The Potential of Ecological Restoration Programs to Increase Erosion-Induced Carbon Sinks in Response to Future Climate Change

Abstract

Erosion-induced carbon sinks are a wild card in the global carbon budget. Soil erosion results in aggregate carbon sequestration by reforming organic–inorganic complexes at depositional areas and plant reserves. The carbon sinks at the depositional sites are rarely considered in the prediction of erosion-induced carbon sink dynamics. The effects of large-scale ecological restoration programs (ERPs) in subtropical regions on soil carbon sinks are still unclear. This study analyzed the potential effects of ERPs on erosion-induced carbon sinks in a red soil hilly region (RSHR) from 2030 to 2060. Based on a land use dataset and two climate scenarios of moderate (RCP4.5) and high emission paths (RCP8.5), three land use change (LUC) patterns were designed: an Ecological Restoration (ER) pattern; a Business-As-Usual (BAU) pattern; and a No LUC pattern. The results of the ER pattern and BAU pattern were compared with those of the No LUC pattern to reflect the role of ERPs in reducing erosion and increasing erosion-induced carbon sinks. The results indicated that the erosion-induced carbon sinks of forestland increased (58 kg km⁻²) in the BAU pattern under the RCP8.5 scenario and erosion-induced carbon sinks of cropland increased (39 kg km⁻²) in the ER pattern under the RCP8.5 scenario. In RCP4.5 and RCP8.5, the erosion-induced carbon sinks of the RSHR increased by 210 Tg and 85 Tg from 2030 to 2060, respectively (1 Tg = 10¹² g). The average annual erosion-induced carbon sink accounted for 3.84% and 1.41% of the annual average carbon sequestration of terrestrial ecosystems, respectively. Neither the BAU pattern nor the ER pattern achieved the purpose of increasing grassland carbon sinks induced by soil erosion. Therefore, the focus of future ERP optimization should be to increase grassland carbon sinks. Our study provides new evidence for research into erosion-induced carbon sinks to mitigate global climate change and a scientific basis for increasing erosion-induced carbon sinks in croplands, forestlands and grasslands in the RSHR of southern China.

1. Introduction

The CO₂ sequestration caused by soil erosion contributes to 5–20% of the global terrestrial carbon sink [1]. However, the carbon cycle disturbance related to atmospheric vertical carbon flux caused by erosion is often ignored in the estimation of a global carbon budget [2,3]. Soil erosion results in aggregate carbon sequestration by reforming organic–inorganic complexes at depositional areas and plant reserves. Deep carbon sequestration in sediment areas can lead to soil organic carbon (SOC) accumulation [4]. Global warming and soil erosion are mutually reinforcing. Soil carbon sequestration plays a key role in mitigating the positive feedback between the climate and terrestrial carbon [5]. Carbon changes in soil erosion are ignored because they are rarely included in the current generation of climate models [6]. The empirical models for measuring soil erosion carbon change vary with the time period and land cover. Such differences can lead to biased predictions of future terrestrial ecosystem carbon sinks. This creates uncertainty in the quest of China to achieve carbon neutrality by 2030 to 2060 through ecosystem sinks as a complementary measure to energy reduction.

Large-scale ecological restoration programs (ERPs) (e.g., the Green for Grain Project and Natural Forest Protection Project) on the land surface affect the carbon change of soil erosion. One of the processes of ERPs that can significantly alter the soil–atmosphere carbon exchange is a large-scale landscape change in terrestrial ecosystems, which in turn affects the deposition process of soil erosion [7,8]. Land use decisions are the key drivers of landscape and land use changes (LUCs) [9,10]. The conversion of marginal croplands to ecological land use (e.g., forestland and grassland) and the adoption of appropriate management practices on agricultural lands can reverse degradative trends and enhance SOC sequestration [11]. Therefore, land use plays a key role in increasing soil carbon sinks, changing the vertical carbon flux between the erosive soil deposition area and the atmospheric carbon pool, mitigating global warming [12]. The ERPs balance the ecosystem services through LUCs, including enhancing soil carbon sinks [13]. Changes in SOC are of particular concern because they can indicate the extent of the restoration of degraded soil systems [14]. It affects climate conditions by regulating greenhouse gas emissions [15]. However, there are few studies on optimizing ERPs of landscape vegetation in erosive soil deposition areas [6,14].

The main application of incorporating the carbon change of soil erosion into current climate models is a scenario analysis. A scenario analysis is an important method used to obtain socioecological driving forces and thus to better understand their potential future effects [16,17]. Future LUC patterns can be achieved by LUC modeling based on different scenarios (e.g., business-as-usual and environmental protection scenarios) [18]. As LUCs are one of the major driving force factors for soil conservation and SOC imports [19,20,21], the increase in soil carbon sequestration can be achieved by applying specific management plans, practices and measures. The understanding of, or insight into, the increase in soil carbon sequestration can be achieved by comparing the impact of LUCs on erosion-induced carbon sinks under different LUC scenarios [22,23]. Global climate change will increase the intensity of soil erosion caused by both rising temperatures and excessive rainfall [24,25]. Precipitation plays an important role in soil carbon sinks [26]. There are a few studies that have predicted erosion rates and the carbon stock by modeling changes in global water erosion using the three alternative (RCP2.6, RCP4.5 and RCP8.5) Shared Socioeconomic Pathway and Representative Concentration Pathway (SSP-RCP) scenarios [27]. Climate change or LUCs are often considered to be single factors, which is difficult for decision-makers to obtain land use planning to deal with climate change [28,29,30]. There is a lack of knowledge to understand the link between increased soil carbon sequestration and climate change based on a reduction in soil erosion and an increase in erosion-induced carbon sinks under different LUCs in ERPs [31]. Understanding the magnitude and rate of change of carbon exchange processes between erosive soil deposits and the atmosphere is very important for improving model-based predictions of carbon sinks.

As mentioned above, the underlying assumption is that ecological restoration programs have the potential to increase erosion-induced carbon sinks in response to future climate change. To fill this knowledge gap, we chose the red soil hilly region (RSHR) of southern China as the study area. The three LUC patterns were designed and simulated using a Future Land Use Simulation Model (FLUS) and the RCP4.5 and RCP8.5 scenarios were scaled down. These scenario outputs were used as the input parameters to simulate soil erosion and erosion-induced carbon sinks at the depositional sites. The future optimization of soil erosion and erosion-induced carbon sinks in croplands, forestlands and grasslands under two climate scenarios based on three LUC patterns under ERPs was discussed. The results substantially expand the understanding of the role of climate change and land use as drivers of erosion-induced carbon sinks at the depositional sites in the RSHR.

2. Materials and Methods

2.1. Study Area

The RSHR is located in southern China (107°49′25″–123° E, 21°22′–31°19′ N) and has a total land area of 796,000 km² (Figure 1). This region has a subtropical humid monsoon climate with an average annual precipitation of about 1500 mm. The landform is dominated by mountains [32]. Anthropogenic pressure combined with heavy precipitation events and mountain landscapes have resulted in severe soil erosion and related ecological problems [33,34]. Therefore, a larger number of ERPs (e.g., the Green for Grain and Natural Forest Protection projects) have been implemented in this area [35]. Due to the implementation of ERPs, land use and land cover have changed dramatically in the last 30 years, affecting the ecological system services (e.g., carbon sequestration) of the RSHR.

2.2. Model LUC Patterns

The land use pattern has been interpreted in our previous research [35,36]. The land use classification, including forestland, grassland, cropland, construction land, unused land and water area, has an accuracy of over 85%. In order to simulate different LUC patterns in the RSHR from 2030 to 2060, three scenarios (Ecological Restoration (ER), Business−As−Usual (BAU) and No LUC) were proposed in this study based on the rate of land use change from 1985–2015. For the ER pattern, the forestland and grassland were set to increase insignificantly and the cropland was set to decrease insignificantly based on the pattern from 1985 to 2000. For the BAU pattern, the forestland and grassland were set to increase significantly and the cropland was set to decrease significantly based on the pattern from 2000 to 2015. The land use types of 2015 were used in the No LUC pattern. The LUC patterns from 1985 to 2000 and from 2000 to 2015 were analyzed using a Future Land Use Simulation Model (FLUS) in our previous research [36]. See the “Material and Methods” part of [35,36] for further details. The data source of this study can be found in the Supplementary Materials.

2.3. Shared Socioeconomic Pathway and Representative Concentration Pathway Scenarios

The Fifth IPCC Assessment Report in 2014 assesses projections for the 21st century from the Shared Socioeconomic Pathway and Representative Concentration Pathway (SSP−RCP) scenarios of a new climate model [37]. The SSP−RCP scheme is a spatial–temporal dependent trajectory of greenhouse gases and pollutant concentrations caused by human activities, including LUCs. The SSP−RCP scheme considers different levels of emissions, including SSP2−RCP4.5 and SSP5−RCP8.5.

The SSP2−RCP4.5 scenario can be thought of as a “middle way” scenario where the historical pattern of development remains unchanged throughout the 21st century. It assumes that the radiative forcing is stable at 4.5 W m⁻² and does not exceed this value until 2100. An important feature is the initial loss of forests of about 43 million hectares from 2000 to 2050 followed by an increase of about 331 million hectares between 2050 and 2100.

The SSP5−RCP8.5 scenario is referred to as “fossil fuel development”. According to the SSP5−RCP8.5 scenario, the global cropland area will significantly expand between 2010 and 2100 (300 million ha, equivalent to about 20%, mainly for grazing and forests, in response to a strong growth in demand for food and feed).

2.4. Statistical Downscaling of the RCP4.5 and RCP8.5 Scenarios

Predictions of climate change are usually based on global scales, usually operating at a 150–300 km resolution [38], which means that predictions on a global scale do not reliably fit the predictions of local regions. The Statistical Downscaling Method (SDSM) was used to generate precipitation outputs at regional spatial scales, incorporating a weather generator and multiple linear regressions [39]. Based on the ArcGIS platform, this paper established the correlation between the historical rainfall data from 1990 to 2009 and the historical rainfall data simulated by CanESM 5.0 in the RSHR combined with the future rainfall data simulated by CanESM 5.0 in 2030 and 2060. The rainfall data for 2030 and 2060 were obtained by statistical downscaling.

2.5. Scenario Simulation of Soil Erosion

The rainfall data, soil attribute data, slope length data, biological measurement data, engineering measurement data and tillage measurement data were used to simulate soil erosion in the RSHR [33]. The rainfall data included historical rainfall data and future rainfall data. Historical daily rainfall was observed in 1990–2009 from 50 selected meteorological sites in the RSHR as well as the daily rainfall data from CanESM 5.0. All were converted into monthly rainfall data. The future rainfall data included the monthly mean rainfall data for 2030 and 2060 in CanESM 5.0.

The Chinese Soil Loss Equation (CSLE) was used to quantify the soil erosion. Containing five parameters, the equation can be expressed as follows [40]:

M = R × K × LS × B × E × T

(1)

where M is the annual soil loss per unit area, R is the rainfall erosivity, K is the soil erodibility, LS is the slope length steepness, B is the biological measurement factor, E is the engineering measurement factor and T is the tillage practice factor.

In the simulation of soil erosion under future climate scenarios, the R value was calculated based on downscaling the rainfall data. The soil erosion modulus in 2030 and 2060 was obtained according to the CSLE. The other five factors all chose the calculation results of 2015.

2.6. Erosion−Induced CO₂ Flux Equation

As shown in Figure 2, the simplified slope of the red soil hilly region was characterized by tree planting at the top, grass planting at the waist and cropland at the bottom. The bottom of the slope was the deposition region of erosion, which was the area of agglomeration and carbon fixation in the reformation of the organic–inorganic complex and the deep burial of the carbon deposition in the sediment deposition areas such as the alluvial plain, reservoir and seabed, leading to the accumulation of SOC [4]. Estimates of the dynamic replacement of the SOC input from the erosional area to the deposition area (D₁) were determined based on the erosion rate, NPP (net primary production), soil carbon content and carbon pool turnover rate. The three fluxes were calculated and their values were added up for each grid; the total flux at the regional scale was obtained by a summation over all grids. The erosion−induced CO₂ flux at the depositional area (D₂) was estimated as the CO₂ emission from the newly buried carbon−rich topsoil. The erosion−induced CO₂ flux during sediment transport (D₃) was determined to be the difference between the CO₂ emissions before and after erosion [7].

Instead of choosing the first year since erosion as the beginning of the simulation period, the N−th year after erosion was selected as the start point. We assumed that the erosion did not exert any impact on the original CO₂ exchange process. The soil carbon composition/decomposition and lateral movement of the organic carbon were simulated as two independent processes with the modeled carbon storage being ${\mathrm{C}}_{\mathrm{unc}}$ (gC m⁻²). The coupled carbon storage (${\mathrm{C}}_{\mathrm{coup}}$, gC m⁻²) was then modeled, including the impact of erosion on the CO₂ sequestration. The difference in carbon storage under the two circumstances was regarded as the erosion−induced CO₂ flux in the erosional area, D₁ (gC y⁻¹). The parameters ${\mathrm{C}}_{\mathrm{unc}}$, ${\mathrm{C}}_{\mathrm{coup}}$ and D₁ were obtained, respectively, from:

$${\mathrm{D}}_1=\frac{{\mathrm{C}}_{\mathrm{coup}}{ - \mathrm{C}}_{\mathrm{unc}}}{\mathrm{T}}{\mathrm{A}}_{\mathrm{ero}}$$

(2)

$$\frac{{\mathrm{dC}}_{\mathrm{unc}}}{\mathrm{dt}}{=\mathrm{I}}_{\mathrm{B}}{ - \mathrm{K}}_0{\mathrm{C}}_{\mathrm{unc}}$$

(3)

$$\frac{{\mathrm{dC}}_{\mathrm{coup}}}{\mathrm{dt}}{=\mathrm{I}}_{\mathrm{B}} - \left({\mathrm{K}}_0{+\mathrm{K}}_{\mathrm{E}}\right){\mathrm{C}}_{\mathrm{coup}}{+\mathrm{C}}_{\mathrm{bel}}{\mathrm{V}}_{\mathrm{ero}}$$

(4)

where ${\mathrm{C}}_{\mathrm{unc}}$ and ${\mathrm{C}}_{\mathrm{coup}}$ are the carbon contents within the corresponding layers in the original soil profile, ${\mathrm{I}}_{\mathrm{B}}$ (gC m⁻² y⁻¹) is the carbon input into the soil and ${\mathrm{K}}_0$ (y) is the turnover rate of the soil carbon with respect to decomposition in the absence of erosion. ${\mathrm{K}}_{\mathrm{E}}$ (y) is the erosion rate of the soil carbon, obtained by calculating the ratio of the soil erosion rate (m y⁻¹) to the depth of the carbon in the top soil layer, which dominates erosion. Both ${\mathrm{C}}_{\mathrm{unc}}$ and ${\mathrm{C}}_{\mathrm{coup}}$ are effectively carbon storage in the top layers, considering that there is no difference in the coupled and uncoupled carbon storage in the deeper layers. These levels are seldom affected by soil erosion and provide no contribution to the erosion−induced CO₂ sink from the deeper layers.

$${\mathrm{C}}_{\mathrm{bel}}{/\mathrm{C}}_0{=\mathrm{C}}_{\min }{/\mathrm{C}}_0+\left({1 - \mathrm{C}}_{\min }{/\mathrm{C}}_0\right){\mathrm{e}}^{{-\mathrm{kV}}_{\mathrm{ero}}}$$

(5)

$$\mathrm{k}=\ln (-({0.01\mathrm{C}}_0{ - \mathrm{C}}_{\min })/{0.99\mathrm{C}}_0)$$

(6)

where ${\mathrm{C}}_0$ (g m⁻³), ${\mathrm{C}}_{\min }$ (g m⁻³) and k (1 m⁻¹) can be determined from the measurements of the carbon concentration in the different layers of the vertical soil profiles. ${\mathrm{C}}_0$ is averaged over the whole top layer (and is, therefore, uniform within the top layer and equal to ${\mathrm{C}}_{\mathrm{bel}}$ when t = 0). ${\mathrm{C}}_{\min }{/\mathrm{C}}_0$ is set to 0.01. The k (1 m⁻¹) is attenuation coefficient of the SOC profile.

As the eroded soil is deposited, newly buried carbon−rich soil releases CO₂ into the atmosphere [4]. The CO₂ flux caused by newly buried SOC mineralization in the sedimentary area is expressed as D₂ (gC y⁻¹).

$${\mathrm{D}}_2={\mathrm{C}}_{\mathrm{SOC}\text{-}\mathrm{surf}}{\overline{\mathrm{V}}}_{\mathrm{ero}}{\mathrm{K}}_{0\text{-}\mathrm{subsurf}}(1-\mathrm{SDR})$$

(7)

where ${\overline{\mathrm{V}}}_{\mathrm{ero}}$ is the mean erosion rate of the grid, ${\mathrm{K}}_{\mathrm{O}\text{-}\mathrm{subsurf}}$ is the turnover rate of the subsoil layer and SDR is a conceptual parameter defined as the ratio of the total sediment exported out of the grid to the total eroded soil within the grid. The turnover rate exponentially decreases with the depth:

$${\mathrm{K}}_{0\text{-}\mathrm{Z}}{=\mathrm{K}}_0{\exp (-\mathrm{u}}_{\mathrm{r}},\mathrm{z})$$

(8)

where ${\mathrm{u}}_{\mathrm{r}}$ is set to 2.6 and the decomposition rate of the newly buried SOC is 40–60% of the top layer, noting that z is usually within the range of 0.2–0.3 m.

We evaluated the erosion−induced CO₂ flux during sediment transport (D₃) by assuming that 63% more SOC was degraded into CO₂ in water than in the soil layers [1].

Table 1. Summary of parameters used in equations for calculating the carbon flux.

Parameter	Carbon Flux	Unit	Description	Equation	Data Source
Aero	D1	m2	Erosion area	(2)	The area of slight, light, moderate, intense, very intense and severe soil erosion in cultivated land, forestland and grassland under two land use patterns (BAU and ERP) in 2030 and 2060.
T	D1	y	Time period	(2)	The value was set to 1.
IB	D1	gC m−2 y−1	Carbon input to the soil	(3), (4)	The value was set to 357 [1].
K0	D1	y−1	Turnover rate of soil carbon with respect to decomposition in the absence of erosion	(3), (4)	The value was set to 0.027 [1].
KE	D1	y−1	Erosion rate of soil carbon obtained by calculating the ratio of the soil erosion rate to the depth of carbon in the top soil layer, which dominates erosion	(4)	The value was set to Vero/0.2 [1].
Vero	D1, D2	m y−1	Erosion rate	(4), (5), (7)	Vero of soil erosion in red soil was 0.16, 1.21, 2.64, 4.91, 8.50 and 13.85, respectively, according to the ratio between the median soil erosion modulus and soil bulk density (1.35) under different water erosion intensities measured by Liang [41]. Vero of cultivated land, forestland and grassland was calculated by weighing the area of soil erosion of different degrees in the three types of land.
Cmin/CSOC-surf	D1	/	Ratio of organic carbon content in the top layer to that in the bottom layer	(5)	The value was set to 0.01 [1].
CSOC-surf	D2	kg m−3	Soil organic carbon content in the top 4.5 cm soil layer	(7)	The SOC of the surface soil layer of cultivated land, forestland and grassland were 9.88, 10.67 and 13.03, respectively [42].
SDR	D2		Ratio of the total sediment exported out to the total eroded soil within the grid	(7)	The SDR of slight, light, moderate, intense, very intense and severe soil erosion in red soil were 0.2, 0.4, 0.6, 0.8 and 0.95, respectively. The SDR of cultivated land, forestland and grassland was calculated by weighing the areas of soil erosion of different degrees in the three land types.
K0-subsoil	D2	y−1	K0 value in the subsoil layer	(7)	Taking 0.2 m as the subsoil depth, the value was 0.016.

shows the values of each parameter in the erosive soil deposition area of the red soil hilly region.

The carbon density of cropland, forestland and grassland was obtained from previous studies and the SOC stock was previously verified [42,43]. In this study, the carbon sinks of the ERPs were estimated using the carbon density data. See Supplementary Materials for the data details.

3. Results

3.1. Scenario Analysis of LUC Patterns

The area and area proportion of different land use are shown in

Table 2. Comparison of land use type areas under three LUC patterns.

Pattern	Time	Cropland		Forestland		Grassland		Water Area		Construction Land		Unused Land
		km2	%	km2	%	km2	%	km2	%	km2	%	km2	%
BAU	2030	115,597	14.53	482,065	60.58	65,078	8.18	39,508	4.96	91,797	11.54	1691	0.21
	2060	103,799	13.04	438,892	55.16	68,085	8.56	50,899	6.40	132,146	16.61	1915	0.24
ER	2030	130,384	16.39	523,203	65.75	63,470	7.98	19,676	2.47	57,474	7.22	1537	0.19
	2060	86,075	10.82	528,018	66.36	76,284	9.59	19,676	2.47	83,957	10.55	1734	0.22
No LUC	2015	185,538	23.32	514,656	64.68	50,948	6.40	19,700	2.48	24,973	3.14	911	0.11

for the three LUC patterns. In the RSHR, the proportion of forestland was greater than 50% and the least land use types were the water area and unused land. From 2030 to 2060, the area of cropland in the ER pattern decreased from 16.39% to 10.82%. The area proportion of forestland increased from 65.75% to 66.36% and the area proportion of grassland increased from 7.98% to 9.59%. During the 30 years, the area of cropland decreased by 44,309 km² and the areas of forestland and grassland increased by 4815 km² and 12,814 km², respectively. The area proportion of forestland, cropland and grassland in the BAU pattern in 2030 was 60.58%, 14.53% and 8.18%, respectively. The area of forestland and cropland decreased by 1.48% and 5.43%, respectively, whereas the area of grassland increased by 0.38% from 2030 to 2060.

3.2. Scenario Analysis of Soil Erosion

The performances of the scenario simulations under the RCP4.5 and RCP8.5 scenarios were R² = 0.745 and R² = 0.798, respectively (Figure 3). According to the relationship between the simulated precipitation and the observed precipitation under the RCP4.5 and RCP8.5 scenarios combined with the simulated precipitation in 2030 and 2060 by the SDSM, the precipitation in 2030 and 2060 was obtained.

In 2030, the soil erosion of the cropland, forestland and grassland in the No LUC pattern was 240 t km⁻², 193 t km⁻² and 370 t km⁻² under the RCP4.5 climate scenario and 261 t km⁻², 229 t km⁻² and 434 t km⁻² under the RCP8.5 climate scenario, respectively. In 2060, the soil erosion of the cropland, forestland and grassland in the No LUC pattern under the RCP4.5 climate scenario was 274 t km⁻², 221 t km⁻² and 421 t km⁻², respectively; that of the RCP8.5 climate scenario was 211 t km⁻², 184 t km⁻² and 347 t km⁻² (Figure 4a,c). From 2030 to 2060, in the No LUC pattern, the soil erosion of the cropland, forestland and grassland in the RCP4.5 scenario showed an increasing trend. In the RCP8.5 scenario, the intensity of erosion showed a decreasing trend. The effect of LUCs on soil erosion in the BAU/ER patterns could be obtained by differentiating the soil erosion in the BAU/ER patterns from the No LUC pattern (Figure 4b,d). The LUCs of the BAU pattern reduced the soil erosion of the cropland and grassland, but increased the soil erosion of the forestland. Under the RCP4.5 and RCP8.5 scenarios, the soil loss of the cropland and grassland decreased whereas the forestland increased. The LUCs of the ER pattern reduced the soil erosion of the forestland and grassland, but increased the soil erosion of the cropland. The soil loss of the forestland decreased under the RCP4.5 scenario. Under the RCP4.5 and RCP8.5 scenarios, the soil loss of the grassland decreased whereas the soil loss of the cropland increased.

3.3. Climate Change Scenario Analysis of Carbon Sinks Induced by Soil Erosion

In 2030, the erosion−induced carbon sinks of the cropland, forestland and grassland in the No LUC pattern under the RCP8.5 scenario were 48 kg km⁻², 106 kg km⁻² and 213 kg km⁻² higher than that under the RCP4.5 scenario, respectively. In 2060, the erosion−induced carbon sinks of the cropland, forestland and grassland under the RCP4.5 scenario were higher at 242 kg km⁻², 155 kg km⁻² and 498 kg km⁻², respectively. From 2030 to 2060, the erosion−induced carbon sinks of the cropland, forestland and grassland showed an increasing trend under the RCP4.5 and RCP8.5 scenarios, but the rate of increase under RCP8.5 was smaller than that under the RCP4.5 scenario (Figure 5a,c).

From 2030 to 2060, the land types in the BAU pattern, ER pattern and No LUC pattern all showed a trend of increase in erosion−induced carbon sinks under the RCP4.5 and RCP8.5 climate scenarios. In the No LUC pattern, the erosion−induced carbon sinks of the cropland, forestland and grassland in the RCP4.5 scenario increased by 402 kg km⁻², 121 kg km⁻² and 1435 kg km⁻², respectively. Under RCP8.5, the erosion−induced carbon sinks increased by 111 kg km⁻², 121 kg km⁻² and 724 kg km⁻², respectively (Figure 5b,d). In the BAU pattern, the erosion−induced carbon sinks of the cropland and grassland decreased under the RCP4.5 and RCP8.5 scenarios whereas the erosion−induced carbon sinks of the forestland increased. Under the RCP4.5 scenario, the cropland and grassland in the BAU pattern reduced the erosion−induced carbon sinks by 78 kg km⁻² and 138 kg km⁻² and 77 kg km⁻² and 157 kg km⁻² under the RCP8.5 scenario, respectively. Under the RCP4.5 and RCP8.5 scenarios, the increased erosion−induced soil erosion carbon sink of the forestland in the BAU pattern was 39 kg km⁻² and 58 kg km⁻², respectively. In the ER pattern, the erosion−induced carbon sinks of the forestland and grassland decreased under the RCP4.5 and RCP8.5 scenarios whereas the erosion−induced carbon sinks of the cropland increased. The forestland in the ER pattern reduced the erosion−induced carbon sinks by 20 kg km⁻² and 78 kg km⁻² under the RCP4.5 and RCP8.5 scenarios, respectively, and the grassland reduced the erosion−induced carbon sinks by 78 kg km⁻² under the RCP8.5 scenario. In the ER pattern, the erosion−induced carbon sinks increased by 39 kg km⁻² under both the RCP4.5 and RCP8.5 scenarios.

4. Discussion

4.1. Impact of ERPs on Carbon Sinks Caused by Soil Erosion

Since the 1980s, 14 ERPs have been implemented in the RSHR to prevent and control soil erosion by land cover. These were selected for the present study; detailed information can be found in Supplementary Materials Data. The characteristics of the implementation of the ERPs included a large−scale financial investment by the government and the large−scale mobilization of the public to participate in labor as well as the maintenance and supervision of ecological protection as social norms. It mainly relies on long−term government funding and planning for programs to address soil degradation and poverty [44]. The ERPs are based on local realities such as soil loss caused by slope croplands and Karst landforms. In addition, the selection of measures in the ERPs is based on rigorous scientific experiments. For example, on the slope croplands, a slope larger than 25° should be reclaimed into terraced fields for cultivation [45] The implementation of ERPs emphasizes the diversity of the vegetation cover, which can increase the SOC storage by increasing the carbon input (especially the underground carbon input), increasing the diversity of the soil microbial community and inhibiting carbon loss during decomposition [5]. Moreover, the implementation cycle of ERPs adopts a pilot and a phased implementation to reduce the abortion of the program. With the coordinated support of local, provincial and central governments, the ongoing program monitoring and evaluation provides the basis for the next stage of program planning and implementation.

In terms of implementation measures, ERPs are mostly based on specific land types (mainly croplands and forestlands) to solve the ecological problems encountered in agricultural production. They have the characteristics of a single restoration target, an emphasis on resource utilization, greater human intervention in the projects and a high restoration cost. In future planning, ERPs can emphasize a systematic ecological restoration based on nature to minimize human interference in the natural ecosystem [46]. In the future, key factors affecting soil erosion and erosion−induced carbon sinks need to be identified and interventions that enhance positive effects should be identified. It is necessary to emphasize the protection and cultivation of forestland in situ and support the development of a fast−growing forest industry. Improvements should be made to cropland infrastructures and a focus placed on improving the quality of arable land (e.g., reducing soil loss, repairing acidified soil and increasing the SOC input).

4.2. Future of the Implementation of ERPs to Meet Global Climate Change

In the No LUC pattern, the soil erosion of different land use types under the RCP4.5 scenario showed an increasing trend whereas that under the RCP8.5 scenario showed a decreasing trend. The erosion intensity of the forestland increased in the BAU and ER patterns, indicating that the change of forestland cover caused by a mild or high ecological restoration had a negative effect on the erosion control for the forestland. The forestland in the ER pattern had the smallest increase in soil erosion under the RCP8.5 climate scenario, which was 0 (Figure 6). The soil erosion of the cropland in the BAU pattern decreased whereas that of the cropland in the ER pattern increased, indicating that a mild ecological restoration could reduce the soil erosion in the cropland. The cropland in the BAU pattern had the largest reduction in soil erosion under the RCP8.5 climate scenario, reaching 28 t km⁻². The decrease in soil erosion of the grassland in the BAU and ER patterns indicated that the change of grassland cover caused by a mild or high ecological restoration had a positive effect on controlling the soil erosion. The grassland in the ER pattern had the largest reduction in soil erosion under the RCP8.5 climate scenario, reaching 14 t km⁻². It appeared that increasing the vegetation cover did not always help to reduce soil erosion.

In the RCP4.5 and RCP8.5 scenarios, the erosion−induced carbon sinks increased by 210 Tg and 85 Tg between 2030 and 2060, respectively. The annual average carbon sink accounted for 3.84% and 1.41% of the annual average carbon sequestration of terrestrial ecosystems, respectively [43]. In the No LUC pattern, the erosion−induced carbon sinks under the RCP4.5 and RCP8.5 climate scenarios presented an increasing trend and the carbon sink under the RCP4.5 scenario additionally increased. The increased degree of the BAU and ER patterns in the RCP8.5 scenario was greater than that in the RCP4.5 scenario, indicating that LUCs driven by ecological restoration had a positive effect on the increase in erosion−induced carbon sinks. Specifically, the forestland in the BAU pattern increased the erosion−induced carbon sinks by 58 kg in the RCP8.5 scenario; the cropland in the ER pattern increased the carbon sink by 39 kg under both the RCP4.5 and RCP8.5 scenarios. In the BAU and ER patterns, the erosion−induced carbon sinks reduced. In the ER pattern, the grassland under the RCP4.5 and RCP8.5 scenarios had a minimum reduction of 78 kg. According to our analysis, neither the BAU nor the ER pattern achieved the goal of increasing the erosion−induced carbon sinks of the grassland. The SOC of the grassland was higher than that of the cropland and forestland, which should be considered as a priority in optimizing ERPs in the future.

In order to achieve a maximum control of soil loss and increase erosion−induced carbon sinks, it is necessary to select appropriate ecological restoration strategies based on iterative and adaptive management. The first step is to reduce man−made disturbances to the existing forestland landscape. Many previous studies have demonstrated that reducing human activity in cropland and forestland has restored soil productivity and nutrients, in particular enhancing SOC storage [47,48]. In addition to carbon sequestration, plantation and shrub forests can also provide other ecosystem services, which are feasible choices in the RSHR where water resources are less limited [49]. Tree species richness has an impact on ecosystem services and the erosion rate decreases with an increase in tree species richness [50]. The second step is that cropland protection measures are effective in the prevention and control of soil erosion, but ignore the protection of the ecological function of the cropland. Engineering measures such as building terraces and warping dams in croplands can improve soil carbon storage [51]. In view of the characteristics of continuous and heavy rainfall as well as soil cohesiveness in the RSHR, new ecological restoration measures should be developed in the cropland to control soil erosion based on soil carbon sequestration. In the third step, grassland had the highest carbon sequestration potential in the soil erosion deposition sites, but the proportion of the grassland cover area on the surface was the lowest. Similar studies on the vegetation cover of ecological restoration sites also found that the SOC storage of grasslands was higher than that of forestlands and croplands [51]. Compared with forestlands, the shrub vegetation biomass density is relatively small and soil carbon accounts for a larger proportion of vegetation biomass carbon [52]. Biological crusts are generated in the ecological restoration process of grasslands, which can protect loose particles of the soil epidermis and water erosion and improve soil quality and productivity [53]. Grass species can be cultivated to make them more suitable for the climate of the region. Ecological restoration measures that give priority to grassland cover can also be implemented when forestland cover is close to the ecological threshold.

5. Conclusions

In this paper, three LUC patterns were designed in two climate scenarios to analyze the potential impact of ERPs on erosion and erosion−induced carbon sinks in the RSHR from 2030 to 2060. It provides a blueprint for a long−term ecological restoration strategy that integrates future LUCs with climate change. Under the RCP4.5 and RCP8.5 scenarios, the annual mean erosion−induced carbon sinks accounted for 3.84% and 1.41% of the annual mean carbon sink of terrestrial ecosystems during 2030–2060, respectively. The specific strategies were as follows: the forestland in the BAU pattern increased the carbon sink by 58 kg km⁻² in the RCP8.5 scenario. The cropland in the ER pattern increased the carbon sink by 39 kg km⁻². The minimum reduction in grassland in the ER pattern was 78 kg km⁻². The potential of grassland to increase erosion−induced carbon sinks should be emphasized in future ERPs. We have provided new evidence for reducing soil erosion potentially related to the enactment of ERPs in the RSHR and have offered a theoretical and practical basis for maintaining the soil carbon sinks of cropland, forestland and grassland sustainability for a subtropical region.