Using the InVEST-PLUS-GeoDetector Model to Predict and Analyze the Pattern of Ecosystem Carbon Storage in the Dongting Lake Basin, China

Abstract

Guaranteeing the ecological security of the Dongting Lake Basin is of paramount importance for national-scale programs, such as the Yangtze River Economic Belt and aquatic conservation projects. Within this framework, carbon storage and its determining drivers act as essential indicators of regional ecological stability. However, the historical trajectory of carbon pools and their response to future multi-scenario land-use transitions remain insufficiently understood. Therefore, this study aims to quantify the spatiotemporal evolution of carbon storage in the Dongting Lake Basin from 2000 to 2020 and project its future dynamics under diverse development pathways. This study, utilizing land use data from 2000 to 2020 and the carbon density database of the Dongting Lake Basin, assessed land use changes over two decades and determined the spatiotemporal distribution of carbon storage. Additionally, using 17 driving factors and various spatial policies, the study projected the land use and land cover changes (LUCC) for 2030 under four scenarios: natural development, ecological protection, economic development, and planned development. The spatiotemporal distribution of carbon storage and its response mechanisms were analyzed for each scenario. The results showed that carbon storage was directly impacted by LUCC, with an overall “decrease-increase-decrease” trend from 2000 to 2020, resulting in a net increase of 3.685 × 10⁶ t. By 2030, the changes in carbon storage under the natural development, ecological protection scenario, economic development, and planned development scenarios were projected to be −1.008 × 10⁷ t, 1.276 × 10⁷ t, 3.292 × 10⁸ t, and −1.200 × 10⁵ t, respectively. Notably, the ecological protection scenario showed a significant positive growth in carbon storage, primarily driven by an increase in forest and wetland areas. Additionally, the spatial distribution of carbon storage exhibited a pattern of “high in the west and low in the east”. These results imply that to achieve the “Dual Carbon Strategy”, future land use planning in the Dongting Lake Basin should prioritize ecological protection and planned development models, including strict control of construction land expansion, increasing ecological land area, and enhancing carbon storage.

1. Introduction

Driven by global climate mitigation efforts, China formally announced during the September 2020 UN General Assembly its intention to have CO₂ emissions culminate prior to 2030 and realize carbon-neutrality by 2060 [1]. The successful attainment of these strategic milestones is largely predicated on the functionality of terrestrial ecosystems, which act as vital reservoirs for atmospheric CO₂ and exert a profound influence on the stability of the global climate system [2]. Terrestrial ecosystems sequester approximately 28% of annual anthropogenic carbon emissions, with carbon storage serving as a fundamental metric for evaluating this efficiency [3]. While carbon pools respond to various environmental factors, land-use transition remains the primary driver of these shifts, particularly as soil—the largest terrestrial carbon reservoir—accounts for over 80% of total storage [4,5,6]. Theoretically, these dynamics are governed by the “pattern-process-function” mechanism and humanland coordination theory, where land-use transitions signify a fundamental reorganization of regional biogeochemical cycles [7]. Enhancing the carbon storage capacity of terrestrial ecosystems is essential for regulating regional climates, understanding land use dynamics, exploring the factors driving changes in carbon storage, and supporting ecological protection alongside sustainable economic and social development [8,9].

As China’s second most expansive freshwater body, Dongting Lake is nestled within the Dongting Lake Basin, which represents a prominent sub-watershed across the Yangtze River’s middle and lower segments. This region is distinguished as a high-priority ecoregion within the Global 200 preservation list [10]. Featuring three Ramsar-designated wetlands, the Basin’s contribution to international biological diversity is paramount. Moreover, the area serves as a pivotal migratory corridor and wintering refuge for waterbirds traversing the East Asian-Australasian Flyway [11]. Beyond its ecological status, it acts as a premier agricultural hub for national grain supplies. The Basin facilitates indispensable natural functions, including flood mitigation, water filtration, climate moderation, and habitat maintenance [12]. However, the region currently faces a critical scientific and managerial dilemma: the inherent conflict between agricultural intensification for food security and the preservation of high-value carbon sinks in wetlands and forests [13,14]. Historically, the high correlation between intensive land utilization in urban agglomerations and carbon emission intensity has triggered imbalances in regional carbon storage patterns [15,16]. This makes the Dongting Lake Basin a significant study for exploring the trade-offs between anthropogenic development and carbon neutrality goals. Consequently, protecting the Dongting Lake Basin is essential for the integrated protection of the Yangtze River [17,18,19].

Different modeling approaches have been used to determine the dynamic evolution of carbon sequestration and its underlying determinants. For instance, the synergistic application of the Patch-generating Land Use Simulation (PLUS) [20], the InVEST suite [21], and the GeoDetector spatial statistical tool [22] provides a robust framework for deciphering carbon fluctuations across multiple scales. Within this integrated architecture, the PLUS algorithm facilitates high-fidelity simulations of landscape transitions, establishing a solid empirical base for future projections. Meanwhile, the InVEST platform elucidates the quantitative nexus between land-use configurations and carbon density, enabling the precise mapping of sequestered carbon. Finally, the GeoDetector tool identifies the correlations between biophysical or socioeconomic variables and carbon storage, highlighting how singular and interactive forces drive regional carbon dynamics.

In recent years, researchers have increasingly integrated the InVEST framework into diverse land-use modeling platforms to examine how LUCC alters carbon sequestration. For instance, the PLUS-InVEST synergy was utilized to anticipate carbon stock variations in the Beibu Gulf metropolitan area through 2060 [23], as well as to scrutinize landscape transformations across Liaoning Province [24]. Similarly, the fusion of CA-Markov and InVEST was deployed to evaluate the efficacy of ecological restoration on carbon stocks within the arid Heihe River Basin [25]. Beyond biomass estimation, the GeoDetector approach has been instrumental in dissecting how demographic density and vegetative mantle govern carbon distribution in the Yellow River’s middle reaches [26]. Furthermore, spatial analysis in Hefei revealed that terrain and productivity significantly shape habitat suitability [27], while social factors like population concentration principally drive land-use transitions in Zhuhai [28]. Collectively, these studies demonstrate that urbanization is a primary driver of carbon loss, yet they often overlook the specific ‘humanland’ contradictions in complex river-lake basins. Addressing this gap is essential for balancing agricultural productivity with carbon sequestration in the Dongting Lake Basin.

This study assesses the spatiotemporal transitions of land-use patterns and carbon sequestration capacity across the Dongting Lake Basin from 2000 to 2020. This specific interval was designated to capture the profound landscape restructuring resulting from rapid urbanization and landmark ecological initiatives, such as the “Returning Farmland to Lake [29]” and “Grain for Green” programs [30]. By scrutinizing the impacts of landscape shifts—specifically the expansion of built-up areas and the preservation of forest-wetland complexes—this study coupled the PLUS and InVEST frameworks to simulate future carbon trajectories for 2030 under four divergent scenarios. Furthermore, the Geodetector framework was implemented to untangle the latent determinants and interactive mechanisms governing the observed spatial heterogeneity. The findings of this study will help optimize land use patterns, ecological protection, and balanced economic development, providing quantitative support and strategic guidance for achieving carbon neutrality regionally.

Based on this, this study investigated the spatiotemporal evolution of carbon storage and its underlying driving mechanisms in the Dongting Lake Basin to support regional carbon management and ecological security. The questions addressed in this study were as follows:

Q1. What are the spatiotemporal evolution characteristics of land use and carbon storage within the Dongting Lake Basin during the 2000–2020 period?

Q2. What are the primary driving factors, and how do their interactions influence the spatial heterogeneity of carbon storage in the Dongting Lake Basin?

Q3. How will regional carbon sequestration fluctuate under different land-use simulation trajectories by 2030, and what strategies can effectively balance economic expansion with carbon storage goals?

2. Materials and Methods

2.1. Study Area

The Dongting Lake Basin (24°38′–30°24′ N, 107°16′–114°15′ E) occupies a central-southern position within China. It is situated between the Yangtze River’s middle reaches to the north and the Nanling Mountains to the south (Figure 1). Spanning approximately 260,000 km², or 14% of the entire Yangtze River drainage area—this region constitutes a vital hydrological network anchored by Dongting Lake [31]. As a strategic hydrological crossroads, the Basin facilitates the convergence of the Xiang, Zi, Yuan, and Li rivers. This intricate drainage network functions as an indispensable nexus for regional water circulation while simultaneously safeguarding the broader ecological integrity of the Yangtze River system [32]. Marked by substantial topographic heterogeneity, the Basin’s geomorphology transitions from rugged mountain ranges in the west to a mosaic of rolling hills and tectonic depressions across the central-southern zones, ultimately flattening into expansive low-lying plains in the northern sector. This physical landscape is governed by a characteristic subtropical monsoonal regime, defined by well-demarcated seasonal shifts and copious meteoric precipitation totaling 1200–1700 mm per annum. Furthermore, the region maintains a stable thermal profile, with mean yearly temperatures typically fluctuating within the 16–17 °C interval. These conditions support rich biodiversity, with forests and wetlands providing valuable habitats for numerous flora and fauna, contributing to their significant ecological and economic value [33]. Because of its ecological and economic significance, the Dongting Lake Basin is designated as a priority area for environmental conservation within the Chinese national framework.

2.2. Data Sources

The 2000–2020 study period was chosen to capture the most representative period of landscape reorganization and socioeconomic transition in the Dongting Lake Basin. With the objective of characterizing these trends, land-use datasets for 2000, 2005, 2010, 2015, and 2020—were systematically partitioned into seven categories: cropland, forestland, grassland, water, construction land, unused land, and wetland. The acquired data were processed through geometric correction, mosaicking, cropping, reclassification, and remote sensing imagery projection transformation. All the datasets were standardized to the WGS1984 UTM Zone 50N coordinate system on the ArcGIS 10.4 platform, ensuring consistent dimensions and accuracy. To optimize spatial resolution and data processing efficiency, the data were resampled to 90 m × 90 m resolution after multiple trials (

Table 1. Datasets and their specifications for the Dongting Lake Basin.

Category	Variable	Data Source
Land Use	Multi-temporal Land-Use (2000–2020)	Resource and Environmental Science Data Platform (https://www.resdc.cn/, accessed on 9 June 2025)
Restricted Areas	Middle Yangtze River City Cluster	Middle Yangtze River Geoscience Data Center (https://cjgeodata.cug.edu.cn/, accessed on 9 June 2025)
	Chengdu-Chongqing Urban Agglomeration	National Catalogue Service For Geographic Information (https://www.webmap.cn/, accessed on 9 June 2025)
Socio-economic	Nighttime Light	Harvard Dataverse (https://dataverse.harvard.edu, accessed on 9 June 2025)
	Population Density	Landscan (https://landscan.ornl.gov/, accessed on 9 June 2025)
	GDP	Resource and Environmental Science Date Platform (https://www.resdc.cn/, accessed on 9 June 2025)
Distance Factor	Distance to Railroad	Open Street Map (https://www.openstreetmap.org/, accessed on 9 June 2025)
	Distance to Highway
	Distance to Primary Road
	Distance to Secondary Road
	Distance to the City Center
Natural Factor	Soil Type	Institute of Tibetan Plateau Research (http://data.tpdc.ac.cn, accessed on 9 June 2025)
	Distance to River	OpenStreetMap (https://www.openstreetmap.org/, accessed on 9 June 2025)
	NDVI	Institute of Tibetan Plateau Research (http://data.tpdc.ac.cn, accessed on 9 June 2025)
	Average Annual Temperature
	Average Annual Precipitation
	DEM	Geospatial Date Cloud (http://www.gscloud.cn, accessed on 9 June 2025)
	Slope
	Aspect

).

2.3. Methods

2.3.1. Land-Use Prediction Scenarios Setting

This research evaluated projected land-use dynamics within the Dongting Lake Basin spanning 2020 to 2030 across four distinct simulation pathways. For each scenario, future land-use requirements for 2030 were first estimated by integrating linear regression with Markov chain techniques, leveraging historical transition patterns derived from the 2000–2010 and 2010–2020 intervals. The divergence among these pathways stems from the deliberate adjustment of conversion probability matrices or the implementation of specific spatial exclusionary constraints:

(1)

Natural Evolution Scenario (NES): This scenario serves as the baseline, assuming the continuation of historical land-use trends without setting any functional restriction areas or planned development zones. The land demand and transition potential were calculated directly from historical data without further adjustment.

(2)

Ecological Protection Scenario (EPS): Under the EPS scenario, the government prioritizes ecological restoration by enforcing stricter policies on expanding construction land. Such measures include converting cropland to forests and lakes and limiting the conversion of forestlands, grasslands, and wetlands to cropland or construction land. The following transition probabilities were adjusted following the restrictions: cropland and unused land were 60% more likely to convert to forest, wetland, and water bodies and 80% less likely to become construction land. Furthermore, the grasslands were 60% more likely to transition to forestland in order to meet the minimum demand for urbanization.

(3)

Economic Development Scenario (EDS): This scenario prioritized meeting economic and social development needs, increasing demand for cropland and construction land. Based on the NDS, the transition probabilities of all land use types (except water bodies) to cropland and construction land increased by 50%. The planned development zones were established within the Middle Yangtze River City Cluster and the Chengdu-Chongqing Urban Agglomeration. At the same time, in 2020, water bodies were designated as functionally restricted areas to meet water demands for production and domestic use.

(4)

Planned Development Scenario (PDS): Within the actual planning framework, the three scenarios should coexist, balancing ecological, production, and living space development needs. Specifically, key ecological areas, rivers, lakes, and wetlands were designated functional restriction areas, prohibiting the transition of forests, wetlands, and water bodies to other land use types. Based on the NDS, transition probabilities of cropland and unused land to forests, wetlands, and water bodies increased by 30%, while conversion to construction land decreased by 30%. Additionally, the transition probability of grassland to forest increased by 20%.

2.3.2. PLUS

The Patch-generating Land Use Simulation (PLUS) model [20], originally designed by the China University of Geosciences, utilizes a random forest model to analyze the expansion of various land use types and their driving forces. This process determines the development probabilities for each land category and quantifies the contribution of driving factors to their expansion over the specified period. The model formula was expressed as:

$$P_{i,k\left(x\right)}^d={\displaystyle \frac{{\sum }_{n=1}^{ m}I\left(h_n\left(x\right)=d\right)}{M}}$$

(1)

where d indicates the transition status, with a value of 1 signifying a conversion to land-use type $\mathrm{k}$ and 0 representing all other cases. Here, $\mathrm{x}$ constitutes a vector of various independent driving factors. The collective prediction of the ensemble is characterized by the indicator function $\mathrm{I}( )$ for the decision tree set, where $h_n\left(x\right)$ denotes the specific category output by the $n$ decision tree corresponding to vector $\mathrm{x}$. Lastly, $\mathrm{M}$ accounts for the total number of decision trees incorporated within the model.

2.3.3. InVEST

Carbon storage is the total amount of carbon stored in vegetation, soil, and dead organic matter within a given area. It is a key indicator reflecting the carbon storage capacity of an ecosystem. The carbon storage was calculated using the carbon module in the InVEST framework with the following formula:

$$\begin{gathered} C_i=C_{i,above}+C_{i,below}+C_{i,soil}+C_{i,dead} \\ {C}_{total}=\sum _{i=1}^n{C}_i\times A_i \end{gathered}$$

(2)

where the total carbon storage ($C_{\mathit{total}}$) is an aggregate metric of carbon sequestered across $n = 7$ land-use categories. For the $i$-th type, the area is denoted by $A_i$ (hm²), while the carbon density $C_i$ (t/hm²) integrates four distinct reservoirs: above-ground ($C_{i,\mathit{above}}$), below-ground ($C_{i,\mathit{below}}$), soil ($C_{i,\mathit{soil}}$), and dead organic matter ($C_{i,\mathit{dead}}$). The parameterization of these densities was based on established literature for the region [34,35,36,37], summarized in

Table 2. Initial carbon density across different land use types in the Dongting Lake Basin.

Land Use Type	${\mathbf{C}}_{\mathbf{a}\mathbf{b}\mathbf{o}\mathbf{v}\mathbf{e}}$	Cbelow	${\mathbf{C}}_{\mathbf{s}\mathbf{o}\mathbf{i}\mathbf{l}}$	Cdead
Cropland	1.80	0.35	80.96	0
Forestland	41.46	8.68	130	1.16
Grassland	0.55	0.14	63.42	0.06
Water bodies	0	0	0	0
Construction land	0	0	44.15	0
Unused land	0	0	30.47	0
Wetland	3.57	13.89	131.61	0.95

.

2.3.4. GeoDetector

GeoDetector is a tool that identifies spatial variations among geographic objects and analyzes the factors driving these patterns [38]. In this study, the model facilitates two distinct analyses: single-factor detection, which gauges the isolated contribution of each driver to carbon storage patterns, and interaction detection, which evaluates how the synergy between multiple factors shapes spatial variability.

(1)

Single-factor Detection

The Factor Detector determines the degree to which an environmental factor explains the spatial stratification of carbon storage. The magnitude of this influence is measured by the $q$-statistic, formulated as:

$$\begin{gathered} q=1-{\displaystyle \frac{{\sum }_{h=1}^L{n}_h{{\sigma }_h}^2}{n{\sigma }^2}}=1-{\displaystyle \frac{SSW}{SST}} \\ SSW={\sum }_{h=1}^L{n}_h{{\sigma }_h}^2 \\ SST={\sum }_{h=1}^{L}n{\sigma }^2 \end{gathered}$$

(3)

where $\mathrm{h} = 1, \dots , L$ denotes the specific strata or sub-regions of the independent variable $X$; $N_{\mathrm{h}}$ and $N$ represent the number of spatial sampling units within stratum $\mathrm{h}$ and across the entire study area, respectively. The terms ${\mathsf{\sigma }}_{\mathrm{h}}^2$ and ${\mathsf{\sigma }}^2$ signify the variance of carbon storage within stratum $\mathrm{h}$ and the global variance of the entire region. Correspondingly, $\mathit{SSW}$ and $\mathit{SST}$ correspond to the sum of within-strata variances and the total sum of squares. The $q$ value is restricted to $[0, 1]$, where a higher value signifies a more robust explanatory capacity of the factor regarding the spatial distribution of carbon storage.

(2)

Interaction Detector

The Interaction Detector evaluates whether the combined influence of two distinct factors ($X1$ and $X2$) enhances or weakens their individual explanatory power, or if they operate independently. The assessment begins by calculating the individual $q$-values for each factor—$q(X1)$ and $q(X2)$—and subsequently determining the interaction $q$-value, $q(X1\cap X2)$. By comparing the interaction result with the standalone values, the relationship can be classified into several types, including the following:

(1)

Nonlinear enhancement: $q\left(X1\cap X2) > q\right(X1) + q(X2)$;

(2)

Dual-factor enhancement: $q\left(X1\cap X2) > \max (q\right(X1),q(X2\left)\right)$;

(3)

Weakening: $q\left(X1\cap X2) < \min (q\right(X1),q(X2\left)\right)$.

3. Results

3.1. Land Use Change from 2000 to 2020

From 2000 to 2020, the land use structure of the Dongting Lake Basin remained largely stable (Figure 2). Forestland consistently covered the largest portion, covering 60.58% to 61.35% of the total area, followed by cropland, covering 27.97% to 29.17%. The primary land use changes included an initial expansion of forestland by 2013.85 km² from 2000 to 2010, followed by a decrease of 768.39 km² from 2010 to 2020, reflecting a relatively slow rate of change. However, cropland and grassland areas steadily declined, with reductions of 3125.69 km² and 1935.89 km², respectively.

The land use transfer matrix from 2000 to 2020 indicated notable shifts in land use types within the study area, particularly in cropland, forestland, and grassland (

Table 3. Land use transfer matrix for Dongting Lake Basin, 2000–2020, in km².

2000	2020
	Cropland	Forestland	Grassland	Water Bodies	Construction Land	Unused Land	Wetland	Total Area	Transfer Out
Cropland	68,372.06 km2	4744.36 km2	283.39 km2	752.91 km2	2062.86 km2	2.41 km2	108.59 km2	76,326.58 km2	7954.524 km2
Forestland	3595.75 km2	152,570.62 km2	538.88 km2	351.18 km2	1440.46 km2	12.82 km2	20.78 km2	158,530.50 km2	5959.88 km2
Grassland	434.45 km2	2133.82 km2	12,515.28 km2	45.68 km2	128.46 km2	0.82 km2	38.26 km2	15,296.76 km2	2781.48 km2
Water bodies	333.33 km2	181.73 km2	8.98 km2	5157.85 km2	62.43 km2	0.17 km2	635.99 km2	6380.48 km2	1222.63 km2
Construction land	256.54 km2	126.00 km2	10.58 km2	35.27 km2	2843.42 km2	0.20 km2	5.14 km2	3277.15 km2	433.73 km2
Unused land	1.88 km2	5.48 km2	1.16 km2	0.45 km2	2.43 km2	13.19 km2	0.19 km2	24.79 km2	11.59 km2
Wetland	206.87 km2	13.96 km2	2.59 km2	238.46 km2	10.39 km2	1.30 km2	1370.63 km2	1844.21 km2	473.58 km2
Total Area	73,200.89 km2	159,775.97 km2	13,360.87 km2	6581.80 km2	6550.45 km2	30.92 km2	2179.58 km2	261,680.47 km2
Transfer in	4828.83 km2	7205.35 km2	845.59 km2	1423.96 km2	3707.02 km2	17.72 km2	808.96 km2

). Cropland exhibited the highest transfer-out rate at 42.23%, predominantly transitioning into forestland and construction land, followed by forestland. The transfer-out rates for other land use types, in descending order, were grassland (14.77%), water bodies (6.49%), wetlands (2.51%), and construction land (2.30%). Unused land had the lowest transfer-out rate at just 0.06%, mainly converting to forestland and cropland.

3.2. Temporal Change in Carbon Storage

Between 2000 and 2020, the carbon storage in the Dongting Lake Basin exhibited a fluctuating trend of “decrease-increase-decrease,” driven by land use changes (

Table 4. Land use carbon stocks (t) in Dongting Lake Basin from 2000 to 2020.

Land Use Type	2000	2005	2010	2015	2020
Crop Land	6.344 × 108 t	6.314 × 108 t	6.169 × 108 t	6.129 × 108 t	6.084 × 108 t
Forestland	28.742 × 108 t	28.785 × 108 t	29.107 × 108 t	29.061 × 108 t	28.967 × 108 t
Grassland	0.981 × 108 t	0.953 × 108 t	0.870 × 108 t	0.865 × 108 t	0.857 × 108 t
Water bodies	0 t	0 t	0 t	0 t	0 t
Construction Land	0.145 × 108 t	0.163 × 108 t	0.210 × 108 t	0.244 × 108 t	0.289 × 108 t
Unused Land	0.755 × 105 t	0.681 × 105 t	1.092 × 105 t	1.085 × 105 t	0.942 × 105 t
Wetland	0.277 × 108 t	0.268 × 108 t	0.284 × 108 t	0.325 × 108 t	0.327 × 108 t
Total	36.49 × 108 t	36.48 × 108 t	36.64 × 108 t	36.63 × 108 t	36.52 × 108 t

). The total carbon storage for the years 2000, 2005, 2010, 2015, and 2020 were 36.49 × 10⁸ t, 36.48 × 10⁸ t, 36.64 × 10⁸ t, 36.63 × 10⁸ t, and 36.52 × 10⁸ t, respectively. Results showed a decline from 2000 to 2005, a subsequent increase from 2005 to 2010, and a gradual decline from 2010 to 2020. Despite these fluctuations, the overall variation in carbon storage over two decades was relatively small, with a net increase of 3.685 × 10⁶ t.

Carbon storage varied significantly across land types over the study period. Forestland was the largest contributor, with carbon storage increasing from 28.742 × 10⁸ t in 2000 to 29.107 × 10⁸ t in 2010, before declining slightly to 28.967 × 10⁸ t in 2020. In contrast, cropland exhibited a continuous decline in carbon storage, dropping from 6.344 × 10⁸ t in 2000 to 6.084 × 10⁸ t in 2020. Grassland also followed a decreasing trend, while wetland and construction land saw a slight increase. Particularly, carbon storage in construction land nearly doubled, rising from 0.145 × 10⁸ t in 2000 to 0.289 × 10⁸ t in 2020, driven largely by urban expansion.

From 2000 to 2020, high carbon storage areas in the Dongting Lake Basin were primarily located in the western part of the Basin, while low carbon storage areas were concentrated in the northeast. This spatial pattern—characterized by “high in the west, low in the east”—aligns with the distribution of land use types. Forestland in the west stored significantly more carbon than cropland in the east (Figure 3). The average annual carbon reserves during this period across the land use types ranked as follows: forestland > cropland > grassland > wetland > construction land > unused land > water areas.

3.3. Variations in Carbon Storage and Driving Forces

The factor detection results (

Table 5. The q-statistics of driving factors for the spatial variation of carbon storage in the Dongting Lake Basin from 2000 to 2020.

Driver Factor	q
	2000	2005	2010	2015	2020	Average Value
Railroad (×1)	0.6496	0.6490	0.6486	0.6484	0.6465	0.6484
Aspect (×2)	0.6502	0.6494	0.6488	0.6484	0.6468	0.6487
Rainfall (×3)	0.6590	0.6581	0.6589	0.6584	0.6567	0.6583
City center (×4)	0.6562	0.6559	0.6568	0.6568	0.6551	0.6562
River (×5)	0.6621	0.6621	0.6631	0.6626	0.6615	0.6623
DEM (×6)	0.6785	0.6791	0.6813	0.6810	0.6797	0.6799
Slope (×7)	0.7012	0.7017	0.7053	0.7051	0.7038	0.7034
Evapotranspiration (×8)	0.6526	0.6523	0.6530	0.6530	0.6515	0.6525
Soil (×9)	0.6728	0.6727	0.6729	0.6718	0.6695	0.6720
Expressway (×10)	0.6522	0.6519	0.6519	0.6519	0.6502	0.6516
Subsidiary road (×11)	0.6528	0.6523	0.6524	0.6526	0.6512	0.6523
GDP (×12)	0.2213	0.2212	0.2213	0.2212	0.2204	0.2211
Temperature (×13)	0.6682	0.6685	0.6706	0.6702	0.6691	0.6693
Population (×14)	0.6500	0.6495	0.6490	0.6486	0.6465	0.6487
NL (×15)	0.6567	0.6571	0.6588	0.6595	0.6592	0.6583
NDVI (×16)	0.7320	0.7331	0.7375	0.7368	0.7390	0.7357
Trunk road (×17)	0.6519	0.6517	0.6516	0.6516	0.6500	0.6514

) revealed that all selected driving factors significantly influenced the spatial distribution of carbon storage in the Dongting Lake Basin (p < 0.01). Among them, NDVI (×16) emerged as the most dominant factor, with its q-statistic consistently exceeding 0.73 across all periods (average q = 0.7357). This was followed by topographic factors such as Slope (×7) and DEM (×6), both of which showed high explanatory power with average q-values of 0.7034 and 0.6799, respectively.

Additionally, the two-factor interactions of drivers influencing carbon storage (Figure 4) further highlighted that, from 2000 to 2020, the interaction between any two driving factors exhibited greater explanatory power for the spatial differentiation of carbon storage than any single factor alone. However, this combined effect was still less than the sum of their explanatory power. This pattern indicates a dual-factor enhancement without evidence of nonlinear weakening, single-linear weakening, independent action, or nonlinear enhancement. The strongest interactive effects from 2000 to 2020 were observed in combinations involving NDVI with soil and slope, all of which had explanatory power exceeding 70%. This further reinforces NDVI’s role as the predominant factor driving the spatial distribution of carbon storage.

3.4. Forecast of Land Use Change for 2030

To ensure the reliability of future simulations, the PLUS model was first validated by utilizing historical land-use data from 2000 and 2010 to reconstruct the 2020 landscape status (

Table 6. Multi-scenario land-use demand prediction for 2030 in km².

Land Use Type	NES		EPS		EDS		PDS
	2030	2020~2030	2030	2020~2030	2030	2020~2030	2030	2020~2030
Cropland	72,323.99 km2	−876.90 km2	71,229.73 km2	−1971.16 km2	74,014.91 km2	814.02 km2	71,658.84 km2	−1542.05 km2
Forestland	159,322.63 km2	−453.33 km2	161,148.03 km2	1372.06 km2	157,133.97 km2	−2642.00 km2	160,040.49 km2	264.52 km2
Grassland	13,174.32 km2	−186.55 km2	12,907.66 km2	−453.21 km2	13,017.69 km2	−343.18 km2	13,084.53 km2	−276.34 km2
Water bodies	6515.95 km2	−65.85 km2	6691.67 km2	109.87 km2	6581.80 km2	0 km2	6603.86 km2	22.06 km2
Construction land	8171.32 km2	1620.87 km2	7214.73 km2	664.29 km2	8649.14 km2	2098.69 km2	7812.73 km2	1262.29 km2
Unused Land	27.18 km2	−3.74 km2	25.39 km2	−5.53 km2	26.24 km2	−4.68 km2	26.20 km2	−4.72 km2
Wetland	2145.08 km2	−34.50 km2	2463.27 km2	283.69 km2	2256.73 km2	77.15 km2	2453.83 km2	274.25 km2

). This validation process yielded a Kappa coefficient of 0.89 and an overall accuracy of 0.96, confirming that the model’s predictive performance aligns with the rigorous requirements of this research. Under the Natural Evolution Scenario (NES), built-up surfaces were projected to expand by 1.621 × 10³ km² relative to 2020, while cropland and forestland coverage receded by 8.769 × 10² km² and 4.533 × 10² km², respectively. In contrast, the Ecological Protection Scenario (EPS) prioritized environmental stewardship, resulting in a substantial augmentation of forestland (1.372 × 10³ km²) and wetland habitats (2.837 × 10² km²), largely at the expense of agricultural land (1.971 × 10³ km²). The Economic Development Scenario (EDS) reflected a paradigm focused on societal infrastructure, characterized by notable increments in both cropland (8.140 × 10² km²) and construction land (2.099 × 10³ km²), albeit coupled with a significant depletion of forest resources (2.642 × 10³ km²). Finally, under the Planned Development Scenario (PDS), cropland area declined by 1.542 × 10³ km², while both forestland and wetland zones exhibited positive growth of 2.645 × 10² km² and 2.742 × 10² km², respectively.

3.5. Changes in Carbon Reserve Under Multiple Scenarios for 2030

The total carbon storage in the Dongting Lake Basin was predicted to increase only by 1.276 × 10⁷ t under the EPS but was predicted to decrease under the other three scenarios (

Table 7. Land use structure and carbon stock projections in Dongting Lake Basin for 2030 in tons.

Land Use Type	NES	EPS	EDS	PDS
Cropland	6.011 × 108 t	5.920 × 108 t	6.151 × 108 t	5.956 × 108 t
Forestland	28.885 × 108 t	29.216 × 108 t	28.488 × 108 t	29.015 × 108 t
Grassland	0.845 × 108 t	0.828 × 108 t	0.835 × 108 t	0.839 × 108 t
Water bodies	0	0	0	0
Construction land	0.361 × 108 t	0.319 × 108 t	0.382 × 108 t	0.345 × 108 t
Unused land	0.828 × 105 t	0.773 × 105 t	0.799 × 105 t	0.798 × 105 t
Wetland	0.322 × 108 t	0.370 × 108 t	0.339 × 108 t	0.368 × 108 t
Total	36.428 × 108 t	36.65 × 108 t	36.20 × 108 t	36.52 × 108 t

). Specifically, EDS resulted in a significant reduction (i.e., 3.292 × 10⁸ t). Across the land use types, grassland had the highest carbon storage, accounting for 78.71% to 79.71% of the total storage, followed by cropland at 16.15% to 16.99%. However, the spatial distribution of carbon storage remained relatively comparable from 2000 to 2000 across all scenarios, with high storage concentrated in forested areas while low storage was found in water bodies and unutilized land (Figure 5).

4. Discussion

4.1. The Responses of Carbon Storage to Land Use Changes from 2000 to 2020

Carbon storage serves as a vital metric for ecosystem service intensity, encapsulating the functional integrity of regional environments [39]. The present investigation analyzed carbon storage trajectories within the Dongting Lake Basin from 2000 to 2020, revealing a tri-phasic “contraction–expansion–contraction” pattern over the two-decade span. Despite this cyclical variability, the region achieved a net carbon stock increment of 3.685 × 10⁶ t, suggesting a reinforced ecosystem service capacity in recent years. However, these trends exhibited significant spatial and temporal heterogeneity, largely governed by transitions in land-use categories. Between 2000 and 2005, carbon storage diminished, primarily driven by the shrinkage of cropland ($7.633 \times {10}^4$ km² to $7.597 \times {10}^4$ km²), grassland ($1.530 \times {10}^4$ km² to $1.485 \times {10}^4$ km²), and wetlands ($1.844 \times {10}^3$ km² to $1.784 \times {10}^3$ km²). These losses were largely precipitated by the proliferation of built-up surfaces, which expanded from $3.277 \times {10}^3$ km² to $3.687 \times {10}^3$ km², highlighting the adverse anthropogenic footprint on carbon sequestration. Subsequently, from 2005 to 2010, a recovery phase occurred as forestland rose from $1.588 \times {10}^5$ km² to $1.605 \times {10}^5$ km² and wetlands rebounded to $1.892 \times {10}^3$ km².

This augmentation of high-carbon-density landscapes offset previous depletions, underscoring the efficacy of afforestation and hydrological restoration. This gain was reversed from 2010 to 2015 as construction land surged to $5.528 \times {10}^3$ km², while cropland and forestland receded to $7.374 \times {10}^4$ km² and $1.603 \times {10}^5$ km², respectively, illustrating the structural trade-offs necessitated by rapid urbanization. The 2015–2020 interval saw a continued decline, with built-up areas reaching $6.550 \times {10}^3$ km². Although wetland conservation efforts increased their extent to $2.180 \times {10}^3$ km², the persistent encroachment of impervious surfaces constrained the overall carbon storage potential. Ultimately, these dynamics reflect the tension between urban expansion and ecological stability, where accelerated modernization competes for space with agricultural and natural lands. Nevertheless, an evolving paradigm of environmental stewardship has fostered the expansion of carbon-dense ecological zones, particularly forestlands, indicating an upward shift in regional environmental quality. The revitalization of wetlands and forests demonstrates the constructive impact of recent policy reorientations toward ecological preservation [40]. Furthermore, the moderated decline in agricultural land suggests an emerging equilibrium between food security mandates and environmental protection strategies [41].

4.2. Complex Interactions Between Environmental Factors and Carbon Storage

The results from the single-factor detection indicated that NDVI, which had an average q-value of 0.7357, was the main driver of spatial variation in carbon storage. These findings contrast previous studies that primarily emphasized anthropogenic factors, such as population density and GDP, as driving factors for carbon storage [42]. This discrepancy may be attributed to the unique nature of the Dongting Lake Basin, where the extensive natural vegetation coverage plays a crucial role in shaping the spatial distribution of carbon storage [43]. As a key indicator of vegetation coverage, NDVI can effectively capture the spatial variability of carbon storage in this region [38].

The interactive detection results further showed that between 2000 and 2020, the combined effect of any two driving factors on the spatial distribution of carbon storage exhibited a greater explanatory power than any single factor, following a “dual-factor enhancement” pattern. Notably, the interaction between NDVI and environmental factors, such as soil and slope, explained more than 70% of the spatial variability in carbon storage, reinforcing the role of NDVI as the primary driver of spatial differentiation of carbon storage in the Basin. The strong interaction between NDVI and soil is likely linked to the influence of regional soil types on vegetation cover [44]. Variations in soil nutrient content and water retention capacity directly affect vegetation growth, impacting carbon accumulation and storage [45]. Similarly, a slope can moderate vegetation growth, amplifying its interaction with NDVI [46], highlighting the role of slope in shaping carbon storage patterns.

In summary, NDVI was a key determinant of carbon storage in the Dongting Lake Basin. In addition, it interacted with soil and slope to influence the spatial distribution of carbon storage. These interactions are not merely additive but enhance the spatial variability of carbon storage [47]. Therefore, effective low-carbon development and ecological planning in the Basin should prioritize the combined effects of vegetation cover, soil conditions, and topography to maximize carbon storage.

4.3. Multi-Scenario Forecast of Carbon Storage in the Dongting Lake Basin for 2030

Using the InVEST modeling approach, this study evaluated changes in carbon storage in the Dongting Lake Basin for 2030 under four land-use scenarios: NES, EPS, EDS, and PDS.

(1)

NES: In 2030, the total carbon storage was projected to be 36.428 × 10⁸ t, marking a decline compared to 2020. This decline may primarily be attributed to construction land expansion, encroaching cropland, forestland, and other land-use types with high carbon storage potential. Specifically, carbon storage in cropland decreased from 6.084 × 10⁸ t to 6.011 × 10⁸ t, while forestland saw a reduction from 28.967 × 10⁸ t to 28.885 × 10⁸ t. Grassland and wetland carbon storage also experienced a slight decline. These trends highlight a consistent decline in carbon storage under the NDS, emphasizing the need for ecological protection measures to prevent further losses.

(2)

EPS: By 2030, the total carbon storage was projected to reach 36.65 × 10⁸ t, showing an upward trend. Under this scenario, forestland saw a substantial increase, with carbon storage reaching 29.216 × 10⁸ t—an increase of 2.488 × 10⁷ t compared to 2020. Wetland carbon storage also increased significantly, from 3.270 × 10⁷ t to 3.695 × 10⁷ t. These outcomes underscore that the EPS effectively curtailed the proliferation of built-up surfaces, thereby safeguarding high-carbon-density zones from anthropogenic encroachment and validating the efficacy of habitat conservation initiatives in bolstering regional carbon reservoirs [25,48].

(3)

EDS: In 2030, total carbon storage was projected at 36.20 × 10⁸ t, with a significant decrease of 3.293 × 10⁷ t. This decline is primarily due to the rapid expansion of construction land, which encroached cropland and forestland. Consequently, cropland carbon storage dropped to 6.151 × 10⁸ t, and forestland carbon storage reduced to 28.488 × 10⁸ t. This scenario indicates that prioritizing economic development reduces carbon storage, consistent with Ye et al. [49], who reported a negative correlation between construction land expansion and regional carbon sequestration.

(4)

PDS: The total anticipated carbon storage for 2030 is estimated at 2030 was 36.52 × 10⁸ t. This study quantified the spatiotemporal transitions of land use and carbon storage in the Dongting Lake Basin from 2000 to 2020, highlighting how landscape changes—such as the proliferation of built-up areas and the preservation of forests and wetlands—have reshaped carbon stock dynamics. To simulate future trajectories, the PLUS model was employed to project land-use patterns for 2030 under four divergent scenarios, while the InVEST framework was integrated to predict subsequent fluctuations in carbon storage. Additionally, the Geodetector framework provided a comprehensive analysis of the driving factors governing the spatiotemporal differentiation of carbon storage across the Basin. The objective was to provide a robust scientific foundation for optimizing land-use configurations to bolster ecological integrity and promote balanced economic development. The results offer quantitative insights and strategic guidance for achieving regional carbon neutrality, with a recorded minimum carbon loss of 1.200 × 10⁶ t under the most favorable management trajectory. Specifically, under the PDS, existing forestlands, wetlands, and water bodies are preserved while accommodating essential urbanization demands. Carbon storage within cropland and forestland was estimated at 5.956 × 10⁸ t and 29.015 × 10⁸ t, respectively. Despite moderate urban growth, the strategic prioritization of forest and wetland habitats resulted in a negligible 1.9% reduction in carbon storage compared to the NDS, demonstrating a clear synergy between economic development and carbon sequestration goals.

Of the four scenarios, the EPS proved to be most effective for increasing carbon sequestration, mainly through the protection and expansion of forest and wetland areas. In contrast, the NDS and EDS showed a significant decline in carbon storage, likely due to the expansion of construction land, which encroaches on high carbon density areas, such as croplands and forests. These results are consistent with similar studies in other regions, reinforcing the detrimental impact of unregulated construction on carbon storage [50]. Meanwhile, the PDS achieved a moderate expansion of construction land while protecting forests, wetlands, and water bodies. This approach strikes a favorable balance between ecological preservation and economic growth, supporting regional development while minimizing the impact on carbon storage and maintaining critical carbon sequestration functions.

Studies have suggested that wetlands and water bodies are significant carbon sinks under frequent water-logging conditions. However, when these ecosystems are disturbed by drainage or destruction, the accumulated organic matter decomposes more rapidly, becoming a carbon source [51]. The higher the wetland aggregation, the stronger their carbon sequestration capacity [52]. Therefore, land use policies should prioritize preserving natural wetlands—such as rivers, lakes, and marshes—and maintaining their natural state to enhance regional carbon sequestration. Nevertheless, future land use planning should prioritize ecological protection and planned development. This includes safeguarding forests, limiting the expansion of construction areas, and preserving wetland aggregation to strengthen regional carbon sequestration and reduce carbon loss.

4.4. Limitations and Future Research

This study used existing research to derive carbon pool parameters through model calibration. Unlike the approaches that rely on national carbon density values [53,54], model calibration offers improved regional applicability, though it remains less precise than direct sampling and survey data. The InVEST framework, while useful, has some limitations, such as assuming fixed carbon storage levels for all land use types. This approach overlooks annual differences in carbon density driven by changes in vegetation age structure and soil microbial activity [55]. Furthermore, the 30 m spatial resolution of the input data may introduce minor uncertainties at the complex land-water interfaces of the basin. Therefore, future studies should integrate field sampling with predictive modeling to improve the accuracy of carbon storage assessments. Building on prior studies [56,57,58], this study selected 17 representative indicators but did not account for the potential effects of climate change on future land use changes. Additionally, while the transition mechanisms identified here are specific to the “river-lake” context of the Dongting Lake Basin, their generalizability to other geographical regions requires further validation. Future studies should refine the selection of driving factors and incorporate climate influence on land use dynamics to improve model applicability. Moving forward, we should adopt more comprehensive evaluation methods to track carbon stock changes in the Dongting Lake basin, providing targeted scientific support for its protection and management.

5. Conclusions

This study quantified the spatiotemporal dynamics of carbon storage across the Dongting Lake watershed from 2000 to 2020. Our findings delineate a temporal trajectory marked by non-linear fluctuations—specifically a “declining–recovering–declining” sequence—largely dictated by variations in the NDVI, which emerged as the paramount determinant governing the spatial heterogeneity of carbon stocks. Predictive modeling for 2030 across divergent pathways suggests that land-use policy interventions significantly modulate the carbon-storage efficacy of diverse terrestrial ecosystems. While the Ecological Protection Scenario (EPS) yielded the maximum potential for carbon accumulation, the Planned Development Scenario (PDS) effectively mitigated carbon depletion while concurrently accommodating regional economic objectives. Consequently, we advocate for a management paradigm that prioritizes ecological integrity alongside strategic development to bolster the Basin’s carbon reservoirs. Critical interventions should involve curbing the unchecked proliferation of built-up surfaces and fortifying the preservation of sylvatic and lacustrine habitats, particularly in zones exhibiting high sensitivity to landscape transitions. These insights provide a robust empirical foundation for tailoring regional carbon management and ecological restoration strategies, supporting China’s overarching “carbon neutrality” objectives.