Multi-Scenario Prediction of Dynamic Responses of the Carbon Sink Potential in Land Use/Land Cover Change in Areas with Steep Slopes

Abstract

Terrestrial ecosystems are vital carbon sinks that can effectively restrain the rise in CO₂ in the atmosphere. How ecosystem carbon storage (CS) in semi-arid watershed areas with slow urbanization is affected by comprehensive factors of the environment and land use, along with its temporal and spatial changes has still not been fully explored. Notably, there is a paucity of research on the temporal and spatial changes and development trends of CS in the rapid deformation belt of slopes from the eastern margin of the Qinghai–Tibet Plateau to the Loess Plateau. Taking Bailong River Basin (BRB) as an example, this study combined GeoSOS-FLUS, the InVEST model, and localized “social–economic–nature” scenario to simulate the long-term dynamic evolution of CS. The aim was to study how topographic factors and land use change, and their interactions impact carbon sinks and gradient effects in steep-slope areas, and then find out the relationship between carbon sinks and topographic factors to explore strategies to improve regional carbon sink capacity. The results showed that the following: (1) CS in BRB increased year by year, with a total increase of 558 tons (3.19%), and showed significant spatial heterogeneity, mainly due to the conversion of woodland and arable land; (2) except for land use type, the relationship between CS and topographic gradient is inverted U-shaped, showing a complex spatial response; and (3) it is estimated that by 2050, under the arable land protection and natural development scenarios, CS will decrease by 0.07% and 0.005%, respectively, encroachment on undeveloped mountain areas, while the ecological protection scenario gives priority to protecting the carbon sinks of woodland and grassland, and CS will increase by 0.37%. This study supports the implementation of targeted ecological protection measures through topographic gradient zoning, provides a reference for policy makers in similar topographic regions to effectively manage the spatial heterogeneity of CS, and helps further strengthen global and regional climate change mitigation efforts.

1. Introduction

Carbon storage (CS) in terrestrial ecosystems is a critical component of the regional carbon cycle and has a significant impact on atmospheric carbon dioxide (CO₂) levels [1,2,3]. The importance of its protection and enhancement is reflected in the Paris Agreement, the REDD+ program, and the Chinese government’s “Double Carbon Target”. However, ecosystem CS is significantly affected by natural environment changes and human activities, especially land use/land cover change (LUCC) [4,5], in which land use (LU) change is particularly critical to the carbon cycle. However, current quantitative studies on the effects of LUCC on CS are inadequate, especially in terms of ecosystem size, pattern and uncertainty [6,7]. Therefore, a comprehensive survey of spatial and temporal scales is needed to accurately capture the dynamics of CS within terrestrial ecosystems.

Combined climate and land impacts may generate positive (net emissions) or negative (net sequestration) biosphere feedbacks in the direction and magnitude which will hinder or facilitate the achievement of regional GHG reduction targets [8]. Therefore, understanding the characteristics of regional CS changes through the lens of LU types is crucial, including the shift in CS engendered by the LU structure, the potential of CS under various scenarios and the effect of land resource management on CS. In the study of CS in terrestrial ecosystems, the InVEST model is an equivalent factor model to evaluate CS based on LU types and changes [9]. Compared with traditional CS estimation methods such as CASA [10,11] and FORCCHN [12], the InVEST model possesses the benefits of less data requirements and higher operational efficiency [13], and its accuracy has been fully verified on a regional scale [14]. However, this model cannot directly predict the response of CS in specific scenarios or complex dynamic landscapes, and has limitations in dynamically evaluating the effects of LU change on CS change. However, integrating an LU dynamic simulation model can serve to productively address this limitation. The GeoSOS-FLUS model is an integrated geospatial simulation and multi-scenario optimization design model developed by Li Xia et al. [15] built on neural networks. This model effectively simulates the complex nonlinear interactions involved in the transformation of various LU types and demonstrates a high degree of simulation accuracy. Therefore, the combination of the GeoSOS-FLUS model and InVEST model not only overcomes the limitations of a single model, but also effectively predicts the future change trend of terrestrial CS, providing strong decision support for regional sustainable development.

Climate change generally affects semi-arid regions, and their carbon sequestration capacity and ecosystem stability are relatively weak. Large-scale drought conditions severely reduce the total carbon content of the ecosystem [16]. With the rise in drought stress, the photosynthetic rate and productivity of forest and grassland gradually decline, and the plant population density and coverage also decrease, leading to the decline in the CS of vegetation and soil [8]. Therefore, the dynamic characteristics of LU types and CS in semi-arid watershed areas deserve more attention. The semi-arid transitional zone between the eastern margin of the Qinghai–Tibet Plateau (QTP) and the Loess Plateau (LP) is particularly noteworthy. This region’s zone of sharp deformation exhibits unique characteristics along environmental gradients on spatial and temporal scales [17]. The complex terrain can produce significant gradient effects in terms of temperature and precipitation. Affected by significant climate differences, different topography, and geomorphic diversity, vegetation cover presents an obvious vertical distribution, which intensifies the spatial heterogeneity of CS. However, research on how variations in topographic gradients impact CS remains relatively limited [18]. Untangling the effects of topographic factors and environmental interactions remains challenging. In addition, a series of challenges caused by natural disasters, for instance, soil erosion and debris flow, restrict the ecological and economic development of this transitional deformation zone [19,20]. To deal with the issue of vegetation degradation and improve the above-ground vegetation, soil organic CS and the carbon sequestration rate, the government implemented programmes such as “Grain for Green” and “Return grazing land to grassland” in 2000. Despite the initial success of these policies, slow vegetation recovery and low biodiversity persisted [21,22]. At present, it is not clear whether these problems are positive or negative for vegetation and soil environment, because there are few direct studies on the interaction of topographic conditions at different scales. Overall, the CS of the transition zone shows complex dynamic patterns along the topographic gradient in space and time. Addressing this knowledge gap will help clarify the relationship between carbon sink allocation patterns and topographic environmental heterogeneity.

The Bailong River Basin (BRB) is experiencing a plateauing climate, resulting in pronounced vertical zonation and a variety of climate types [20]. Due to factors such as climate change, population growth, expanding land development, significant regional topographic shifts, and intense tectonic activity, the CS of the ecosystem is facing considerable challenges. Therefore, this study aims to quantify and compare the carbon sink potential of LUCC along the elevation gradient in the steep slope area, and studying the CS change of different topographic gradients is helpful to understand its distribution characteristics and influencing mechanism. This study specifically aimed (1) to evaluate and monitor the combined effects of LUCC on ecosystem CS; (2) clarify the influence and gradient effect of topographic factors and LUCC and their interaction on CS through spatial autocorrelation; (3) explore the future change in CS under different scenarios, and the probability and significance of the transition between various LU types; and (4) focus on sustainable management, make recommendations, and look forward to future research prospects.

2. Materials and Methods

2.1. Research Area

The BRB region is situated in a rapidly deforming zone, marking the transition between the QTP and the LP (Figure 1). This area is characterized by significant topographic variations, active tectonic activity, extensive weak rock formations, and frequent heavy rainfall [20]. According to the geological disaster data from the Gansu Geological Environment Monitoring Institute, significant debris flow activities are currently observed, primarily in the middle and lower reaches of the BRB main stream, along fault zones, and areas with a weak lithology. In recent years, due to factors such as population relocation and urban expansion, there have been notable changes in the landscape’s ecological structure, resulting in significant shifts in landscape ecological risks [10].

2.2. Data Sources and Preconditioning

2.2.1. Data Source and Preprocessing

Various influencing factors were collected, including statistical, climatic, and socio-economic data (Figure 2). These driver data were then standardized into a 100 m resolution grid, and the corresponding distance grid was calculated using Euclidean distance. The sources of the potential drivers are detailed in

Table 1. Spatial basic data source.

Dataset	Data	Year	Attribute/Spatial Resolution	Data Resources
Land use datasets	Land use data	2000, 2010, 2020	TIFF/30 m	Globeland30 (http://globeland30.org/ (accessed on 20 November 2024))
Socio-economic data	Administrative boundaries	2022	SHP	National Catalogue Service for Geographic Information (https://www.webmap.cn/ (accessed on 20 November 2024))
	Road network data	2020
	River network data	2022
	GDP	2020	TIFF/1 km	Resource and Environmental Science Data Platform (https://www.resdc.cn/Default.aspx (accessed on 20 November 2024))
	Population density	2019
Climatic environmental data	DEM	2020	TIFF/30 m
	Slope
	Aspect
	Average annual precipitation	2019	TIFF/1 km
	Average annual temperature	2019

. The LU data are categorized into six types: arable land, woodland, grassland, water areas, construction land, and other land. This study used data from different years, mainly following the principles of data availability and integrity. Climate data and some socio-economic data were based on the latest observational and statistical information, and were interlinked in time span, reflecting the socio-economic pattern and infrastructure distribution at the time and reflecting the dynamic changes in the study area to the greatest extent. Through data standardization, the consistency of the data and reliability of the results were ensured. The technical route is shown in Figure 3.

2.2.2. Carbon Density Data

To ensure the accuracy of the data, the CD data used in previous studies with similar latitudes and the same climatic zone as the study area were selected, as well as the CD data revised with national data. The data were mainly derived from the actual data collected by Liu et al. (2012, 2013) [11,12], the arid region data of China and the 2010 China Terrestrial Ecosystem CD dataset (https://nesdc.org.cn/) [14]. Regional CD is determined through the revision of national CD data, and the revision method refers to the research of Alam, Giardina, and Ryan (2000) [13,15]. In the revision process, the relationship between precipitation, biomass, and soil CD, as well as the relationship between temperature and biomass CD were considered [9,23,24]. The CD data in this study integrated both the literature and the actual conditions of the research area, with the detailed data presented in

Table 2. Carbon density data (kg/m²).

Land Use Type	Cabove	Cbelow	Csoil	Cdead
Arable land	0.456	0.745	1.84	0
Woodland	4.072	1.32	4.69	0.652
Grassland	0.086	0.538	1.79	0.129
Water	0	0	0	0
Construction land	0	0	3.35	0
Other	0.063	0.495	1.22	0

.

2.3. CS Simulation Framework Based on LUCC

2.3.1. GeoSOS-FLUS Model

An artificial neural network is an advanced algorithm designed to simulate the structure of biological neurons, primarily used for estimating nonlinear functions with multiple input variables. By incorporating additional driving factors, the accuracy of the results is enhanced, enabling a more detailed representation of the interactions and competition among different land types [25].

In the GeoSOS-FLUS model, the likelihood of land conversion is influenced not only by the probabilities derived from the neural network’s output based on the driving factors, but also on the influence of neighborhood density (Formula (1)), inertia coefficient (Formula (2)), conversion cost, and competition among lands, indicating the strength of expansion ability. Finally, the total probability of land type conversion was determined (Formula (3)).

$${\mathsf{\Omega }}_{p,k}^t={\displaystyle \frac{{\displaystyle {\sum }_{M\times N}con(c_p^{t-1}=\mathrm{k})}}{N\times N-1}}\times w_k,$$

(1)

where ${\mathsf{\Omega }}_{p,k}^t$ signifies the field density, while ${\displaystyle {\sum }_{M\times N}con(c_p^{t-1}=\mathrm{k})}$ is an overview of the pixels. Additionally, $w_k$ reflects the weight assigned to the influence of different LU areas.

$$A{i}_k^t\left\{\begin{array}{l}A{i}_k^{t-1}(\left|D_k^{t-2}\right|\le \left|D_k^{t-1}\right|)\\ A{i}_k^{t-1}\times {\displaystyle \frac{D_k^{t-2}}{D_k^{t-1}}}(0>D_k^{t-2}>D_k^{t-1})\\ A{i}_k^{t-1}\times {\displaystyle \frac{D_k^{t-1}}{D_k^{t-2}}}(D_k^{t-1}>D_k^{t-2}>0)\end{array}\right.$$

(2)

In this case, $A{i}_k^t$ represents the inertia coefficient for the kth land type at iteration time t. On the other hand, $D_k^{t-1}$ and $D_k^{t-2}$ indicate the differences between the pixel count of the Kth land type and the desired number in the previous and the second-to-last iterations, respectively.

$$T\mathit{Pr}o{b}_{p,k}^t=s{p}_{p,k}\times {\mathsf{\Omega }}_{p,k}^t\times A{i}_k^t\times (1-{\mathit{sc}}_{c-k})$$

(3)

where $T\mathrm{Pr}o{b}_{p,k}^t$ is the total probability of pixel p transforming into land-type k at the iteration number t, $s{p}_{p,k}$ represents the adaptive probability of the neural network, $s{c}_{c-k}$ represents the cost of converting LU-type c to type k, and $1-s{c}_{c-k}$ represents the difficulty of conversion.

2.3.2. InVEST Model

The InVEST model was employed to estimate the CS in the study area. Within the InVEST framework, the CS module includes four key carbon pools: the carbon in the above-ground biomass (all living plant material above the soil), the carbon in subsurface biomass (living plant roots), the soil carbon pool (organic carbon in mineral soil), and the carbon in dead organic matter (such as fallen plant material and dead trees). The total regional CS was calculated by summing the contributions from these four carbon pools across different LU types [26,27].

$${\mathrm{C}}_x={\displaystyle \sum _{j=1}^J{A}_{xj}(C_{aj}+C_{bj}+C_{sj}+C_{dj})}$$

(4)

Formula: ${\mathrm{C}}_x$ is the CS of region X, unit is kg; $A_{xj}$ is the area of LU type in region x; $C_{aj}$, $C_{bj}$, $C_{sj}$, and $C_{dj}$ refer to the above ground CD, underground CD, soil CD, and dead organic matter CD (kg/m²) of LU-type j, respectively.

2.3.3. Scene Setting

Building on the land development and conservation goals outlined in the Master Plan of Territorial Space of Gansu Province (2021–2035) and the Horizontal Ecological Compensation Agreement for the BRB (Gannan–Longnan Section), three distinct scenarios were established. These scenarios were designed to forecast and assess the spatio-temporal changes in CS under various future conditions.

Natural development scenario (ND): This scenario assumes that the conversion of all LU types is driven only by natural development, not taking into account the effect of policy, planning, and human intervention. This setting reflects a “no-policy intervention” baseline scenario for assessing the likely trajectory of LU change under natural development conditions. Arable land protection scenario (ALP): This scenario emphasizes prioritizing the conservation of arable land in the baseline scenario by controlling the expansion rate of construction land and reducing the conversion of farmland into construction land as much as possible, which can ensure the stability of food supply, avoid the ecological degradation caused by excessive urbanization, and promote the sustainable development of the region. Ecological protection scenario (EP): Based on the ecological protection principles outlined in the Territorial Ecological Restoration Plan of Gansu Province (2021–2035), this scenario focuses on strengthening the preservation and recovery of ecological land. It seeks to reduce the likelihood of woodland and grassland being converted into built-up areas, while promoting the conversion of arable land into woodland and grassland. The aim was to decrease the likelihood of woodland and grassland shifting to construction land, while encouraging the transformation of arable land into forests and grasslands. The cost matrix used in these scenarios is provided in

Table 3. Scenario of the cost matrix setup.

	ND						ALP						EP
Type	A	B	C	D	E	F	A	B	C	D	E	F	A	B	C	D	E	F
A	1	1	1	1	1	1	1	0	0	0	0	0	1	1	1	1	1	1
B	1	1	1	1	1	1	1	1	1	0	0	1	0	1	0	0	0	0
C	1	1	1	1	1	1	1	1	1	1	1	1	0	1	1	0	0	0
D	1	1	1	1	1	0	0	0	0	1	0	0	0	0	0	1	0	0
E	1	1	1	1	1	1	1	1	1	0	1	1	0	1	0	0	1	0
F	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1	1

Note: A represents arable land; B stands for woodland; C denotes grassland; D corresponds to water areas; E signifies construction land; F indicates other land types. A value of 1 signifies that the conversion between them is possible, while 0 denotes that conversion is not allowed.

.

2.3.4. Accuracy Verification

The model’s accuracy was assessed by comparing the 2020 modeled LU results with the actual LU data for that year. An overall accuracy of 93.68% was achieved, with a kappa coefficient of 0.834, signifying a high degree of validity (values above 0.80 are typically considered valid). These results demonstrate a strong correspondence between the predicted and actual spatial distribution of LU. The model efficiently captures the spatial–temporal patterns of LU alterations, meeting the study’s objectives.

2.4. Topographic Gradient

2.4.1. Topographic Relief

Topographic relief is defined as the vertical difference between the highest and lowest points within an analysis window. It serves as a key indicator for the quantitative characterization of terrain features and the classification of landforms. The optimal statistical unit [] for this index was determined using the mean change point analysis method, and the value was then derived through the grid neighborhood calculation tool in ArcGIS.

$$RA=C_{\max }-C_{\min }$$

(5)

In this formula, $RA$ represents the topographic relief, $C_{\max }$ denotes the maximum elevation (m) within the analysis window, and $C_{\min }$ refers to the minimum elevation (m) in the same window.

2.4.2. Topographic Niche Index

The topographic niche index is a metric that aligns with the elevation and slope attributes at any given point within the analysis space. It provides a comprehensive reflection of the spatial variation in topographic conditions.

$$TNI=\lg \left[({\displaystyle \frac{E}{E_0}}+1)\times ({\displaystyle \frac{S}{S_0}}+1)\right]$$

(6)

In this context, $TNI$ represents the topographic niche index, $E$ and $E_0$ correspond to the elevation (in meters) and the average elevation (in meters) of any grid cell within the space, respectively, while $S$ and $S_0$ denote the slope value (°) and the average slope value (°) of the grid cell. Generally, the niche index is lower for a raster terrain with both low elevation and slope, and higher for terrain with greater elevation and slope.

2.4.3. Topographic Data Grading Standards

To investigate the spatial distribution of CS across varying topographic gradients, the regional topographic features—elevation, slope, topographic relief, and topographic niche index—were divided into ten classes using the quantile method. Each level was labeled based on its value range, as detailed in

Table 4. Classification of DEM, slope, landform relief gradients, and terrain niche gradients.

Rank	DEM/m	Slope/°	Landform Relief	Terrain Niche
I	566–1415	0–12.24	0–388	0.22–1.00
II	1415–1704	12.24–17.29	388–507	1.00–1.13
III	1704–1914	17.29–21.01	507–590	1.13–1.22
IV	1914–2110	21.01–24.21	590–651	1.22–1.30
V	2110–2316	24.21–26.87	651–710	1.30–1.38
VI	2316–2525	26.87–29.53	710–770	1.38–1.44
VII	2525–2748	29.53–32.19	770–831	1.44–1.52
IX	2748–3018	32.19–35.11	831–905	1.52–1.59
X	3018–3362	35.11–39.10	905–1014	1.59–1.69
Ⅹ	3362–4831	39.10–67.83	1014–1573	1.69–2.30

.

2.5. Exploratory Spatial Analysis

In this section, by combining regional CS and spatial autocorrelation analysis, the complex interaction between CS and regions with dramatic topographic changes is revealed in order to promote regional sustainable development.

2.5.1. Spatial Autocorrelation

The analysis of spatial autocorrelation uncovers the distribution patterns and key hotspots by examining spatial disparities and their statistical significance. The formula for this calculation is presented as follows [28]:

$$I_{xy}={\displaystyle \frac{{\displaystyle {\sum }_{i=1}^n{\displaystyle {\sum }_{j=1}^n}{w}_{ij}}(x_i-\overline{x})(x_j-\overline{x})}{{\displaystyle {\sum }_{i=1}^n}{\displaystyle {\sum }_{j=1}^n{w}_{ij}{\displaystyle {\sum }_{i=1}^n{(x_i-\overline{x})}^2}}}}$$

(7)

$$I_i(d)={\displaystyle \frac{n{x}_i-\overline{x}{\displaystyle {\sum }_{j=1}^n}{w}_{ij}(x_i-\overline{x})}{{\displaystyle {\sum }_{i=1}^n{(x_i-\overline{x})}^2}}}$$

(8)

Moran’s I and LISA statistics, denoted as $I$ and $I_i(d)$, respectively, were used to measure spatial autocorrelation. The term $\overline{x}$ indicates the mean value, while $x_i$ and $x_j$ refer to the CS of neighboring units i and j. Additionally, $w_{ij}$ symbolizes the spatial weight matrix, and n denotes the total number of grid cells.

2.5.2. Bivariate Spatial Autocorrelation Between CS and Topographic Gradient

At the spatial scale, there exist response patterns and dependence characteristics between carbon sequestration and the topographic gradient, influenced by linear correlation and aggregation effects. The global bivariate Moran’s I index reflects interactions between watershed ecosystem services and topographic gradients across prolonged time spans.

$$I_{xy}={\displaystyle \frac{n}{{\displaystyle {\sum }_i{\displaystyle {\sum }_j}{w}_{ij}}}}\times {\displaystyle \frac{{\displaystyle {\sum }_i{\displaystyle {\sum }_j}{w}_{ij}}(x_i-\overline{x})(y_i-\overline{y})}{\sqrt{{\displaystyle {\sum }_i(x_i-\overline{x})}}\sqrt{{{\displaystyle {\sum }_j(y_i-\overline{y})}}^2}}}$$

(9)

$I_{xy}$ is the binary Moran’s I index; Other variables are the same as explained in Section 2.5.1.

3. Research Results

3.1. Dynamic Evolution of Land Use from 2000 to 2020

3.1.1. Spatiotemporal Characteristics of Land Use

Between 2000 and 2020, woodland and arable land were the predominant LU types in the BRB, collectively making up 95.37% of the total area. Woodlands are chiefly found in the northwest, west, and south, whereas cultivated land is predominantly concentrated in the east and northeast. Both land types show a trend of initially expanding and then contracting in area throughout this period. The proportional distribution of different land types is as follows: woodland (75.98%), arable land (19.40%), grassland (3.74%), construction land (0.58%), and water area (0.30%) (Figure 4).

3.1.2. Spatial and Temporal Dynamics of CS

The CS in the study area showed significant spatial heterogeneity, and its distribution features showed north–south polarization and local differences, which was highly consistent with the topographic characteristics and landscape distribution in the study area (Figure 5). To the west, the region is adjacent to the high-altitude QTP, while the LP lies to the east and the Qinling Mountains to the south, all of which contribute to a high-density CS zone with abundant forest and grassland resources. In contrast, the densely populated areas in the east show the opposite trend.

Between 2000 and 2020, there was an overall upward trend in CS, but between 2010 and 2020, CS declined slightly (by 30 tons). The primary factor driving this change is the rapid increase in construction land between 2010 and 2020, coupled with the reduction in both forest and agricultural land areas (Figure 6). Although the government has implemented relevant policies such as ecological restoration and territorial spatial planning, these measures have promoted ecological restoration to a certain extent. However, these measures have been insufficient to effectively offset the loss of CS resulting from the conversion of agricultural land into urban areas and the deterioration of woodlands.

3.2. Impacts of Land Use Type Changes on CS

Between 2000 and 2020, the LU conversion area of BRB reached 1933.22 km² (Table 4), and the LU conversion area was as follows: grassland (945.91 km²) > arable land (419.11 km²) > woodland (545.66 km²). During this process, grassland CS showed a decreasing trend and was mostly converted into woodland, followed by grassland. The cultivated land and woodland are basically in mutual conversion, among which the CS of cultivated land is reduced by 342.35 tons, and the CS of woodland is reduced by 178.55 tons. The area of arable land converted to woodland accounts for 82.31% of the total conversion amount of arable land, and the area of woodland converted to farmland accounts for 72.39% of the total conversion amount of woodland. Additionally, the transformation between cultivated land and forestland exceeds that between other land types, resulting in a more significant impact on ecosystem services. In addition, the conversion area of water bodies and other land types is relatively small, and the impact on CS is weak. In general, shifts in LU types have a considerable impact on ecosystem CS, predominantly reflected in grassland, cultivated land, forestland, and built-up areas, with the alterations exhibiting a complex pattern of various combinations.

3.3. Spatial Dependence Mechanism Underlying CS

3.3.1. Gradient Effect of Terrain Driving CS

BRB is located in the area of rapid deformation; the topography is diverse, showing obvious features of a valley, loess plateau, and alpine landform (as shown in Figure 7). Figure 8 shows the distribution of CS under different elevations (DEM), gradients, topographic relief, and topographic niche index gradients.

The variation in CS with altitudes presents a significant gradient (Figure 8): under different DEM gradients, the cumulative value of the ninth gradient (about 3000 m) is the highest, and that of the third gradient (below 2000 m) is the lowest. The CS decreases first and then increases with altitude, which is related to the frequent human activities, and the high incidence of landslide and debris flow in the low-altitude area. The forest cover area is the main area at 2130–3362 m (grade IV–IX), and the alpine meadow, bare land, and snow cover area are the main areas at grade X, and the CS is low.

There are also significant differences in CS among different slope zones. Areas with gentle slopes are predominantly found in the northeastern, central, and southern regions of the basin. These areas are mainly used for agriculture and construction, and have been significantly impacted by human activities. When the slope increases, the suitable hydrothermal conditions and the decrease in anthropogenic influence are conducive to vegetation growth, the increase in mountain forest and grassland cover, and the corresponding increase in CS. However, when the slope reaches grade VI, vegetation decreases, and soil and water conservation function declines, resulting in a slight decline in CS. The spatial distribution of topographic relief and slope is similar, so the variation characteristics of CS on topographic relief gradient are basically the same as that of the slope gradient.

The variation in the topographic niche index further revealed the distribution characteristics of CS. The regions with a high topographic niche index (grade X) are mainly concentrated in the south of Diepe county, the east of Zhouqu, and the west of Wenxian. The low-value areas (grade I–III) are primarily found in the southern part of Wenxian County, as well as in the northern regions of Dangchang and Wudu District, mostly agricultural and urban areas. The median area (grade IV–IX) is dominated by woodland with less human disturbance, and the fluctuation of CS increased with the topographical niche index.

Overall, with the increase in altitude and slope, forest and grassland gradually become the main types of LU, and CS peaks between 2500 and 3300 m above sea level, while areas with a slope of 27–40° and relief degree of 700–1000 are the most suitable for vegetation growth, which contributes to the accumulation of soil organic carbon.

3.3.2. Spatial Correlation Between CS and the Topographic Gradient

A strong and consistent positive relationship exists between the topographic gradient and carbon sequestration (CS) in the study area, leading to a clear stepped pattern in spatial distribution (Figure 9): the region with a low elevation and gentle slope is gradually transitioning to a region with high altitude and good vegetation cover. Low-CS areas such as urban construction land are often surrounded by high-CS areas. On the whole, the CS in the west and northwest of the study area formed a significant high–high accumulation pattern, indicating that the forest resources in this region are rich and have a positive effect on carbon accumulation. The area with a high altitude and large topography is rich in forest resources, which forms favorable conditions for carbon accumulation. The urban construction land with a relatively low elevation and flat terrain shows a low–low accumulation pattern and faces the risk of gradual shrinking of ecological carbon reserves.

The low- and high-concentration areas decreased gradually (90 km², 19.25 km², 30.25 km² and 652.5 km², respectively), indicating that urban expansion and the intensification of human activities had an impact on the ecosystem. At the same time, the decrease in high–high-aggregation area (34 km², 132.25 km², 114.5 km² and 1502.75 km², respectively) reflects the destruction of forest resources and the decline in CS. The low–low aggregation pattern increased from 2000 to 2010, indicating rapid urban expansion during this period. However, with the successive implementation of ecological policies from 2010 to 2020, the low–low aggregation area was reduced, and the ecological environment was partially alleviated. The low–high-aggregation area decreased slightly, showing a declining trend for ecological quality. In areas with a steep terrain, CS is relatively high, while in agricultural areas with a high population density, flat terrain, and vast land, CS declines significantly. There is a correlation mechanism between this distribution and topographic gradient, showing synergistic effects (39.6% in 2000, 34.7% in 2010, 34.9% in 2020) and tradeoff effects (13.7% in 2000, 11.89% in 2010, 11.9% in 2020). However, these effects gradually waned over time, suggesting that the carbon sink capacity of regional ecosystems is declining in general.

3.4. Spatio-Temporal Characteristics of LUCC and C Storage Under Multi-Scenario Simulations

The natural development scenario mirrors the historical pattern, characterized by a modest increase in both forest area and built-up land, largely driven by the “return farmland to forest” policy. These changes are particularly evident in the peripheral areas of cities like Wudu and Dangchang, where significant alterations in carbon sequestration (CS) have occurred. Notably, grassland degradation is expected to continue, potentially leading to economic losses for livestock and a decline in biodiversity. As a result, the ecological benefits will be further weakened by 2050, and the cumulative CS will be 1.1 tons less than in 2020 (Figure 10). At the same time, with the passage of time, social and economic complexity is increasing, and how to balance the value of ecological investment and economic development has become a major challenge. What is more serious is that the limited space for agriculture is becoming increasingly apparent, and the potential risk of food crisis cannot be ignored.

In the ALP scenario, carbon sequestration (CS) showed fluctuations over time. By 2030, CS had decreased by 14.7 tons compared to 2020, but in 2040 and 2050, it saw slight increases of 1 tons and 0.6 tons, respectively. Notable changes in arable land area are particularly observed in regions like Dangchang and Wudu (Figure 11). As urbanization and industrialization progress, arable land is not consistently developed or occupied. When a reduction in arable land leads to decreased food production, the conversion of other land types, particularly those not dedicated to construction, becomes necessary to compensate for the lost agricultural area. This reflects the tradeoff between short-term economic benefits and long-term ecological sustainability in the ALP scenario. In this case, the priority protection of cultivated land may not be able to fully meet the needs of urban development, further highlighting the contradictions and challenges in land resource allocation.

In contrast, ecological conservation scenarios offer a more promising solution for maintaining CS. Compared with 2020, CS in 2030, 2040, and 2050 will increase by 19.6 tons, 22.2 tones, and 16 tons, respectively (Figure 10 and Figure 11). This suggests that regions with rich natural resources are better able to implement green development paths, thereby reducing potential environmental pressures. However, given that the region often faces natural disasters such as landslides and soil erosion, the management of limited resources and space urgently needs to be strengthened to ensure the effectiveness and sustainability of ecological protection measures. In the EP scenario, prioritizing the conservation of existing carbon sinks and promoting the conversion of arable land to woodland and grasslands not only significantly improves the overall carbon sequestration capacity, but also strengthens the resilience of ecosystems. This further highlights the key significance of achieving long-term sustainable development on the basis of balancing ecological protection with socio-economic development.

The ND, ALP, and EP scenarios show significant differences and complex trade-offs in all aspects. The ND scenario is dominated by the natural development of urbanization and industrialization, emphasizing short-term economic growth. Nevertheless, this involves the sacrifice of a persistent loss of arable and ecological land, leading to ecosystem degradation, a significant decline in carbon sink capacity, and raising long-term sustainability issues. ALP scenarios try to strike a balance between agricultural demand and economic development by strictly protecting arable land and prioritizing food security and agricultural stability. However, due to insufficient attention to ecological LU, ALP scenarios may weaken the overall ecological function and limit the resilience of ecosystems. In contrast, the EP scenario significantly enhances carbon sink capacity and ecosystem stability by prioritizing the protection and restoration of ecological land such as woodland and grasslands, while enhancing biodiversity and ecological service functions to provide stronger protection against natural disasters and climate change. However, this scenario is subject to short-term economic pressures and the impact of reduced arable land on food production. The overall comparison shows that the ND scenario is shortsighted by sacrificing ecology and agriculture for economic development. Although the ALP scenario has excellent performance in agricultural protection, its ecological recovery ability is limited and it is difficult to achieve long-term ecological balance. Through an overall improvement in ecological functions, EP scenarios show outstanding advantages in carbon sink capacity, ecological benefits and long-term sustainability. Although the EP scenario may face short-term pressures, it lays a solid foundation for coordinated regional development and future sustainable development, demonstrating long-term strategic value over other scenarios.

4. Discussion

4.1. Response of CS to Land Use Change

Land use is essential in determining the carbon dynamics within terrestrial ecosystems [29], thereby influencing the CS mechanism and causing fluctuations in [30]. The specific shift involves the rise in high-density LUCC types and the fall in low-density LUCC types, thereby improving regional CS [31,32]. The study reveals that CS in the region has exhibited a fluctuating pattern since 2000 (Figure 5,

Table 5. Shifts in land use-related CS in the BRB.

	Vegetation CS (tons)			Soil CS (tons)			Total CS (tons)
	2000	2010	2020	2000	2010	2020	2000	2010	2020
Arable land	481.95	488.32	481.93	738.37	748.13	738.34	1220.32	1236.45	1220.26
Woodland	8087.01	8490.72	8472.43	8012.02	8411.98	8393.87	16,099.03	16,902.70	16,866.30
Grassland	99.62	48.34	48.32	306.37	148.66	148.61	405.99	196.99	196.93
Water	—	—	—	—	—	—	—	—	—
Construction land	—	—	—	13.42	16.33	40.25	13.42	16.33	40.25
Other	0.18	0.29	0.01	0.40	0.63	0.02	0.58	0.92	0.02
Total	8668.76	9027.66	9002.69	9070.58	9325.73	9321.08	17,739.34	18,353.39	18,323.77

). This variation is strongly linked to the enactment of China’s farmland-to-forest policy in 2002, which aimed to secure ecological land, restore degraded farmland, prevent soil erosion, and substantially expand woodland areas. However, rapid economic growth, coupled with the expansion of urban areas and the decline in forested regions, has led to widespread vegetation destruction and ecosystem degradation [33]. Consequently, the average yearly growth in carbon sequestration has slowed, and certain forested areas have been converted into agricultural and urban land. The main cause of carbon loss in the region originates from the growth of urban areas, and the decline in woodland also contributes significantly to CS loss. This is consistent with the findings of Sadat et al. (2020), who also observed that urbanization and the expansion of urbanized land have a major effect on reducing CS [34]. LU patterns are closely linked to biomass, vegetation coverage, and the CS capacity of various land types, which directly influence regional CS changes [35].

Table 6. Land use changes and associated carbon dynamics in the BRB, 2000–2020.

Land Use Type		Vegetation C (tons)		Soil CS (tons)		Total CS (tons)		Change in Area (km2)	Change in Area (%)
Transfer Out	Transfer to	Increase	Decrease	Increase	Decrease	Increase	Decrease
Arable land	Woodland	157.30	—	—	188.24	—	345.54	449.13	82.31
	Grassland	0.24	—	1.76	—	1.52	—	30.48
	Water	—	2.04	1.33	—	3.37	—	11.08
	Construction land	8.30	—	6.60	—	—	1.70	54.97
	Other	0.00	—	0.00	—	0.00	—	0.01
								545.66
Woodland	Arable land	—	106.25	127.16	—	233.41	—	303.38	72.39
	Grassland	—	31.75	44.23	—	75.98	—	92.76
	Water	—	9.03	9.12	—	18.15	—	16.91
	Construction land	1.19	0.0119	3.23	—	4.42	—	5.99
	Other	0.03	0.0003	0.04	—	0.07	—	0.08
								419.12
Grassland	Arable land		1.82	—	13.30	—	11.48	230.44	24.36
	Woodland	231.87	—	—	322.98	—	554.85	677.33	71.61
	Water	—	2.39	0.78	—	3.16	—	12.43
	Construction land	3.67	—	1.60	—	—	2.07	25.67
	Other	0.00	—	0.00	—	0.00		0.04
								945.91
Water	Arable land	0.77	—	—	0.50	—	1.27	4.18	37.46
	Woodland	2.48	—	—	2.50	—	4.98	4.64	41.58
	Grassland	0.14	—	—	0.05	—	0.19	0.75
	Construction land	0.53	—	0.00	—	—	00.53	1.59
								11.16
Construction land	Arable land	—	1.17	-—	0.93	0.24	—	7.75	82.45
	Woodland	0.30	—	—	0.81	—	1.11	1.501	15.97
	Grassland	0.00	—	0.00	—	0.00	—	0.02
	Water	—	0.04	0.00	—	0.04	—	0.12
								9.40
Other	Arable land	0.01	—	—	0.01	—	0.01	0.09
	Woodland	0.09	—	—	0.11	—	0.0020	0.22
	Grassland	0.02	—	0.00	—	—	0.0002	0.22
	Water	—	00.18	0.0008	—	0.0026	—	1.45	73.23
Total		404.5	154.68	195.85	529.43	340.36	923.73	1.98
Sum		249.8178		−333.5792		−583.3696		1933.22

emphasizes the effect of various LU types on CS, revealing trends in soil CS, vegetation CS, and total CS. The results show that forests, grasslands, and arable land contribute more than 95% to the CS of terrestrial ecosystems. Nevertheless, future development trends may pose significant challenges to ecosystems and food security, especially as policy directions and development pathways may lead to marked changes in LU patterns and spatial distribution [36]. In response to the strategy of “carbon neutrality, carbon peak”, the best model of regional low-carbon development is proposed. According to the results, CS is expected to rise in 2050 under the ecological protection scenario. However, in the ND scenario, the growth of land for economic activities negatively impacts the ecological environment, although this model provides high LU flexibility and allows for the natural process of urbanization and industrialization, it fails to effectively guarantee the systematic protection of ecology and agriculture. This approach will lead to a disorderly reduction in arable and ecological land, and ultimately increase carbon loss. In scenarios focused on arable land protection, while there may be a short-term increase in CS, the long-term effect is uncertain, and CS may eventually decline. A continued emphasis on farmland protection policies may not foster stable CS growth in the future [26]. Forests are under increasing pressure from both natural and human-induced degradation [37]. In contrast, the EP scenario effectively avoids ecosystem degradation, promotes the steady growth of forest CS by rationally adjusting the conversion of agricultural land and other land types to forests, and is relatively cheap in economic cost. Therefore, future planning and management should reduce human intervention, and focus on strengthening the protection of forest land and grassland to improve carbon sequestration capacity. The conclusion that rational socio-economic development and land adjustment can effectively mitigate carbon loss is consistent with existing research [38,39,40,41], and supports ecological conservation measures that can effectively address regional ecosystem imbalances and ensure the long-term stability of CS.

4.2. Effect of the Transition Deformation Zone on CS Change

As a critical regional ecosystem, transitional zones exhibit significant transition characteristics of terrain, climate, soil, and vegetation across different spatial scales and geographical locations [42,43]. These characteristics profoundly impact the CS of ecosystems. The transition region from the eastern margin of the QTP to the LP is a rapidly changing ecological ecotone. Its CS fluctuates sharply with topographic factors, regulating the evolution of vegetation types and physiological structures. A complex mountain topography and diverse ecological functions lead to an uneven distribution of CS and significant spatial heterogeneity [44]. Vegetation distribution and growth are significantly impacted in ecosystems where the alpine climate gradually shifts to semi-arid conditions. The regulation of water and heat allocation further affects soil decomposition activities, thus profoundly impacting the carbon sequestration capacity of vegetation and soil [45]. The results show that CS in the study area presents an inverted U-shaped curve as the topographic gradient increases, resembling the findings of Wang et al. (2022) [46]. This is attributed to the concentration of farmland and building land in flat areas at low altitudes, where human activity is evident and economic development is rapid. The high-altitude area has superior natural resources and favorable hydrothermal conditions. In addition, CS presents a stepped distribution pattern in altitude, peaking at approximately 2500 meters above the sea level (Figure 8). This altitude marks the transition area from the eastern margin of the QTP to the LP. With an increased altitude, topographic rain formed by warm, moist air rising along the hillside effectively alleviates regional moisture limitations. Under sufficient moisture conditions, high vegetation coverage can effectively prevent soil from being exposed to intense sunlight, thereby reducing soil temperature and respiration rates, and increasing soil CS [47]. However, with further increases in altitude, various factors in the alpine environment significantly affect plant growth. Tree and shrub communities are gradually replaced by shrubland and alpine grassland. Vegetation CS decreases rapidly along the altitude gradient, with soil CS becoming more dominant at higher altitudes [48]. The dramatic changes in the transition zone with increasing altitude make the climate increasingly unfavorable for crop growth. The grassland area continues to expand, and the agricultural system gradually shifts from crop-based to livestock-based. Livestock production increases carbon emissions while reducing CS capacity. This corresponds with the discoveries made by other scholars. Sujan Chaudhary et al. believe that trees in low-altitude forests have higher CS, and the highest soil CS occurs at higher altitudes [49]. Similar findings have been made by Gebrewahid et al. [50]. Nevertheless, other academics have found partially contrasting conclusions, pointing to the possibility of other climate forces playing a key role in shaping the observed trade-offs [51], such as in regions like the high altitudes of Ecuador, where there is low seasonality in precipitation, organic matter accumulates in the soil, which increases soil CS [52]. Conversely, prolonged precipitation seasons at high altitudes in Peru may hinder the accumulation of organic matter in the soil [51]. Therefore, with the increase in altitude, the drastic change in climate in the transition zone is also an important factor affecting the regional vegetation CS and soil CS. In addition, plant community structure, biological drivers, and variables affecting soil dynamics (such as soil type, quantity and quality of soil organic matter) may also contribute significantly to CS, which is also a research direction that needs to be considered in the future to further improve the understanding of dynamic changes in CS.

4.3. Strategies for the Sustainable Management of CS

The expansion of human activities has progressively intensified the degradation of LU carbon sequestration (CS) capacity [33,37,53]. Overgrazing, agricultural practices, and construction projects have led to a significant deterioration of surface vegetation, including forests and grasslands, in the BRB region [10,54]. This, coupled with frequent occurrences of soil erosion, landslides, and debris flows, has further undermined the ecosystem’s CS potential. The analysis of LU changes in soil and vegetation carbon sequestration (CS) indicated that arable land had a more pronounced influence on the decline in soil organic CS in the BRB. This LU type accounted for 35.6% of the reduction in soil CS observed in the study area. In the scenario simulation, safeguarding agricultural land boosted carbon sequestration, confirming that artificial management could adjust the change in farmland CP in the long or short terms [36]. Therefore, the government should prioritize the “balance between arable land occupation and compensation” in BRB to guarantee the sustainable development of regional soil organic carbon [55,56]. In addition, woodland contributed the most to the increase in vegetation CS in BRB, and the decrease in woodland directly affected the reduction in overall CS. The decline in the woodland area reduced vegetation carbon by 147 tons, accounting for 95% of the decrease in vegetation CS within the study area. EP was the only scenario that could continuously and steadily increase CS. It demonstrates that ecological conservation measures are essential to improve the CS capacity of terrestrial ecosystems. Thus, moving forward, it is crucial to guarantee that economic growth proceeds within the boundaries of the ecological red line. This involves optimizing the regional LU structure, focusing on expanding the areas of forests and grassland, and actively promoting forestry conservation initiatives, as well as land reclamation and greening efforts [57,58]. Moreover, steering the transformation of land into forests will contribute to maintaining the stability of the regional ecosystem and safeguarding the carbon balance. Ultimately, these actions will foster the sustainable growth of low-carbon infrastructure and achieve the transition to carbon sink.

In addition, according to the results of the study on the correlation effect of topographic gradient and CS, the following relevant ecological protection measures and suggestions were put forward: (1) in the low-value (I–IV gradient) area (Figure 12), namely the densely populated area along and north of BRB, and the agricultural area in northern Wenxian County, we should strictly control human behavior, strengthen the intensive use of land, alleviate the contradiction between people and land, and pay attention to the construction of environmental protection projects such as urban greening and ecological livable. (2) In the topographic gradient of the high-value (IV–IX gradient) section, namely the eastern part of Diepe county, the southern part of the Dangchang, the southeastern part of the Wudu, it is necessary to establish and expand ecosystem-service-protection areas, strengthen forest management and protection, attach importance to forest and grass restoration, and enhance environmental carrying capacity and carbon sink capacity. (3) In the section of the high topographic gradient (X gradient), that is, the bare land, alpine desert and other low vegetation cover areas in counties, the protection of alpine meadows should be strengthened, rational grazing should be carried out, the interference of human activities should be reduced, the vulnerability of ecological environment should be reduced, and the ability to resist risks should be improved.

4.4. Deficiency and Prospects

In this study, model selection, data correction, and local scenario settings were used to ensure the accuracy of the experimental results, but there are still some uncertainties. Although the CD data were revised with the help of meteorological factor correction models to bring them closer to estimates of regional CD, this correction was completely eliminated uncertainties, particularly in terms of inter-annual fluctuations and regional variations [59]. Specifically, the InVEST model may have overlooked several key factors, such as microbial activity and photosynthetic rate, in the process of CS estimation [60]. In addition, the model does not adequately account for the influence of the internal structure of LUCC and dynamic changes in vegetation age on CS potential [61]. Therefore, future studies should further refine the model and incorporate more environ-mental variables, human activities, and biochemical factors to improve the accuracy of CS estimates.

5. Conclusions

This study introduces a coupled model and a multi-driver framework to dynamically simulate the spatio-temporal patterns of CS across different scenarios. It also integrates the spatial response mechanisms underlying heterogeneity, aiming to enhance the effective management of CS in regions experiencing rapid slope deformation.

(1)

Over the past two decades, CS in the BRB has risen by 558 tons, primarily driven by the conversion of arable land to woodland. Spatially, CS is higher in the western and northwestern parts of the area, whereas the eastern and northeastern areas exhibit lower CS levels. This distribution is highly consistent with the topographic characteristics and landscape pattern.

(2)

The variation trend for CS in different scenarios is significantly different. In the ND scenario, the disorderly expansion of economic land at the cost of ecological environment leads to excessive reduction of cultivated land and ecological land, and finally aggravates the loss of CS. At the same time, although the cultivated land protection policy has a certain positive effect, its long-term effect is uncertain, which may eventually lead to the decline in regional CS. On the contrary, strengthening ecological protection can effectively expand the area of forest and grassland, and reduce the encroachment of construction land on ecological land, thus significantly improving the regional carbon sink capacity and bringing significant environmental benefits.

(3)

Along the elevation gradient, CS has an inverted U-shaped relationship with the topographic gradient, showing a synergistic effect, and forming an obvious stepped spatial distribution. CS peaks between 2500 and 3300 m above sea level, with superior natural resources in areas with slopes of 27° to 40° and topographic relief of 700 to 1000, resulting in carbon accumulation. However, these steep slopes still require soil and water conservation, and EP projects to further enhance carbon sequestration capacity. Therefore, it is very important to conduct topographic gradient zoning management for steep-slope areas, and a long-term ecological monitoring system should be established to evaluate the effect of ecological protection measures and timely adjust management strategies.

Through the spatial heterogeneity of CS driven by topographic gradients and the customization of ecological corresponding strategies combined with future scenarios, this study provides important insights for fine ecological management in similar areas with rapid slope deformation. Future work could further consider the interaction of soil type, latitude, and localization drivers, and explore optimal LU management options constrained by CS targets under the dual scenarios of economic growth and ecological conservation.