Introduction

Since the Industrial Revolution, the rapid development of industrialization has triggered a surge in carbon emissions, and the resulting global warming, frequent extreme weather and other problems have attracted increasing attention1,2,3. COP26 indicated that there is an urgent need to introduce relevant policies to reduce emissions, control temperatures and achieve carbon emission targets by the middle of the twenty-first century4. As a major carbon emitter, China clearly proposed “dual carbon” goals in 2020, setting the tone for future high-quality development and ecological civilization construction. As important global carbon sinks, terrestrial ecosystems have irreplaceable ecological value for maintaining the balance of the carbon cycle and mitigating climate change5,6. Land use / land cover (LULC) represents the spatial classification of terrestrial ecosystems, and the structural changes in LULC directly drive the spatial differentiation of regional carbon stocks by changing the surface cover pattern7,8,9,10. Therefore, examining the spatiotemporal patterns of carbon stock dynamics and their drivers is fundamental for sustaining the regional carbon balance, enhancing ecosystem sequestration capacity, developing evidence-based management policies, and advancing “dual carbon” targets.

At present, domestic and international scholars generally use remote sensing inversion, field surveys and model simulation methods to assess carbon reserves11,12,13. Remote sensing inversion has the ability to quickly acquire data over a large scale, but its spatial and temporal resolutions are limited by sensor performance and satellite operating cycles14. Although field surveys can obtain accurate data, they are costly and have limited coverage. Model simulation can quickly generate multiscale carbon stock data based on input data through a modeling framework15. Among the existing models, the InVEST model has attracted much attention from scholars because of its advantages, such as simple operation, easy parameter acquisition, and multiscale visualization expression16. For example, the InVEST model was used to assess the carbon stock in the Beht River Basin in Morocco17, Ningwu County in Shanxi Province18, and Vietnam19. In terms of LULC prediction, the PLUS model has attracted attention because of its excellent predictive ability. Researchers have employed the PLUS-InVEST model to examine ecosystem carbon dynamics across both the Beibu Gulf Economic Zone of Guangxi20 and the Shanxi mining area21. Although the above studies have successfully predicted future regional carbon stock, they are still insufficient for analyzing the driving factors of carbon stock evolution.

In 2022, Hubei Province issued the “Outline of the Planning and Construction of Wuhan Metropolitan Area (Wuhan-Ezhou-Huangshi-Huanggang WEHH)”. Under the new development pattern, the strategic position advantages of the WEHH metropolitan area are highlighted. Based on this outline, this paper utilizes the coupled PLUS-InVEST-GeoDectetor model to study the spatiotemporal variation in carbon stock in the WEHH metropolitan area from 2000 to 2020 and predict the carbon stock distribution pattern in 2030. Finally, the driving factors of the spatial heterogeneity of carbon stock are explored to provide a scientific reference for the ecological development of the metropolitan area.

Materials and methods

Overview of the study area

The WEHH metropolitan area (113°41′ ~ 116°8′E, 29°30′ ~ 31°36′N) is a national-level metropolitan area centered on Wuhan and linked with the three cities of Ezhou, Huangshi and Huanggang(Fig. 1). It is located at the junction of four major plains and the eastern part of Hubei Province, with an area of 32,000 square kilometers. The metropolitan area boasts a mix of low mountains, hills, and plains, with more than one thousand large lakes (≥ 6 hectares), resulting in abundant water resources. Located in the northern subtropical zone, the climate type is a northern subtropical monsoon climate, with an average annual rainfall exceeding 1200 mm. Furthermore, the region boasts a rich array of plant and animal resources, including numerous national nature reserves, such as the Dabie Mountains. As the core area of the Wuhan metropolitan area, it has a permanent population of approximately 23 million, accounting for nearly 40% of the population in the province. Its GDP in 2024 exceeded 3 trillion yuan, accounting for more than 50% of the province’s total economic output, and its comprehensive strength is relatively strong. With the proposal of the two national development axes of the Yangtze River Delta-Chengdu-Chongqing and the Beijing-Tianjin-Hebei-Guangdong-Hong Kong-Macao, the WEHH metropolitan area, as the intersection, is striving to build a modern “nine-province thoroughfare” three-dimensional transportation network with the Wuhan international comprehensive transportation hub as the core, integrating Yangtze River-Han River water transportation, high-speed rail-high-speed land transportation and the Tianhe-Huahu aviation gateway.

Fig. 1
figure 1

Geographical location of the study area. This map was created using ArcGIS Pro 3.4 (ESRI) software. Administrative boundaries at all levels are from the National Geographic Information Public Service Platform (https://www.tianditu.gov.cn/). DEM data is from the European Space Agency’s Copernicus Dataspace Ecosystem (https://dataspace.copernicus.eu/). Basemap URL: https://www.arcgis.com/home/item.html?id=6ea270a2b9b84e55b1fb0c3114e109fe.

Full size image

Data source and preprocessing

Table.S1 and Table 1 show the main data sources of this study, in which LULC data are integrated into six land types, and each distance element is obtained by Euclidean distance processing. In this paper, the data table on the carbon density in Hubei Province compiled by Zhang Bin22 is revised and used as the carbon density data (t/hm2).

Table 1 Carbon density data table for various types of land in Hubei Province.
Full size table

Research methods

InVEST model

The InVEST model’s carbon stock and sequestration module was applied in this study to quantify the temporal variations in the carbon stock. The calculation formula are as follows33,34,35,36,37,38:

$$C_{i} = C_{i - above} + C_{i - below} + C_{i - soil} + C_{i - dead}$$
(1)
$$C_{total} = \mathop \sum \limits_{i = 1}^{n} C_{i} \times S_{i}$$
(2)

where Ci is the total carbon density of the LULC (t/hm2); Ci-above, Ci-below, Ci-soil and Ci-dead are the aboveground biomass carbon density (t/hm2), underground biomass carbon density (t/hm2), soil carbon density (t/hm2) and dead organic matter carbon density (t/hm2) of the LULC type i, respectively; Ctotal is the total carbon stock of terrestrial ecosystems (t); and Si is the area of LULC type i (hm2).

Carbon density correction

Discrepancies may exist between the literature-summarized carbon density data for various LULC types in Hubei Province and the actual values within the study area. The existing carbon density data need to be corrected to improve their accuracy39,40. The corrected carbon density data are shown in Table 2.

Table 2 Corrected carbon density data table for each type of land in the study area.
Full size table

PLUS model future LULC multi-scenario simulation

In this paper, the PLUS model is employed to simulate future LULC under multiple scenarios across the WEHH metropolitan area. The PLUS model includes two main modules—LEAS and CARS—and utilizes linear regression and Markov chain methods to simulate future LULC demands under different scenarios41.

Based on the research of Wang Baosheng42 and combined with expert opinions, this paper sets the domain weights (Table.S2) and selects 13 driving factors for land use simulation in 2020. After testing, the Kappa coefficient is 0.8034, and the overall accuracy is 87.54%. The results show that the PLUS model has a good simulation effect (Fig. 2). On this basis, this study takes the 2020 LULC data as a benchmark and uses the PLUS model to simulate the LULC of the metropolitan area under three scenarios—natural development (Q1), ecological protection (Q2) and cultivated land protection (Q3)—in 2035 (Table.S3). Among them, the Q1 scenario is the continuation of the land use and cover change (LUCC) trend from 2005 to 2020, and the parameter settings are consistent with the simulation parameters of the historical period; the Q2 scenario prioritizes ecological conservation and strictly prohibits the conversion of ecological land to other uses; and the Q3 scenario enforces strict policies to safeguard cultivated land and minimize its conversion. With the exception of construction land, all the other LULC types are eligible for conversion into cultivated land.

Fig. 2
figure 2

Comparison of actual land use and simulation results in 2020. Map generated using ArcGIS Pro 3.4 (ESRI).

Full size image

Geodetector

This study utilized the Geodetector, a statistical approach proposed by Wang Jinfeng et al.43 to investigate spatial heterogeneity in geographical phenomena and its causal factors. Specifically, single-factor and interaction detection functions were used to identify the drivers behind the spatial variation in carbon stock within the study area and to assess how these factors interact44. The factor detection calculation formula is as follows:

$$q = 1 - \frac{1}{{n\sigma^{2} }}\mathop \sum \limits_{h = 1}^{L} N_{h} \sigma_{h}^{2}$$
(3)

where q lies within the interval [0, 1], which is the key indicator for evaluating the control effect of factor X on the spatial distribution of the dependent variable Y. A large value corresponds to a strong influence of X on Y, and vice versa; L represents the categorical grouping of the independent variable X; Nh denotes the sample count per layer, while N represents the total sample size; σh2 represents the variance of the dependent variable within the layer; and σ2 represents the total variance of the dependent variable within the study area.

Results and analysis

LULC scenario simulation

Table 3 displays the 2035 LUCC per scenario relative to the 2020 baseline across the study area. The three scenarios uniformly indicated negative trajectories for cultivated land area compared with that in 2020 but with significant variations in depletion intensity. The cultivated land area decreased by 7.12% in the Q1 scenario; the loss intensified under the Q2 scenario, with a decrease of 9.13%; and the Q3 scenario effectively curbed the loss of cultivated land, with a decrease of only 3.74%, and the loss area decreased by 941.40 km2 compared with that in the Q2 scenario; the areas with reduced cultivated land under the three scenarios were generally located near the original towns (Fig. 3a–d). The woodland and waters areas both tended to increase under the Q1 and Q2 scenarios, with woodland increasing by 2.77% and 5.57% respectively, with the expansion occurring mainly in the northern and northwestern parts of Macheng, the northern part of Hong’an County, the northern part of Huangpi District and the southwestern part of Yangxin County; the waters areas increased by 0.41% and 5.74%, respectively, with the main expansion areas in the southwestern part of Caidian District, Qichun County and the surrounding areas of Yangxin County; and under the Q3 scenario, the areas of the two decreased by 7.51km2 and 165.8km2, respectively. The grassland area tended to decrease in both the Q1 scenario and the Q3 scenario, with decreases of 31.01% and 48.9%, respectively, while in the Q2 scenario, the area increased by 0.1km2. The area of construction land increased under the three development scenarios, with growth rates of 40.64%, 38.11% and 34.94% in the Q1, Q2 and Q3 scenarios, respectively. The cultivated land preservation and ecological governance policies effectively constrained construction land sprawl. The expansion of construction land is manifested mainly in the large-scale occupation of cultivated land and expansion along rivers, roads and railways. The area of unutilized land tends to decrease in all three scenarios, and the difference in the declines are relatively small.

Table 3 Land use change area under different scenarios.
Full size table
Fig. 3
figure 3

Land use simulation results under different scenarios in 2035. a-d, (a) 2020; (b) 2035 (Q1); (c) 2035 (Q2); (d) 2035 (Q3); Q1: natural development scenario; Q2: ecological protection scenario; Q3: cultivated land protection scenario. Map generated using ArcGIS Pro 3.4 (ESRI).

Full size image

Temporal and spatial variation characteristics of carbon stock

Temporal and spatial variation characteristics of carbon stock in metropolitan area from 2000 to 2020

The carbon stock in the metropolitan area fluctuated periodically between 2000 and 2020, culminating in a slight overall reduction. The carbon stock during 2000, 2005, 2010, 2015 and 2020 was 455.24, 458.29, 457.18, 458.18, and 454.81 Tg, respectively. From 2000–2005, the carbon stock increased rapidly, with an increase of 3.05 Tg, an increase of 0.67%. The key reason was that woodland and construction land expansion generated carbon stocks increases that were significantly higher than the losses from cultivated land contraction. From 2005–2015, carbon stock experienced fluctuating adjustments, first decreasing but then increasing; from 2015–2020, the carbon stock decreased by 3.37 Tg. The main reason for this decrease was the shrinkage of woodland, grassland and waters areas during this period. The reduced LULC types were all high-carbon density LULC types, which led to a sharp decline in the total carbon stock. The net decrease in carbon stock was 0.43 Tg during the entire observation period, among which the attenuation from 2015 to 2020 dominated the overall trend of the carbon stock changes, indicating that the carbon sink function in this region is at risk of accelerated degradation in the future. Cultivated land and woodland land contribute significantly to the carbon stock, with average contribution rates reaching 47.11% and 46.34%, respectively. The average contribution rates of waters, construction land, grassland and unutilized land are relatively low, at 4.69%, 1.78%, 0.07% and 0.01%, respectively (Table 4).

Table 4 Area and carbon stock of different land types from 2000 to 2020.
Full size table

Spatial analysis revealed that the distribution of carbon stock in urban terrestrial areas manifested a distinct northeast-high/west-low spatial gradient. Moreover, as time progressed, the low-value areas along the river cut through the high-value areas in the east, eventually forming a symmetrical distribution pattern of high-value areas with the Yangtze River as the axis (Fig. 4a–e). Specifically, high carbon stock regions are located mainly in the Dabie Mountains in northeastern Huanggang, southeastern Ezhou, southern Huangshi and northern Wuhan. The abovementioned areas have dense woodland distributions and strong carbon fixation capacities. Low carbon stock areas are distributed mainly in cities, towns, and near rivers and lakes with intensive anthropogenic activities. The rapid development of Wuhan, Huanggang, Huangshi, Ezhou and their core towns has led to the conversion of substantial woodland and cultivated land into construction areas with low carbon densities. This land conversion drives regional carbon stock depletion, creating zones of low carbon value. Similarly, rivers and lakes, which are inherently low in carbon density, also commonly form such low-value zones.

Fig. 4
figure 4

Carbon stock of metropolitan area distribution from 2000 to 2020. a-e, (a) 2000; (b) 2005; (c) 2010; (d) 2015; (e) 2020. Map generated using ArcGIS Pro 3.4 (ESRI).

Full size image

To intuitively visualize the spatial dynamics of the carbon stock in the metropolitan area for 2000–2020, the changes are divided into three categories: decrease, basically unchanged, and increase. The data in Fig. 5a–e and Table 5 show that the carbon stock in most areas remained basically unchanged in the four periods, with unchanged areas accounting for 95.32%, 94.64%, 93.35% and 94.74% of the total area. However, the carbon stock in some areas changed significantly. From 2000–2005, the decreased carbon stock area covered 602.21 km2, mainly in the vicinity of Wuhan, and the increased carbon stock area covered 902.19 km2, showing a scattered distribution. From 2005–2010, the areas with increased carbon stock were still scattered, but the area of decreased carbon stock increased to 1027.10 km2, which was more obvious than that in the previous period, and there was a trend of spreading to the southeast along the river. From 2010–2015, gains in carbon stock were largely clustered in the north and southwest of the metropolitan area, and losses extended further southward and eastward than they did in the preceding timeframe. The area in which the carbon stock decreased from 2015 to 2020 was 1000.21km2, indicating a spatial distribution pattern of “concentrated in the west and scattered in the east”. The area of increase decreased severely—reaching only 689.68km2—and was located mainly in the southwest corner and scattered in the central and northern parts. Overall, from 2000–2020, the areas in which the carbon stock decreased and increased accounted for 7.38% and 5.99% of the total metropolitan area, respectively, with a difference of 447.64 km2. The spatial distribution in the carbon stock in most areas of the metropolitan area at different times remained basically unchanged. These changes confirm the periodic fluctuation characteristics and slight negative trend of the carbon stock in the metropolitan area.

Fig. 5
figure 5

Spatial distribution of carbon stock changes in different periods from 2000 to 2020. a-e, (a) 2000–2005; (b) 2005–2010; (c) 2010–2015; (d) 2015–2020; (e)2000–2020. Map generated using ArcGIS Pro 3.4 (ESRI).

Full size image
Table 5 Carbon stock change area and proportion.
Full size table

Temporal and spatial variation characteristics of carbon stock under different scenarios in 2035

From a temporal perspective, the carbon stock under all three scenarios is projected to decrease by 2035 relative to that in 2020, although the magnitude of the decrease varies. Minimization of carbon stock loss is markedly greater under the Q2 scenario than under the other two scenarios. Under the Q1, Q2 and Q3 scenarios, the predicted carbon stocks are 450.87 Tg, 453.65 Tg and 449.57 Tg, respectively. These represent reductions of 3.94 Tg, 1.16 Tg, and 5.24 Tg, respectively, from the 2020 level, highlighting that the Q2 scenario has superior effectiveness in mitigating metropolitan carbon stock depletion by 2035. Under the three scenarios, the changes in grassland and unutilized land are limited and have no obvious effect on the regional carbon stock (Fig. 6). The changes in carbon stock in the other LULC types exhibit relatively large fluctuations. Under the Q1 scenario in 2035, the cultivated land carbon stock decreased by 14.63 Tg. Conversely, woodland and construction land increase by 6.01 Tg and 4.64 Tg, respectively. Under the Q2 scenario, the carbon stock in cultivated land experienced a larger reduction (− 18.78 Tg), whereas that in the other three LULC types increases. The carbon stock in woodland significantly increase (+ 12.09 Tg). Under the Q3 scenario, the carbon stock in construction land increases by 3.99 Tg, whereas those in cultivated land and waters decline by 7.69 Tg and 1.32 Tg, respectively. These shifts indicate that cultivated land, woodland, waters, and construction land will be the primary LULC types influencing the future metropolitan carbon stocks.

Fig. 6
figure 6

Changes in carbon stock by region under different scenarios in 2035.

Full size image

Spatial analysis revealed that the carbon stocks in the three scenarios in 2035 are still characterized by distinct northeast-high/west-low spatial gradients (Fig. 7a–c). From 2000–2035, the carbon stock in most regions remained basically unchanged in the Q1, Q2 and Q3 scenarios, with unchanged areas accounting for 96.12%, 95.03% and 97.40% of the total area, respectively (Fig. 8a–c). The carbon stock reduction areas under the three scenarios are basically the same, all located around Wuhan and on both sides of the Yangtze River coast, and the regional distribution of increase is slightly different. Under the Q1 scenario, areas with reduced and increased carbon stock account for 3.04% and 0.84% of the total area, respectively. The areas experiencing increased carbon stock are primarily concentrated in the northern/southwestern parts of Wuhan and the northern/northeastern parts of Huanggang. Under the Q2 scenario, the proportion of areas experiencing increased carbon stock increases to 1.68%, with the areas of increase primarily concentrated in the northern and northeastern regions of the study area. Under the Q3 scenario, these areas significantly decreased in size, with only scattered patches remaining in the northeast.

Fig. 7
figure 7

Carbon stock distribution under different scenarios in 2035. a-c, (a) 2035 (Q1); (b) 2035 (Q2); (c) 2035 (Q3); Q1: natural development scenario; Q2: ecological protection scenario; Q3: cultivated land protection scenario. Map generated using ArcGIS Pro 3.4 (ESRI).

Full size image
Fig. 8
figure 8

Spatial distribution of carbon stock changes under different scenarios in 2035. a-c, (a) 2020–2035 (Q1); (b) 2020 –2035 (Q2); (c) 2020–2035 (Q3); Q1: natural development scenario; Q2: ecological protection scenario; Q3: cultivated land protection scenario. Map generated using ArcGIS Pro 3.4 (ESRI).

Full size image

LULC transfer and carbon stock response in metropolitan area from 2000 to 2020

LUCC alters the structure of the original land cover and thus cause changes in regional carbon stock. As shown in Table 6 and Fig. 9, from 2000–2020, the carbon stock in the metropolitan area decreased by 0.43 Tg. The key reasons for the decrease are the large-scale conversion of carbon-dense woodland to carbon-sparse cultivated land (− 7.25Tg) and construction land (− 0.52Tg) and the transfer of carbon-dense cultivated land to carbon-sparse construction land (− 8.03Tg) and waters (− 1.44Tg). Although conversions from cultivated land, waters, construction land, and unutilized land to carbon-rich LULC generated localized carbon stocks, the limited scale of these transitions proved insufficient to offset losses from the degradation of carbon-dense LULC types. Consequently, net carbon stock depletion occurred throughout.

Table 6 Land use transition and carbon stock response in urban areas from 2000 to 2020.
Full size table
Fig. 9
figure 9

Land use type changes from 2000 to 2020. Map generated using ArcGIS Pro 3.4 (ESRI).

Full size image

the observation period. In summary, construction land expansion constituted the primary driver of regional carbon loss, with large-scale transfers of cultivated land and woodland to urban uses representing the dominant land conversion mechanism driving this decline.

Quantitative attribution analysis of temporal and spatial differences in carbon stock

According to the data in Table 7, the ten selected natural and socioeconomic factors significantly drive the spatial carbon stock distribution patterns in the metropolitan area (p value < 0.1). Systematic analysis revealed marked variations in the explanatory capacity of the factors driving carbon stock heterogeneity, with natural drivers consistently having greater influence than socioeconomic factors. Among them, slope (0.6229) and elevation (0.5478) emerged as the dominant drivers of carbon stock spatial variation, demonstrating a significantly stronger explanatory capacity than the other factors, followed by the NDVI (0.4712), average annual temperature (0.3157), soil type (0.2861) and annual precipitation (0.1433). Socioeconomic factors have minimal influence on the carbon stock spatial patterns, with GDP (0.1209), the night light index (0.0863), and population density (0.0722) demonstrating notably weak explanatory capacity. Among all the drivers, aspect (0.0405) has the least predictive power for carbon distribution heterogeneity. Elevation and slope demonstrate invariant explanatory capacity for carbon stock distribution heterogeneity, with persistently high effect stability, and the explanatory power of the remaining factors varies in different periods, which is reflected mainly in the following: from 2000 to 2015, slope and elevation accounted for the main explanatory power, whereas in 2020, NDVI and slope accounted for the main explanatory power, the explanatory power of average annual temperature and annual precipitation fluctuated significantly, and the explanatory power of slope aspect and soil type had no obvious temporal change trend, whereas socioeconomic drivers exhibited a sustained increase in explanatory influence, and the explanatory power of each factor on the spatial differentiation of carbon stock in the three scenarios simulations in 2035 was basically consistent with that in 2020. In general, elevation determines the potential for carbon stock formation by controlling temperature, precipitation, vegetation and soil development; slope affects carbon stock distribution mainly by controlling soil erosion and the intensity of anthropogenic activities; and GDP directly reflects regional socioeconomic conditions. Consequently, slope, elevation, NDVI, mean annual temperature, precipitation, soil type, and GDP collectively govern the spatiotemporal carbon stock dynamics across the metropolitan area.

Table 7 Geographic detector factor detection results.
Full size table

Bivariate interaction detection among the 10 factors revealed exclusively synergistic or nonlinearly amplified effects, with no instances of independence or antagonism observed.

That is, the spatial differentiation of the carbon stock in the metropolitan area was the result of the joint action of multiple factors (Figs. 10a–e; 11a–c). Among them, slope–NDVI interactions consistently dominated the carbon stock spatial structuring across all the periods, with q values of 0.7338, 0.7324, 0.7463, 0.7340, 0.7454, 0.7317, 0.7322, and 0.7413; in addition, the explanatory power of the interaction between all the factors and the NDVI gradually increased, and the changes in 2020 were particularly obvious. In general, the interaction between slope and other factors maximizes the impact of individual factors on the spatial heterogeneity of carbon stock, whereas elevation and the NDVI have a secondary but substantial common influence.

Fig. 10
figure 10

Results of two-factor detection from 2000 to 2020. a-e, (a) 2000; (b) 2005; (c) 2010; (d) 2015; (e) 2020.

Full size image
Fig. 11
figure 11

Results of dual-factor detection under different scenarios in 2035. a-c, (a) 2035 (Q1); (b) 2035 (Q2); (c) 2035 (Q3); Q1: natural development scenario; Q2: ecological protection scenario; Q3: cultivated land protection scenario.

Full size image

Discussion

Reasonableness of the research results

This study revealed that between 2000 and 2020, the carbon stock in the metropolitan area clearly experienced periodic fluctuations and an overall slight negative trend. The study by Ke Xinli et al.31 identified antagonistic coupling between urban expansion and cropland conservation as a primary catalyst for carbon stock depletion, which is consistent with the conclusion of this study that the dual transformation paths of “cultivated land → construction land” and “woodland → construction land” are the core mechanisms of carbon loss. Zhang Bin et al.45 reported that the carbon stock in the Wuhan metropolitan area is low in the central and western parts and high in the north and south in terms of spatial distribution, which is similar to the conclusion of this study that the carbon stock distribution is high in the northeast and low in the west. Huang Yan et al.46 demonstrated through PLUS-InVEST modeling that the Q2 scenario yielded the peak carbon stock projections—these results are empirically aligned with our findings of minimal carbon depletion under the Q2 scenario. Therefore, the results of this study are reasonable.

Carbon density correction and its impact on the results

To calibrate carbon densities, contemporary mean annual temperature and precipitation data were integrated with land-type-specific coefficients published by Zhang Bin for Hubei Province. After data collection and discussion, we believe that this step was necessary. The data summarized by Zhang Bin and others have problems such as the use of the provincial average to replace the value for a small range, and the data were collected too early to reflect the current situation. The terrain of Hubei Province is relatively complex, with the overall terrain being high in the west and low in the east. The high-altitude areas in the western mountainous areas have different water and heat conditions from those in the eastern plains and hills, and their land carbon density greatly affects the provincial average. In addition, the data were collected too early. In recent years, there has been an increase in extreme weather in Hubei Province, with frequent high temperatures and heavy rainfall, which affect land carbon density.

Since the total carbon density of each LULC type after correction is greater than the original value and the high-low ranking remains unchanged, except for the overall high results in the final carbon stock, the resulting carbon stock change trends are generally consistent.

Uncertainty analysis

Methodological uncertainties inherent in the research framework may introduce measurable deviations in the findings. First, only 13 driving factors were considered in the simulation of future LULC, which reduced the reliability of the model results to a certain extent. In addition, the LULC demand forecast, field weights and cost transfer matrix under different scenarios were based mainly on the results of previous studies and were subject to high subjective influences. Second, when the InVEST model was used, the different carbon densities of the different woodland types were ignored, which could easily lead to deviations in the calculation results. Third, when the GeoDectetor model was used for attribution analysis, the selection of driving factors was not comprehensive enough. Therefore, future research could combine previous research with the current reality to more comprehensively, objectively, and scientifically predict model parameters such as LULC demand, incorporate more driving factors for LULC simulation and attribution analysis, and conduct field measurements of carbon density in different woodland types to improve data accuracy.

Conclusion

Compared with the areas in 2020, cultivated land and unutilized land declined across all the scenarios, while construction land continued to expand. Woodland and waters increased in the Q1 and Q2 scenarios but decreased in the Q3 scenario, and grassland increased only slightly in the Q2 scenario. Cultivated land and ecological protection policies significantly inhibit the expansion of construction land. From 2000–2020, the carbon stock in the metropolitan area showed periodic fluctuations, and overall slightly negative changes (a net decrease of 0.43Tg). The attenuation from 2015 to 2020 dominated the overall trend of the carbon stock changes, indicating that the carbon sink function in the region is at risk of accelerated degradation in the future. Further analysis revealed that the contribution rates of cultivated land and woodland to the carbon stock are the most significant. From 2000–2035, urban terrestrial carbon stocks manifested a distinct northeast-high/west-low spatial gradient. Moreover, as time progressed, the low-value areas along the river cut through the high-value areas in the east, eventually forming a symmetrical distribution pattern of high-value areas with the Yangtze River as the axis. The carbon stock in most regions remained basically stable during the study period. Future scenario simulations revealed that compared with the Q1 and Q3 scenarios, the Q2 scenario significantly suppressed carbon loss, and cultivated land, woodland, waters and construction land were the main LULC types that will affect the carbon stock in metropolitan areas in the future. From 2000–2020, construction land expansion constituted the primary driver of regional carbon loss, with large-scale transfers of cultivated land and woodland to urban uses representing the dominant land conversion mechanism driving this decline. The spatial differentiation of carbon stock in metropolitan areas is the result of the synergistic effect and positive promotion of multiple factors. LUCC is the primary cause. The explanatory power of natural factors is greater than that of socioeconomic factors, and the explanatory power of the latter continues to increase. Among them, the interactions between slope, elevation or NDVI and the other factors constitutes the dominant driving combinations.