in specific regions. To enhance our knowledge of soil erosion patterns and mechanisms in tropical regions, it is necessary to examine soil erosion in the three major watersheds of Hainan Island from 1991 to 2021, which encompass significant geographical features such as tropical island water sources and tropical rainforest national parks. The study employed the China Soil Loss Equation (CSLE) model, slope trend analysis, Pearson correlation analysis, land-use transfer matrix, and spatial attribution analysis to examine soil erosion under different scenarios. The research results indicate that scenarios driven by the combination of natural and human factors have the greatest impact on soil erosion changes in the entire study area. Co-driven increases affected 53.56% of the area, while co-driven decreases affected 21.74%. The 31-year soil erosion showed an overall increasing trend. Human factors were identified as the primary drivers of increased soil erosion in the Nandu River basin, while a combination of climate and anthropogenic factors influenced the decrease in soil erosion. In the Changhua River basin, climate and human activities contributed to the soil erosion increase, while human activities primarily caused the decrease in soil erosion. In the Wanquan River basin, climate intensified soil erosion, whereas human activities mitigated it. This study underscores the significant combined impact of human activities and natural factors on soil erosion in tropical regions. It emphasizes the importance of considering human-induced factors when implementing soil erosion control measures in tropical regions.

1. Introduction

Soil erosion research provides the essential means for humans to use the earth’s resources appropriately and preserve and restore environmental conditions. The United Nations has identified soil conservation as one of the major challenges of the 21st century [1,2]. Soil degradation due to water erosion processes poses several threats to terrestrial ecosystems [3]. Quantifying soil erosion is an important part of regional ecological security assessment [4]. Soil erosion is a natural phenomenon, and moreover, the most important manifestation of land degradation worldwide caused by inappropriate human activities. Climate change and anthropogenic impacts may cause accelerated or minimized soil erosion formation processes in certain specific areas [5]. Consequently, soil erosion represents a global issue for society, as it often degrades soil and water resources and triggers economic losses in several regions of the world [6].

In terms of the broad categories of influencing factors, there are differences in the way that natural and anthropogenic factors affect soil erosion. The great variability of human activity on spatial and temporal scales prevents a comprehensive understanding of how humans influence regional erosion [7]. Several studies have shown that land-use/cover change, due to industrialization, inappropriate agricultural management, urbanization, overgrazing, and deforestation may increase soil erosion rates to a large extent [8]. Precipitation, a natural factor, and vegetation change, an anthropogenic factor, have a complex relationship with soil erosion change. Vegetation affects precipitation distribution and increases the runoff resistance coefficient, as well as changes the physical and chemical properties of the soil and can increase its erodibility [9]. Climate, especially temperature and precipitation, can strongly influence changes in vegetation and thus indirectly impact soil erosion by altering hydrothermal conditions at the surface [10]. The spatial distribution of soil erosion in the world indicates that the threat of soil erosion is most severe in agricultural areas with heavy rainfall [11]. Intense rainfall and windy weather directly drive the generation of soil erosion, and coupled with regional differences in topography, there is clear spatial heterogeneity in the response of soil erosion to climate change.

Soil erosion models provide effective methods for detecting and predicting soil loss [12,13]. Multiple models have been developed to estimate soil erosion in regions and watersheds [14]. Models are generally divided into empirical models and physical models, such as the revised universal soil loss equation (RUSLE) and the Water Erosion Prediction Project (WEPP) model [15,16]. Among the various models, the universal soil loss equation (USLE) and RUSLE have earned global recognition for predicting soil erosion at different spatial scales due to their simple and powerful model structure and compatibility with ArcGIS [17,18]. However, USLE and RUSLE were designed for gentle slopes (<18%) and when these models are applied to steep slopes, large errors occur [19,20]. On the other hand, the CSLE is developed based on the RUSLE equation, modified according to the actual conditions of Chinese soil properties, and can calculate slope factors in the range of 9–55% [19,21]. In comparison with RUSLE, the modification of CSLE conforms with the practical situation of vegetation management and soil erosion control in China [4]. CSLE considers the effect of vegetation cover on soil erosion. Vegetation cover can slow down water flow and enhance soil conservation capacity, playing a crucial role in controlling soil erosion [8]. However, the parameters in the CSLE model need to be adjusted and calibrated based on specific regions and vegetation types to improve accuracy. The model is now widely used in soil erosion prediction and assessment in China [22,23].

Current soil erosion studies are mostly based on arid and semi-arid regions as the study area, and there are relatively few studies on soil erosion mechanisms in tropical regions. Climate change leads to a shift in the dry–wet zone, which largely alters the spatial and temporal variability of hydro-climatic variables (e.g., rainfall) [24,25]. Tropical forests are one of the planet’s major climate regulators [26]. Tropical rainforest, monsoon rainforest is the zonal vegetation type of Hainan Island, located in the eastern coastal monsoon climate zone of China, which is warm and humid and disturbed by typhoons and tropical cyclones. The impact of deforestation on climate varies according to latitude, with high latitudes in the Northern Hemisphere causing a drop in temperature, and deforestation in the tropics causing a temperature rise [27]. The complex vegetation structure and high diversity of flora and fauna are the distinguishing features of tropical rainforests [28]. Tropical rainforest vegetation is evergreen with tall trees and a large thickness of deadfall layer. The tropical rainforest canopy dominates forest ecosystem functions such as carbon sequestration and sink enhancement, conservation of water resources, and regulation of regional climate change. The monitoring of vegetation types and their characteristics is essential for the management of tropical forests and their climatic conditions [29]. There are, however, differences in vegetation types and climatic conditions between arid and semi-arid regions and tropical regions. Vegetation in arid zones is dominated by dry grasses and shrubs, and vegetation dynamics are strongly influenced by precipitation. Vegetation in the semi-arid zone is generally sparse and dominated by drought-tolerant annuals and other short-growth cycle plants. Regional vegetation types and climatic characteristics lead to differences in the response patterns of soil erosion from those of the widely studied arid and semi-arid regions [30,31].

In most studies analyzing the driving factors of soil erosion, a variety of human and natural factors are often selected. These factors include precipitation, elevation, GDP, and population density, among others, to analyze the driving factors behind the simulated results obtained through soil erosion modeling. For example, Wang et al. [32] conducted a simulation and quantitative attribution analysis of soil erosion in different geomorphological types within a typical karst basin using the RUSLE model and the geodetector method. Factors such as land-use type, slope, elevation, and vegetation cover were considered. In analyzing the driving factors of soil erosion, vegetation and topographic factors are often considered more extensively. Using the RUSLE model and Geodetector method, Wang et al. [33] investigated the temporal and spatial dynamics of soil erosion in the Yellow River basin. They also assessed the contribution of vegetation cover to soil loss since the implementation of the Grain-for-Green project. Zhao et al. [34] conducted a study to identify the areas with the most severe erosion in different geomorphologic types. While these studies adequately considered factors with regional characteristics for the analysis of influencing factors, the selected factors may lack objectivity despite their representativeness. In this study, we take a comprehensive perspective of both human and natural factors and set up different scenarios to identify the dominant factors of soil erosion. We hope that this research can provide scientific references for soil and water conservation in tropical watersheds.

China is one of the most populous countries in the world and soil erosion is caused mainly by intense human activities and natural factors such as deforestation, urbanization, land-use change, regional meteorology, geology, and topography. Therefore, it has become a national environmental problem [8,35]. Tropical ecosystems with healthy soils can sustain multiple ecosystem services and provide support for local livelihoods [36]. Hainan Island is located on the northern edge of the tropics and the transition zone of dry and wet tropical climate, developing and preserving the largest tropical rainforest ecosystem in China, which plays a huge role in soil conservation ecological services [37]. Hainan Island has undergone rapid changes in land use/cover, such as the cultivation of large cash crops like rubber and betel nut, as well as rapid population growth and development of land for construction due to the policy support of the International Tourism Island and the Pilot Free Trade Zone, in the past decades [38,39]. Therefore, the above changes may lead to the destruction of tropical rainforests, increased soil erosion, and even serious damage to the tropical biodiversity and ecological environment of Hainan Island [37]. Li et al. [40] study also pointed out that, similar to other tropical regions, Hainan Island is also facing severe soil erosion. In the past, scholars conducted soil erosion research on Hainan Island together with southern China or the red soil regions [41]. Nowadays, scholars often include soil conservation services as an important ecosystem service in their ecosystem research [42]. The proposal for ecological conservation and the development strategy of building Hainan Province into an international tourist island have raised new and higher requirements for soil and water conservation. In order to promote soil erosion prevention and control, as well as the protection and rational utilization of soil and water resources, dynamic monitoring of soil erosion has been conducted in various regions of Hainan Province. According to the 2017 National Government Work Report, the entire island of Hainan has been designated as a prone area for soil and water loss. Therefore, quantifying the spatial attribution of soil erosion in tropical watersheds, revealing the patterns of change and mechanisms of soil erosion in tropical regions, holds significant theoretical and practical significance for enhancing land productivity in red soil hilly areas, effectively controlling soil erosion, and maintaining ecological security in similar regions.

2. Study Area

Hainan Island has dense tropical rainforests, crisscrossed river systems, abundant rainfall, and towering mountains; it is located at latitude 18.10°–20.10°N and longitude 108.37°–111.03°E, lies off the Leizhou Peninsula across the sea, covering an area of about 33,900 km² with 18 cities and counties and having a coastline of 1823 km [43]. The study area, as a whole, experiences a typical tropical monsoon oceanic climate, with an average annual temperature of 23.42–25.34 °C and a mean annual precipitation of 1449–2336 mm. The study area is located in the central zone of Hainan Island, which is different from the eastern coastal region characterized by strong economic development, high population concentration, and frequent human activities. It is also distinct from the western region, which has relatively dry climate conditions and less human activity. The study area contains most of the three major watersheds of Hainan Island. Among them are the Longtang hydrological station in the north on the Nandu River, the Baoqiao hydrological station in the west on the Changhua River, and the Jiaji hydrological station in the east on the Wanquan River (Figure 1). The study area also includes the origin of the three major rivers in Hainan Island, and the regional tropical rainforest is widely distributed, with a large elevation range and obvious geo-climatic features of research significance.

3. Method

This study aims to investigate the drivers of soil erosion in three major watersheds on Hainan Island from 1991 to 2020 by constructing a CSLE model and applying Slope trend analysis, Pearson correlation analysis, and spatial attribution analysis methods (Figure 2). The detailed study steps are shown in Figure 1. The objectives were (1) To analyze the annual average variation in each hydrothermal factor (rainfall, temperature, reference crop potential evapotranspiration) and aridity index using GIS technology and slope trend analysis methods, as well as the spatial variation trends multi-year. (2) To analyze the spatial correlation between soil erosion and aridity index, and to explore the effect of land-use shift changes on soil erosion using the land-use shift matrix. (3) To identify, through spatial attribution analysis, the dominant drivers influencing soil erosion change—natural, anthropogenic, or combined natural–anthropogenic drivers.

3.1. Data Sources

The meteorological data, MOD13Q1 data, land-use data, and soil data utilized in this research were sourced from different platforms. The meteorological data were obtained from the China Meteorological Data Service Center, which is part of the National Meteorological Information Center (http://data.cma.cn (accessed on 5 December 2020)). The MOD13Q1 data were acquired from the Goddard Space Flight Center, NASA (https://ladsweb.modaps.eosdis.nasa.gov/search (accessed on 10 December 2020)). The land-use data originated from the Resource and Environment Science Data Center, Chinese Academy of Sciences (http://www.resdc.cn/ (accessed on 20 December 2020)). Lastly, the soil data were obtained from the World Soil Information Database (http://westdc.westgis.ac.cn (accessed on 13 January 2021)).

3.2. Linear Regression Analysis

To investigate the spatiotemporal patterns of Hydrothermal factor variation, we used trend analysis, whose slope coefficients [10] were used to determine the trends in variations, which can illuminate long-term changed direction for precipitation, temperature, evapotranspiration, and aridity index. The trend formula is

$$Slope=\frac{n\times {\sum }_{i=1}^{n}i\times X_i-({\sum }_{i=1}^{n}i)\times ({\sum }_{i=1}^{n}i)\times ({\sum }_{i=1}^n{X}_i)}{n\times {\sum }_{i=1}^n{i}^2-{({\sum }_{i=1}^{n}i)}^2}$$

(1)

The slope is the trend of the dependent variable during a certain time series, where n is the sequence number of monitoring years, from 1 to 30; where i is the year, and X_i is the dependent variable in the year i. In general, if the slope > 0, the change in the dependent variable exhibits an upward trend, while if the slope < 0, the change in the dependent variable exhibits a downward trend [44].

Correlation coefficients were calculated between the annual Hydraulic erosion (dependent variable) and the annual aridity index (independent variables) using Pearson correlation.

$$r_{xy}=\frac{{\sum }_{i=1}^n\left(x_i-\overline{x}\right)\times \left(y_i-\overline{y}\right)}{\sqrt{{\sum }_{i=1}^n{\left(x_i-\overline{x}\right)}^2\times {\sum }_{i=1}^n{\left(y_i-\overline{y}\right)}^2}}$$

(2)

where r_xy is the correlation coefficient between the variables x and y, CC ∈ (−1,1). Here, CC represents the value of the Pearson correlation coefficient. x_i, y_i are the independent and dependent variables, respectively, and $\overline{x}$, $\overline{y}$ are the means of variables x and y, respectively.

3.3. Aridity Index

The aridity index was defined as the reciprocal of the wetness index, as determined by the Vysotskii model, and the ratio of potential evapotranspiration (ET₀) to precipitation was used to express the aridity index (AI). The aridity index is used to indicate the degree of drought in a region and further indicates the moisture status of the study area. Before calculating the aridity index, the ET₀ (reference crop) is first calculated. ET₀ is an important part of the water cycle, and energy balance. The Penman–Monteith model, revised by the Food and Agriculture Organization of the United Nations (FAO), was used to calculate the ET₀ in the study region [45]. The specific calculation is as follows.

$${ET}_0=\frac{0.408\times \left(R_n-G\right)+\gamma \times \frac{900}{T+273}\times U_2\times \left(e_s-e_a\right)}{∆+\gamma \times \left(1+0.34\ast U_2\right)}$$

(3)

$$\begin{gathered} R_n=\left(1-\alpha \right)\times \left({\alpha }_s-b_s\times \frac{n}{N}\right)\times R_a-\delta \times \left(\frac{T_{max}^4+T_{min}^4}{2}\right) \\ \times \left(0.56-0.08\sqrt{e_a}\right)\times \left(0.1+\frac{0.9n}{N}\right) \end{gathered}$$

(4)

$$G=0.14\left(T_i-T_{i-1}\right)$$

(5)

In the equations, R_n represents the net surface radiation of the reference crop canopy, MJ/(m²·d). Δ represents the slope of the saturated water vapor pressure versus temperature curve, kPa/°C. γ is the wet and dry table constant, kPa/°C. T represents the average daily air temperature, °C. G represents the soil heat flux, MJ/(m²·d). e_a represents the actual water vapor pressure, kPa. e_s represents the saturated water vapor pressure at the evaporating surface, kPa. U₂ represents the wind speed, m/s. R_a represents the top-level solar radiation, MJ/(m²·d). δ represents the Boltzmann constant. n represents the actual sunshine hours, h. N represents the maximum sunshine hours, h; a_s is the fraction of external atmospheric radiation reaching the ground under full cloud cover (n = 0). b_s is the fraction of external atmospheric radiation reaching the ground on a clear day (n = N). α is the surface reflectance and takes the value of 0.23. T_i refers to the average temperature of month i, and T_i−1 refers to the average temperature of month i−1, K [46].

$$e_s=\frac{e^0\left(T_{max}\right)+e^0\left(T_{min}\right)}{2}$$

(6)

$$e^0\left(T_{min}\right)=0.6108\times exp\left(\frac{17.27T_{min}}{T_{min}+237.3}\right)$$

(7)

$$e^0\left(T_{max}\right)=0.6108\times exp\left(\frac{17.27T_{max}}{T_{max}+237.3}\right)$$

(8)

The formula e⁰ indicates the saturation pressure at the moment of the highest or lowest temperature, kPa. T_max and T_min refer to the maximum absolute temperature and the minimum absolute temperature, respectively, in K [47]. After calculating the ET₀, the aridity index (AI) equation is as follows.

$$W=\frac{P}{{ET}_0}$$

(9)

$$AI=\frac{{ET}_0}{P}$$

(10)

In the equations, W, AI, P, and ET₀ are wetness index, aridity index, precipitation, and potential evapotranspiration, respectively. According to the original definition of the Vysotskii model, AI ≤ 0.99 is wet type, 1 < AI ≤ 1.49 is semi-humid, 1.50 < AI ≤ 3.99 is semi-arid, and AI ≥ 4.00 is arid. A positive value indicates increasing aridity, while a negative value indicates decreasing aridity and increasing wetness [48].

3.4. Transition Matrix Method

The transfer matrix method was employed to examine land-use transfer changes during the study period and to analyze the relative contribution of unchanged land use and land-use type conversion areas to long-term soil erosion changes. The general form of the transfer matrix is as follows [49]:

$$s_{ij}=\left[\begin{array}{c}{s}_{11}\\ s_{21}\\ \dots \\ s_{n1}\end{array}\begin{array}{c}{s}_{12}\\ s_{22}\\ \dots \\ s_{n2}\end{array}\begin{array}{c}{\dots s}_{1n}\\ {\dots s}_{2n}\\ \dots \dots \\ {\dots s}_{nn}\end{array}\right]$$

where s represents the area in hectares (ha), n represents the number of land-use and land-cover types before and after the transfer, and i, j (where i, j = 1,2, …, n) represent the land-use types before and after the transfer respectively. s_ij denotes the area in hectares (ha) of land type i converted to land type j.

3.5. Spatial Attribution Analysis

To distinguish the effects of human activities and climatic factors on the changes in soil erosion, three forms of soil erosion modulus (i.e., actual, predicted, and residual values) are all calculated at the pixel scale. X_A is calculated from the soil erosion model and the effect of climate change and human activities on the annual change in soil erosion can be determined by the trend of slope change (S_A) of X_A. A positive or negative value of S_A indicates that the soil erosion condition is intensified or reduced respectively in that period. X_P represents the extent to which climate change affects soil erosion, and natural influences are selected to be substituted into the model for estimation, with positive/negative values of S_P representing the increase/decrease change in the impact of climate change on soil erosion. A positive or negative value indicates an increase or decrease in climate change. X_R is the difference between equations X_A and X_P. Trends in X_R (S_R) indicate changes in soil erosion from causes other than climate change, with positive and negative values indicating an increase in soil erosion driven by anthropogenic factors and a decrease in soil erosion driven by anthropogenic factors, respectively. X_R is calculated by the following Equation (11) [50].

$$X_R=X_A-X_P$$

(11)

Six scenarios were classified according to trends in the three forms of S_A, S_P, and S_R to illustrate the driving mechanisms of climatic and anthropogenic factors on soil erosion changes in the study area (

Table 1. Distinction between different scenarios of the involvement of natural and human factors in soil water erosion changes.

SA > 0	SP	SR	SA < 0	SP	SR	The Role of Climate and Human Activities
IC	>0	<0	DC	<0	>0	Climate Dominance
IH	<0	>0	DH	>0	<0	Human activity dominated
IB	>0	>0	DB	<0	<0	Climate and human activities together

Note: IC, IH, and IB refer to increases in soil water erosion driven by climatic, anthropogenic, and climate–anthropogenic factors, respectively. Conversely, DC, DH, and DB refer to a decrease in soil water erosion dominated by each driver, respectively.

).

3.6. Soil Erosion Model

The CSLE model was developed by the RUSLE to account for the specific soil properties in China. This study employed the CSLE model to monitor soil erosion dynamics. The equation can be expressed in the following basic form [51]:

$$A=R\times K\times L\times S\times B\times E\times T$$

(12)

The unit for factor A is t·hm⁻²·a⁻¹, representing the soil erosion modulus. The units for the rainfall erosivity factor (R) and soil erodibility factor (K) are MJ·mm·hm⁻²·h⁻¹·a⁻¹ and t·hm⁻²·h·hm⁻²·MJ⁻¹·mm⁻¹, respectively. The factors L and S represent the slope length and slope steepness, and they are dimensionless. The vegetation cover and biological measures factor (B), engineering measures factor (E), and tillage measures factor (T) are also dimensionless.

To calculate and obtain these seven factors, a combination of remote sensing techniques and field surveys was employed. The specific algorithms for obtaining these factors can be referenced from the Technical Guidelines for Dynamic Monitoring of Soil Erosion in 2020 and the Classification and Grading Standards for Soil Erosion (SL190-2007).

4. Result Analysis

4.1. Spatial and Temporal Characteristics of Hydrothermal Factors

To calculate and analyze the potential evapotranspiration and drought indices of reference crops in the study area from 1991 to 2020, the study performed a presentation of the multi-year average and trend variation characteristics of hydrothermal factors including rainfall and temperature (Figure 3). The hydrothermal factors in the study area are significantly spatially heterogeneous. From the western and northern parts of the study area to the southeastern areas, the rainfall gradually increases, with variations ranging from 1449 to 2336 mm and a mean value of 1988 mm. The slope trends of multi-year rainfall reveal that only some areas within the western Changhua River basin in the study area follow a decreasing trend (negative values), while all other areas follow an increasing trend of rainfall (positive values). The annual mean temperature gradually decreases from the western and northeastern part of the study area to the southeastern region, with variations ranging from 23.42 to 25.34 °C and a mean value of 24.26 °C. The trend of slope variation in temperature is exactly opposite to the spatial distribution characteristics of annual mean temperature. Their slope values are all consistently positively distributed, with values increasing from the western and northern regions to the central region with a small range of variation (0.00018–0.00053).

The high-value area of potential evapotranspiration of the reference crop is similar to the range of high-value areas of the annual average temperature distribution in the west. The range of potential evapotranspiration minimums is similar to the area where the range of temperature minimums and the range of precipitation maximums are distributed. During 1991–2020, the multi-year average variation in potential evapotranspiration ranged from 1246 to 1611 mm, with a mean value of 1336 mm. The distribution area where the potential evapotranspiration showed a decreasing change was small. It is noteworthy that, in addition to the areas distributed in the range of the minimum annual average potential evapotranspiration, the potential evapotranspiration in the lower Nandu River region also shows a significant growth trend. In the study area, all areas belong to the wet zone (AI ≤ 0.99), except for the watershed extending from part of the western Changhua River to the Baoqiao station, which is a semi-wet zone (1 < AI ≤ 1.49). The trend of the aridity index is polarized, with most of the Nandu River and the lower Wanquan River basin showing a decreasing trend change, and the Changhua River as a whole and a few other areas showing an increasing trend change. That is, the wetness increases in the humid regions and the aridity increases in the semi-humid regions.

The spatial distribution of climatic factors reveals that the southeastern region extending outward from Qiongzhong, with high rainfall, low temperature, and low potential evapotranspiration, is the most humid area in the study area. The middle and lower reaches of the Changhua River basin, with the Baoqiao hydrological station as the node, show low precipitation, relatively high temperature, and high evapotranspiration, and are also the concentrated distribution area of the semi-humid zone in the study area. The spatial trends of the climate factors in the study area are intuitively obvious and the spatial distribution is consistent with previous studies [43].

4.2. Correlation Analysis of Soil Erosion and Aridity Index

In exploring the role of natural factors on soil erosion drivers, it is necessary to seek an integrated factor to characterize the state of interactions and constraints among natural factors to reflect watershed differences. Furthermore, it is necessary to differentiate it from human effects. Therefore, this composite factor is mainly climatic factors such as rainfall, temperature, wind, and solar radiation. The interactions and constraints of these factors produce the wet and dry conditions of the land surface, which can be indicated by the aridity index. Therefore, the correlation between the aridity index and soil erosion is explored to analyze the spatial distribution pattern of soil erosion influenced by meteorological factors. In general, soil water erosion showed a marked increasing trend (S = 3.48/a) (Figure 4). The aridity index was negatively correlated with soil erosion in most areas of the Nandu River and Wanquan River basins in the study area, with correlation values of −0.44 to 0.00 for a wide distribution (Figure 4). The areas where the aridity index was positively correlated with soil water erosion were relatively concentrated in the Changhua River basin. During the period 1991–2020, the regional differentiation of the increasing and decreasing trends of drought indices in the study area was evident in terms of spatial distribution (Figure 3). Combining the criteria of the aridity index for classifying dry and wet areas, we obtained those humid areas (AI ≤ 0.99) in the study area that showed a negative correlation with soil water erosion, while semi-humid areas (1 < AI ≤ 1.49) showed a positive correlation with soil water erosion.

The correlation of soil water erosion with the aridity index in the study area was statistically analyzed by dividing different sub-watersheds and based on different land uses (Figure 5). The average correlation value between soil water erosion and aridity index was negative (CC = −0.12) except for the Changhua River II-1 watershed (CC = 0.01644), where all other watersheds had negative values. The basin with the most significant negative correlation was the Wanquan River III2 basin (CC = −0.1954), and the weakest was the Nandu River I-2 basin with only −0.06072. The correlation between land-use-based aridity index and soil water erosion was predominantly negative. The most prominent negative correlation was for paddy fields (CC = −0.1765), followed by rural building land (CC = −0.1283). The weaker negative correlation was for bare land (CC = −0.04785), followed by orchards (CC = −0.05499).

4.3. Analysis of the Influence of Human Activities on Soil Water Erosion

The relative contribution status of interannual variation in soil water erosion in areas with unchanged land-use types and in areas with changed land-use types was analyzed through an in-depth discussion of the effects of human activities on soil water erosion changes. The land-use transfer matrix was used to identify land-use changes for the 31 years between 1990 and 2020. As reflected by the land-use (cover)-type transfer matrix, the largest percentage of all land uses in the study area was transferred to woodland (73.08%), followed by cropland (18.60%) (

Table 2. Transfer matrix of land-use (cover)-type changes in the study area from 1990 to 2020 (%).

1990/2020	Grassland	Cropland	Construction Land	Woodland	Bare Land	Water Area	Sum
Grassland	0.28	19.44	1.62	75.60	0.01	3.05	4.79
Cropland	0.51	70.60	6.31	16.37	0.03	6.17	18.60
Construction Land	1.67	35.86	40.90	17.05	0.06	4.46	1.02
Woodland	0.25	35.36	1.87	60.64	0.01	1.87	73.08
Bare land	0.81	24.93	1.49	9.32	0.00	63.46	0.33
Water area	0.33	16.70	1.02	9.14	0.02	72.79	2.19
Sum	0.32	40.72	3.06	51.38	0.02	4.51	100.00

). After identifying the regions where different land-use types have undergone transfer changes and the regions where land-use types have remained unchanged using the land-use transfer matrix, the areas are overlaid with soil erosion to identify the relative contributions of different regions to soil water erosion (

Table 3. Relative contribution of unchanged land-use and land-use-type shift areas to multi-year changes in soil water erosion (%).

Unchanged Area	Area Percentage	Type of Change	Area Percentage
Cropland	8.91	Cropland–Woodland	41.25
Woodland	90.67	Grassland–Woodland	21.60
Water area	0.04	Woodland–Water area	12.80
Construction Land	0.36	Grassland–Cropland	9.06
Grassland	0.02	Cropland–Construction land	6.64

).The relative contribution of unchanged land-use areas to interannual variation in soil water erosion was in the order of woodland (90.67%) > cropland (8.91%) > construction land (0.36%) > water area (0.04%) > grassland (0.02%) > bare land. The relative contributions of land-use change zones to the interannual variation in soil water erosion are, in descending order, cropland–woodland (41.25%) > grassland–woodland (21.60%) > woodland–water area (12.80%) > grassland–cropland (9.06%) > cropland–construction land (6.64%) (Table 3). The shift in land-use cover to woodland may be based on the construction of Hainan Tropical Rainforest National Park (2019) in the study area and the national project to return farmland to woodland since 1999 [52].

4.4. Spatial Attribution Analysis of Soil Erosion Driving Mechanisms

The scenarios driven by a combination of natural and anthropogenic factors have the most pronounced impact on the variation in soil erosion (Figure 6). The combined effect causes a large spatial distribution of the change in soil water erosion growth with a contribution of 53.56% (IB), followed by 12.77% (IH) for human activities and only 3.69% (IC) for climate domination. The spatial distribution of the joint effect causing changes in soil water erosion reduction was relatively wide, with a contribution of 21.74% (DB), followed by anthropogenic-led 4.92% (DH) and climate-led 3.31% (DC). An analysis of the spatial attribution results of soil water erosion under the different scenarios combined above found that human activity factors contribute more to soil erosion than natural factors.

From the watershed division, the increase in soil water erosion led by anthropogenic activities and co-led by anthropogenic climate in the Nandu River I-1 watershed was prominent, with 56.13% and 25.26%, respectively (

Table 4. Relative contribution of each watershed to soil water erosion under different scenarios (%).

Watersheds/Scenarios	IC	IH	IB	DC	DH	DB
I-1	22.70	56.13	25.26	52.82	21.47	46.63
I-2	17.99	7.33	15.18	6.59	12.75	11.79
I-3	6.50	14.27	5.20	9.27	2.03	10.82
II-1	8.32	4.59	12.91	10.23	17.70	10.06
II-2	9.50	7.01	9.49	6.07	9.88	9.05
III-1	1.74	2.67	1.57	2.94	3.21	2.56
III-2	14.61	6.13	12.25	6.35	15.61	4.34
III-3	18.64	1.88	18.15	5.72	17.34	4.74

). Not only that, the distribution area of this watershed is larger than other watersheds in all three scenarios of soil water erosion reduction. Other watersheds driving significant changes in increased soil water erosion are the Wanquan River III-3 watershed (joint human–climate: 18.15%) and the Nandu River I-2 watershed (joint human–climate: 15.18%). The Nandu River I-2 and I-3 basins and the Changhua River II-1 basin accounted for a relatively large share of the change in soil water erosion reduction driven by joint human–climate effects, at 11.79%, 10.82%, and 10.06%, respectively.

5. Discussion

Understanding the relationship between drivers and changes in soil erosion can help us better assess soil erosion dynamics [53]. The analysis of driving mechanisms is a prerequisite for controlling, predicting, and preventing soil erosion problems. Based on the spatial attribution results of soil water erosion under the above scenarios, it is important to further explore the reasons for the increase/decrease in soil water erosion changes in the study area due to human activities or climatic factors, and to compare and contrast the differences in soil erosion driving mechanisms between the tropics and other climatic zones and identify the characteristics of soil erosion in the tropics. In addition, a detailed analysis of the causes of soil water erosion increase and decrease under the dominance of climatic factors, anthropogenic factors, and combined climate–anthropogenic activities will be carried out to provide countermeasures for soil water erosion management in the study area, aiming to provide a reference for other similar areas.

5.1. Divergence of Soil Erosion Driving Mechanisms in Different Climatic Zones

The dominant factors driving soil erosion vary between climatic zones. The investigation compares differences in soil erosion driving mechanisms in three particular regions of China, the upper Yellow River, Yunnan Province, and the Qinghai–Tibet Plateau, where climatic types span multiple climatic zones and where the characteristics of climatic and geographic elements differ significantly from those of the three watersheds of Hainan Island. Most of the upper reaches of the Yellow River have a highland continental climate. Li et al. [54] pointed out that soil erosion was on the rise in the upper reaches of the Yellow River during the study period due to natural factors. After the implementation of the policy of returning farmland to forests after 2000, soil erosion due to human activities has been on the decline, and it can be seen that forests have a stronger ability to control soil erosion than grasslands. Rainfall precipitation in Yunnan Province is mainly influenced by the Indian Ocean monsoon and has a high spatial variability under the influence of complex topography. However, its vegetation is more sensitive to the driving effect of soil erosion than rainfall [4]. In contrast to previous historical years when land-use changes drove changes in soil erosion increases and decreases, in recent years, the increase in erosion on the Tibetan Plateau has been driven mainly by extreme rainfall events [55]. The contribution of anthropogenic factors to changes in soil water erosion is higher than natural factors in the three watersheds of Hainan Island. Clearly, land development, urbanization, and agricultural expansion, as well as the construction of reservoirs and roads, have altered water flow paths and land-use patterns, increasing the risks of soil erosion, and significantly impacting changes in soil erosion [56]. While the driving role of natural factors, such as precipitation and other meteorological factors, in soil erosion cannot be underestimated, the role of human factors in tropical watershed water and soil loss should also be of concern.

Most studies have only distinguished between human activities and natural factors to examine the driving factors influencing changes in soil erosion in each study area. Six scenarios were distinguished in this study to identify the main drivers of soil erosion change in the study area. The climate–anthropogenic co-driven scenario is the one that drives the dominance of soil erosion changes in the three major watersheds of Hainan Island. Consequently, the causes of the spatially driven mechanisms influencing soil erosion variability in the study area are relatively complex compared to other regions, and there is an urgent need to further unravel the problems and causes of the differences in hydraulic erosion in the basin that led to its existence.

5.2. Analysis of the Causes of Soil Erosion Changes under the Action of Each Dominant Factor

The mechanisms driving soil water erosion in the major catchments and their drivers are described according to the spatially attributed distribution of soil water erosion shown in Figure 6, taking into account the relative contribution of each catchment (Table 4). The study also identifies areas where the spatial attribution contribution of soil water erosion is significant and analyzes which factors play a dominant role and how they drive increased or decreased changes in soil water erosion. There are differences in the drivers that play a dominant role in the increase and decrease in soil water erosion in the three catchments, but the human activity factor is more influential in the decrease change. Below are the details of the three main catchments.

(1) The Nandu River basin demonstrated a prominent proportion of anthropogenic-driven changes in soil water erosion increase (IH = 77.73%) and climate–anthropogenic-driven change decrease (DB = 69.24%). This may be related to the flat topography of the lower reaches and the distribution of large hydrological measures such as the Songtao Reservoir upstream, development and construction projects, and the high level of disturbance due to urbanization development.

(2) The climate of the Changhua River basin is unique compared to other basins, and the impact of the semi-humid climate due to geographical factors on soil erosion is different from the causes of soil erosion in other basins. This, combined with human efforts to improve the regional economy, has led to a combined climate–human driver of soil erosion (IB = 22.40%). However, for the sake of ecological protection, human activities are driving the change in soil erosion reduction (DH = 27.58%).

(3) The Wanquan River watershed is the wettest in the study area and has the largest percentage of climatic factor-driven changes in increased soil water erosion (IC = 34.99%). Therefore, at the level of continuous wetness, intervention in watershed soil and water conservation through human activities resulted in a significant change in the decrease in soil water erosion driven by human activity factors (DH = 36.16%).

In the study area as a whole, the wetter areas are subject to a combination of climate–anthropogenic-driven changes in increased soil erosion, resulting in more pronounced changes in soil water erosion than in other areas. In the context of considering the influence of natural conditions such as climate and topography, people use water resources from some rivers and reservoirs for hydraulic engineering measures such as flood control and storage, inter-basin water transfer, and hydropower generation [57]. This results in a spatial distribution of changes in soil water erosion reduction driven by a combination of natural and human activities (DB). In terms of univariate drivers, while the combined climate–anthropogenic (IB) effect drives a wide range of increasing changes in soil erosion in the study area, the areas where human activities (IH) are the dominant driver of increasing changes in soil erosion are very close to the former distribution. Furthermore, near the dense distribution areas of climate–anthropogenic-driven changes in soil erosion increase, it can be seen that single-factor-dominated changes in soil erosion decrease have large, clustered distribution areas near both reservoirs and river mainstems. For example, at the downstream main stream of Nandu River (I-1), near Songtao Reservoir (I-3), near Daguanba Reservoir (II-1), downstream of Changhua River (II-1), near Niu Luling Reservoir (III-3), etc. These two distributional phenomena further illustrate that both decreasing and increasing changes in soil water erosion are driven by the dominant role of human activities. However, the uncontrollable factors of natural conditions still cause some soil erosion. However, if people can combine climate, topography and other natural conditions, rational use of water resources, and vegetation restoration measures, they can properly play a role in reducing soil water erosion, and vice versa.

5.3. Causes and Countermeasures of Watersheds with Dense Spatial Distribution

The study presents a comprehensive analysis of the spatial characteristics of the Nandu River Basin, Wanquan River Basin, and Changhua River Basin. The Nandu River, the largest river on Hainan Island, has witnessed the construction of hydropower facilities and reservoirs in the upper reaches to mitigate drought disasters and support agricultural irrigation downstream. Notably, the Songtao Reservoir, the largest hydroelectric project in Hainan Province, exhibits soil erosion in its reservoir area. The expansive basin of Songtao Reservoir experiences abundant rainfall, a humid climate, low potential evapotranspiration, intricate topography, diverse soil types, and dense tropical rainforest vegetation. However, anthropogenic factors such as deforestation, construction activities for economic development, and the monsoon climate of tropical islands contribute to soil erosion, particularly during heavy rain events that transport eroded soil into water bodies. The majority of economic forests and cultivated land within the Songtao Reservoir watershed lack sufficient soil and water conservation measures, with economic forests comprising a significant portion (35%) of the vegetation. The conversion of woodland to other land-use types between 1991 and 2020 has exacerbated problems, including a reduction in soil and water conservation capacity. Therefore, it is crucial to implement rigorous soil and water conservation measures, enhance vegetation cover, and adopt alternative practices such as terracing or fish scale pits for rubber forest plantations to replace traditional downhill cultivation methods. Additionally, soil and water conservation efforts should not be neglected in downstream plain areas where development, construction projects, and agricultural irrigation take place, apart from the reservoir areas.

The Wanquan River Basin, although the smallest in terms of land area among the three basins, experiences concentrated and intense precipitation, which is the main driver of soil water erosion changes in the area. The high rainfall reaches a threshold where vegetation becomes overwhelmed, leading to oversaturated soil and reduced water absorption capacity. Insufficient regulation and storage capacity of river channels in the basin, coupled with steep slopes, result in significant surface runoff in the mountainous regions, causing rain-induced floods and increasing the risk of soil erosion. Among the three sub-basins within the Wanquan River Basin, the Wanquan River III-3 watershed exhibits the highest increase in soil water erosion (32.72%), followed by the Nandu River I-2 watershed (23.64%) and the Wanquan River III-2 watershed (23.47%), under the combined influence of climate and human activities. Consequently, the analysis focuses on understanding the causes of soil water erosion in the Wanquan River III-3 basin. The Wanquan River III-3 watershed includes the Niu Luling Reservoir, an important ecological area and a key site for public welfare forest protection. It serves as a crucial water source in central Hainan Province. This watershed experiences the wettest climate in the study area and falls within the minimum aridity index range (Figure 3g). It also shows the highest percentage increase (2.31%) in soil water erosion driven by climate factors. However, significant ecological damage and water source security issues are present in the reservoir area. Long-standing practices such as relay cropping of betel nut and other economic forests, destruction of natural forests, and encroachment upon forest land by natural secondary forests have led to the expansion of planted forests, partial loss of original vegetation, destruction of biodiversity and ecosystem functions, and increased soil water erosion. Therefore, it is essential for relevant management authorities to actively enforce regulations and promote legal knowledge dissemination to address these challenges.

The Changhua River basin, located at a higher elevation compared to other basins in the study area, experiences lower rainfall, higher temperatures, and high potential evapotranspiration. It contains a semi-humid zone, which is rare in the study area (Figure 1 and Figure 3). Extensive afforestation, agriculture, urbanization, and river engineering activities have significantly altered the basin’s surface characteristics. The increasing aridity in the watershed makes it highly susceptible to soil erosion caused by sudden heavy rainfall and flooding. Analysis of a 31-year trend reveals distinctive distribution patterns of soil water erosion along the topography and water system (Figure 4). The combined impact of climate and human activities accounts for a relatively significant proportion (22.90%) of the increase in soil water erosion in this watershed area (II-1 and II-2 combined), displaying unique spatial characteristics. Dryland farming practices in the basin contribute to soil and water conservation and facilitate water storage. However, due to the limited arable land resources and high altitude, reliance on mountain biological resources for livelihood has led to indiscriminate deforestation and excessive clearance of forest land, resulting in high sand content in rivers and soil erosion [58]. It is vital to cease deforestation and protect soil conservation forests along both sides of the main channel of the Changhua River. Furthermore, cultivation on steep slopes exceeding twenty-five degrees should be prohibited within the river management area. Activities endangering the stability of riverbanks and embankments must be avoided, prioritizing effective soil erosion prevention measures.

6. Conclusions

From 1991 to 2020, the study area experienced a polarized trend of dryness and wetness. Humid areas saw an increase in wetness, while semi-humid areas became more arid. Overall, humid areas had a negative correlation with soil water erosion, while some semi-humid areas had a positive correlation. The correlation between soil water erosion and the aridity index was generally negative, with paddy fields showing the strongest negative correlation. Land-use changes in the study area were mainly characterized by the conversion of land to woodland and cropland. Forest land had the highest impact on inter-annual variation in soil water erosion. The changes in soil erosion are influenced by a combination of natural and human factors. The dominant factors driving soil water erosion vary across the three basins, with human factors playing a significant role in reducing soil erosion. In the Nandu River Basin, human activities drive an increase in soil water erosion, while a decrease in soil erosion due to a combination of climate and human factors is observed. In the Changhua River Basin, soil water erosion is increased driven by the combined effects, and a decrease driven primarily by human activities. In the Wanquan River Basin, soil erosion increases due to climatic reasons, while human factors lead to a reduction in soil erosion. This study endeavors to identify the drivers of soil erosion in watersheds containing a convergence of features such as high elevation, heavy rainfall, high vegetation cover, dense distribution of rivers, and water source origins. The study is of great theoretical and practical importance for improving land productivity in red soil hilly areas, effectively controlling soil erosion and maintaining ecological security in similar areas.

This study identifies that climate–human activity scenarios jointly contribute to the largest proportion of the observed increase in soil water erosion. However, the study primarily describes human activities through changes in land use. Therefore, it is crucial to conduct research on the impact and contribution of extreme climate events, major engineering projects, and other factors causing soil water erosion. Such studies are of significant importance for identifying the forces driving soil water erosion.