1. Introduction
A significant part of the global carbon cycle is served by terrestrial ecosystems, which annually absorb an average of 2.5 Pg of carbon, or 25% of the emissions from fossil fuels [
1,
2,
3]. The effects of global warming on the economy, environment, and resources, make it a severe problem for humanity today [
4,
5,
6,
7]. The increasing carbon storage (CS) is now thought to be the most practical, eco-friendly, and practical way, to reduce the greenhouse effect [
7,
8]. Some researchers have pointed out that land use and land cover (LULC) is one of the most important elements affecting ecosystems carbon sequestration [
9,
10,
11], about 1.6 Pg of carbon emissions per year are attributable to LULC variations [
12]. China is the top carbon emitter, and its rapid development over the past few decades has been accompanied by unreasonable territorial spatial planning that has caused CS to decline in the majority of areas. China is currently facing numerous challenges on its path to becoming carbon neutral [
13,
14]. Therefore, for future land resource allocation and ecosystem function optimization, it is essential to comprehend the synergistic relationship of LULC and CS and dynamically modify land use planning [
15].
The combined measurement-statistics approaches, such as the forest census [
16] and the bookkeeping model [
17] employed in China to estimate forest carbon storage, have accurate results, and are extensively utilized to solve the problem of CS estimation. However, the actual measurement process is laborious and time-consuming. As an illustration, Tang et al. set 14,371 sample squares in advance to estimate the total carbon pool (89.27 ± 1.05 Pg) in China [
18], making it challenging to use such methods to quickly obtain the evolution characteristics of CS. A number of process models have been developed based on a significant amount of measured experience and the use of remote sensing and geographic information system, such as the BIOME-BGC model [
19] and the TEM model [
20]. These process models have obvious advantages for CS evolution analysis, but their generality is weak due to the high number of parameters and tailored design for particular vegetation types [
21]. The Integrated Valuation of Ecosystem Services and Trade-offs (InVEST) model was developed to solve the above problems and quickly realize the calculation of different carbon pools. It only requires the carbon density data to estimate and map the CS of terrestrial ecosystems. The InVEST model has been widely used in regional carbon sequestration and emission functional mismatch analysis [
22], historical CS evolution [
23], and future prediction [
24]. Since CS estimation in the InVEST model is dependent on LULC data, acquiring accurate LULC historical and future simulation spatial patterns is the foundation for a reasonable CS analysis.
LULC simulation entails both distribution prediction and demand estimation. For future LULC prediction, the enhanced Cellular Automata (CA) model is usually used. The CA model is a discrete grid dynamics model that includes both space and state and can be used to model the dynamic process of complex systems [
25]. Currently, there are various improved CA models, such as Logistic-CA [
26], ANN-CA [
27], CLUE-S [
28], and FLUS [
29]. However, these models are unable to explore the underlying driving mechanisms of LULC change, which results in low simulation accuracy. The Patch-generating Land Use Simulation (PLUS) model, which integrated a CA model with a patch-generating simulation strategy, contains two parts: the land expansion analysis strategy (LEAS); and an improved CA model [
30]. Based on the spatial data of driving factors and LULC, the LEAS part uses a random forest algorithm to calculate the spatial growth suitability of each land use type. The results of the former are input into the CA model, which uses a descending threshold mechanism to determine the LULC distribution. The prediction of LULC demand is most widely based on the Markov model, which assumes that future trends are consistent with historical changes [
31,
32]. However, because of this, it is difficult for this model to consider the individual development characteristics of the different study area, especially the lack of consideration for future policy implementation [
33]. The Multi-objective Optimization (MOP) model, which enables users to design different objective functions and constraints based on different development objectives and set the quantity demanded for each land use type separately, has been used in urban and ecological barrier area [
34,
35].
In summary, MOP is a “top-down” model for optimizing the quantitative structure of LULC, PLUS is a “bottom-up” spatial simulation function, and InVEST is a spatial-temporal CS evolution analysis approach. In order to conduct a multi-scenario simulation analysis of LULC and CS, here we proposed a MOP-PLUS-InVEST coupling model (MPI). To validate the three stated objectives of the MPI model: (1) more accurate land use target planning; (2) high accuracy patch-level LULC simulation; and (3) modelling and investigation of CS evolution. The Beijing-Tianjin-Hebei city agglomeration (BTH) was chosen. Our main goals are as follows: (1) explore the spatial-temporal characteristics of LULC; (2) conduct dynamic simulations of CS at the patch scale; and (3) evaluate CS at the regional scale. This is to solve the problem of coordinated and sustainable development of regional ecology and economy, and to assist government departments in formulating and adjusting land-use planning.
4. Results
4.1. Spatial-Temporal Characteristics of LULC
LULC in 2000 and 2010 were utilized as input, in order to confirm the PLUS model’s simulation accuracy, the simulation results of LULC in corresponding years under the BAU scenario were obtained. Comparing the simulation results to the real LULC in 2010 and 2020 revealed that the accuracy was 91% and 84%, respectively, and the Kappa coefficients were 0.86 and 0.77, respectively. This demonstrated that the accuracy met the needs of the research.
The random forest algorithm is a component of the land analysis approach in the PLUS model, which used to assess the influence mechanism of driving factors on land use type change [
30,
32]. Based on the distribution of LULC change and driving factors, the contribution value of each driving factors was examined (
Table 2).
Table 2 shows that topographic relief clearly influences all land use forms, excluding bare ground. The plain regions with low height, better topography, and hydrothermal conditions, have a higher concentration of agricultural activity. For the growth of grassland and forest, mountainous with a high altitude and complicated topography are better suited. The determinants of built-up expansion include population increase and the construction of road systems. The main causes of the expansion of bare land are human activities and unfavorable soil conditions.
According to
Figure 5a,b,f,g, the transition between cropland, woodland, and grassland, best reflects the historical evolution. A lot of croplands were being invaded by the rapid rise of the built-up area. As shown in
Table 3, the following were the primary LULC changes during this time: (1) There was a trend toward less cropland and water. 11,829.93 km
2 cropland decreased, of which 66.38 % was converted to built-up, which was then replaced by grassland (21.26%). A total of 207.81 km
2 of water were lost, primarily due to agricultural and built-up areas (80.20%). (2) The amount of forest, built-up, and bare land, showed an upward tendency. Cropland and grassland provided 99.06% of the growth in forest, which increased by 469.02 km
2. The built-up area had increased by 75% to around 26,207.06 km
2 in 2020 compared to 2010, and the growth rate was still increasing. The loss of grassland was primarily responsible for the expansion of bare land. (3) An inverted “U”-shaped trend of growing and then dropping in both grassland and wetland. The grassland area increased by 0.62% and was largely constant. Wetland area declined by 9.06%, primarily due to a shift to cropland and water. (4) There were obvious spatial variations in the changes of the various land types. As seen in
Figure 5f,g, the Taihang Mountains to the north and west of the study area were the primary locations for the spread of grassland and forest. The North China Plain study area’s central and eastern regions had the greatest growth in built-up, with the pattern of growth showing a spread outward from the urban core. Both the water conservation area and the eastern Bohai Sea coastline zone were affected by the growth of water. These results were in line with earlier research on the mechanisms affecting land use expansion, and they once more demonstrate how topography, hydrothermal conditions, and economic status, can all have an impact on LULC change.
The LULC simulation results of the three scenarios in 2030 are shown in
Figure 5c–e. They demonstrated that the differences in simulation outcomes clearly had an impact on goals. Under BAU scenario (
Figure 5c,h), the growth of forest and grassland in the mountainous areas of the western part of the study area was at a standstill. Built-up in the central and southern plains of the study area expanded significantly, with a growth rate of 25.83%. The wetland area showed a decreasing trend (at 26.36%), mainly transformed into forest and water. Social construction under the BAU scenario would continue the trend of 2000–2020, with a policy focus favoring socioeconomic development. The socio-economic benefits of LULC increased by about 22.08% and the ESV decreased by about 3.13%. Under the EDP scenario (
Figure 5d,i), the growth of built-up was at a stagnant state compared to 2020. The area of forest and wetland increased by 26.96% and 11.25%, with 37.99% reduction in bare land. Cropland, water, and grassland, decreased by 8.86%, 7.96%, and 6.18%, respectively. These changes resulted in a 5.52% increase in ESV and a 1.08% decrease in economic benefits in the study area under the EDP scenario. Under the EEB scenario, the three types of ecological land (i.e., forest, grassland, and wetland) showed an increasing trend with a slowdown in the growth of built-up land (
Figure 5e,j). Cropland and water showed a decreasing trend, and the area of bare land remains stable. Compared to 2020, the ESV increased by 1.69% and the socio-economic benefits increased by 6.40%. In summary, the degree of LULC change under the EEB scenario was between the BAU scenario and the EDP scenario, so the development objectives of the EEB scenario better reflected the balance between ESV and socio-economic benefits.
4.2. Dynamic Simulation of CS at Patch Scale
The aboveground, belowground biomass, and topsoil organic carbon densities of different land use types were calculated using overlay analysis based on the LULC data and carbon density data, (
Table 4).
To explore the characteristics of CS under the three scenarios, the historical and future simulation LULC results were input into InVEST, as shown in
Figure 6. The carbon storage was characterized by “high in the northwest and low in the southeast”, and the maximum and minimum values of carbon storage are 22,608 Mg/km
2 and 5648 Mg/km
2, respectively. As shown in
Table 5, forest, cropland, and grassland, contributed the most to the carbon storage. Through
Figure 6f,g, the expansion of build-up in the plains and the encroachment of forest in higher elevations lead to a general decreasing trend of CS in the study area. Compared with 2000–2010, the area of CS decline increased and the degree of decline deepened from 2010–2020. From 2000 to 2010, CS in the study area increased by 9.78 Tg, mainly in the “Northern Sand Control Zone” in the northwestern part, showing a connected aggregation. During 2010–2020, the CS in the study area decreased by 3.27 Tg, and the carbon sink capacity only part of the “Water Containment Ecological Function Reserve” showed an increasing trend.
For the three scenarios, the spatial pattern of CS in 2030 was the same as that in 2010–2020. The increase in CS was most significant in the northwestern part of the study area under the EDP scenario, followed by the EEB scenario, while the change was not obvious under the BAU scenario. Under the BAU scenario, CS in most areas of the study area showed a decreasing trend, with the most significant decrease in the North China Plain, and the CS increase area was mainly located in the Bohai Sea coastal zone (
Figure 6c,h). Under the EDP scenario, the expansion of forest in the northwest and wetlands in the northeast led to the increase of regional carbon stock at 148.01 Tg, while the change in CS in the North China Plain was smaller (
Figure 6d,i). The variation of CS in EEB scenario was between EDP and BAU scenarios. As shown in
Table 5, the study area had the highest CS at 2684.89 Tg under EDP scenario, which is an increase of 5.83% (148.01 Tg) compared to 2020. Under EEB scenario, future CS was 2599.39 Tg, which is an increase of 2.46% compared to 2020. The CS was the least in BAU scenario (2525.86 Tg), which is a decrease of 0.43% (11.02 Tg).
4.3. Evolution Prospect of CS at Regional Scale
The results in the previous
Section 4.2 shows the distribution pattern and evolutionary characteristics of CS at the patch scale. To further explore the CS variation at the regional scale, the CS of 13 cities in the study area were counted with the help of geographic information technology, as shown in
Table 6. Combined with the spatial distribution of CS changes under different time periods and scenarios (
Figure 6), the 13 cities showed significant spatial heterogeneity. In terms of spatial location from 2000 to 2010, CS in the northern part of the study area showed a decreasing trend, and in the central and southern parts of the study area showed an increasing trend. From 2010 to 2020, CS in the northeast of the study area showed an increasing trend, while the CS in the southwest showed a decreasing trend. For each city, Chengde had the highest CS in 2010 and 2020, with 620.06 Tg and 620.92 Tg, respectively. In 2010 and 2020, both Baoding and Beijing hold CS greater than 200 Tg, Shijiazhuang, Tangshan, Handan, Xingtai, and Cangzhou, were larger than 100 Tg, and Langfang had the smallest CS at less than 50 Tg. From 2000 to 2010, the CS in Zhangjiakou and Chengde showed a significant increasing trend, with an amount of 8.70 Tg and 1.27 Tg, respectively. The CS in Handan and Tangshan showed a minor decrease trend, by 0.23 Tg and 0.19 Tg, respectively. From 2010 to 2020, the CS in Qinhuangdao, Chengde, Tangshan, Tianjin, and Beijing, showed a clear increasing trend, with an increase of 3.39 Tg, 0.86 Tg, 0.77 Tg, 0.67 Tg, and 0.59 Tg, respectively. The CS in Zhangjiakou, Baoding and Shijiazhuang showed a clear trend of decrease, by 5.73 Tg, 1.82 Tg, and 0.94 Tg, respectively.
The results also showed that the changes of CS under the three scenarios have significant differences, both the northern and western parts of the study area showed an increasing trend in CS, while the central and southern parts showed a decreasing trend in CS. Under BAU scenario, the CS of all 13 cities showed a decreasing trend, with the most obvious decrease in Beijing (1.08 Tg). Under EDP scenario, the CS of all 13 cities showed an increasing tendency, with the increase of CS in Chengde, Zhangjiakou, Qinhuangdao, and Baoding, being greater than 10 Tg. Under EEB scenario, the CS of four cities (i.e., Tianjin, Langfang, Cangzhou, and Hengshui) were relatively stable, while the CS of the remaining cities show an increasing trend, with the most significant increase (15.70 Tg) in Chengde. Therefore, the carbon sink function in Hengshui and Cangzhou faces certain risks in the future.
5. Discussion
5.1. MPI Coupling Model
Simulating LULC and CS dynamics is critical for rational land use planning under the background of China’s quest of “carbon neutrality” [
55]. Some researchers have pointed out that the traditional estimating models of CS have the disadvantages of extensive parameter setting and difficult operation, so the simulation models of LULC have the shortcomings of low accuracy, poor disclosure, and poor applicability [
21]. To overcome the above problems, this work suggested a MOP-PLUS-InVEST coupling model (named MPI) that fully exploits the benefits of each of them. The MOP model is a quantitative optimization model for LULC that can determine demand for each land use type according to the study area’s historical land use evolution characteristics and current territorial spatial planning policies. The PLUS model then simulates the spatial distribution of LULC based on the demand and growth probability of each land type, and during this process, spatial restrictions can be imposed on multiple regions and land use types as needed to meet the specific needs of ecological restoration, arable land conservation, and so on. Finally, by integrating the LULC simulation results into the InVEST model for investigating CS changes at several scales, the results of the spatial and temporal evolution of CS are swiftly produced. The MPI model thus aids in the formulation and adjustment of land use planning. Furthermore, the results of the MPI model can be integrated with the results of other models to broaden its applications, such as biodiversity and soil conservation [
56], calculating ESV [
57], and performing landscape pattern analysis [
58].
5.2. Evolution of LULC and CS
LULC changes have a major impact on the carbon sink function of terrestrial ecosystems [
33,
34]. Rapid urbanization and economic development have resulted in a decrease in CS in major portions of China. This research is to determine whether a “win-win” situation of ecological function and economic value can be achieved through a reasonable LULC spatial distribution. Under the BAU scenario, ESV decreased by 3.13% and economic benefit increased by 22.8%, and under the EDP scenario, ESV increased by 5.52% and economic benefit decreased by 1.08%. However, under the EDP scenario, ESV and EC increased by 1.69% and 6.40%, respectively. The result proved that the EEB scenario, which weighs economy and ecology, is a reference for BTH to achieve a coordinated and sustainable development approach. This is also consistent with a number of related studies, such as Li et al.’s study on the ecological barrier zone in Sichuan-Yunnan, China, which noted that a feasible scenario to a green economy involves balancing the ecological carrying capacity and the spatial arrangement of economic construction [
35].
LULC changes in BTH are still dominated by the mutual transfer of cropland, forest, and grassland, while the implementation of the “Grain for Green Project” has obviously accelerated the transfer of cropland out to another two types. The increase in the total area of four categories of ecological land: water bodies, wetlands, woodlands, and grasslands, is the result of the implementation of ecological conservation policies in BTH [
36]. Water is an important part of the ecological land use, and continuous water systems play a vital role in the improvement of the neighborhood environment and the full play of ecological services, but the area of water bodies in BTH still shows a trend of shrinking, and water pollution and water degradation are still urgent issues that need to be resolved.
Currently, the CS in BTH is about 2536.88 Tg, of which the aboveground biomass, belowground biomass, and topsoil organic carbon pool, account for 34.81%, 27.35, and 37.83%, respectively. It is expected that the CS in BTH will be 2525.86–2684.89 Tg in 2030. Cropland and forests, which together make up 37.58% and 31.95% CS, are the most significant land use types. Vegetation’s ability to act as a carbon sink is influenced by its phenology. Between 2000 and 2020, new forests were created on 1247.51 km2 of cropland and 5912.63 km2 of grassland, but the CS in northwest BTH expanded slowly since it was at the early age. The capacity of plants to absorb carbon increases with size, and as the forest in the BTH will continue to grow in all three scenarios, it has the potential to be a very significant carbon sink.
5.3. Potential Ecological Threat
As mentioned above, the interconversion of cropland, grassland, and forest, accounts for the majority of the carbon gains in BTH. However, local governments must also be aware of the potential ecological risks posed by such phenomena. Water performs significant roles, even though it contributes less to CS than other land use types [
17]. However, due to the current protection laws not being perfect, water bodies in BTH are at considerable risk of degrading. Forest ecological processes are significantly impacted by deforestation and fires, hence stringent conservation of forests must persist. On the other hand, investigating ecological changes at various scales and levels is a crucial step in developing a future plan for ecological issues. Because different sizes and different research subjects will lead to variances in outcomes, Liang et al.’s simulation analysis of China’s CS demonstrates that the future CS in BTH tends to decline [
21]. As a result, a multi-scale investigation was carried out in this study, and the findings indicate that the carbon sink functions of Hengshui and Cangzhou cities in the southeast of BTH are in jeopardy and require policymakers’ attention. Moreover, according to the outcome envisaged by the EEB scenario, the restoration of ecological functions and economic development are not completely contradictory, so ecological protection cannot be implemented in an all-encompassing way.
5.4. Limitations and Prospects
A new approach suitable for exploring the simulation of LULC and CS evolution at the regional scale was proposed in this research, but there are still some points that need to improve. First, the MOP model’s practical application is easily constrained by the level of detail of relevant data in the study region because it is a highly targeted LULC structure optimization method. Similar research has been cited in this paper to strengthen the LULC limitations in addition to closely adhering to the land use planning of BTH. Second, the partial lack of the LULC data and driving variables will result in missing pixels on the result images, which inherently affects the results’ correctness. Third, there are still some uncertainties in the CS estimation results. Although mature carbon density data were used, the results are uncertain because the InVEST model ignores variations in carbon density within different land use types. This is because carbon density is likely to undergo subtle dynamic changes over time due to the human activities and environmental changes.
Future studies should concentrate on: (1) integrating various techniques to look at more standardized and effective ways to improve LULC structure; (2) examining ways to build dynamic systems of land use and carbon density drivers, taking into account the dynamic effects of human activity and natural environmental changes on the ecosystems’ capacity to store carbon; and (3) making empirical measurements to examine the variation within different land use types and validating model findings with adequate empirical data to raise the precision of regional terrestrial ecosystem CS evaluations.
6. Conclusions
In rapidly urbanizing regions, economic development and anthropogenic disturbances have resulted in substantial LULC, which has a significant impact on ecosystem functions and services. It is critical to assess the spatiotemporal evolution of LULC and terrestrial ecosystem CS in these areas, especially for the BTH in China.
In this study, a MOP-PLUS-InVEST coupling model was proposed to simulate the LULC and CS in 2030 in the BTH, based on the historical trends from 2000 to 2020. The evolution of LULC in BTH were mostly represented in transfers between cropland, forest, and grassland. The historical changes were characterized by the reduction of cropland and water, and the increase of forest, built-up and bare land, among which 11,286.7 km2 of built-up land occupied a large amount of ecological land. The BAU scenario demonstrated the most dramatic increase in built-up, with small changes in forest and grassland, and hence the economic benefits rose. In the EDP scenario, forest and wetland developed dramatically, while built-up remained stagnant, restoring the ecological function. In the EEB scenario, ecological land increased while built-up expanded slowly, resulting in economic and ecological benefits. The constant shrinking of water is a pressing issue that must be addressed in the future. In terms of CS evolution, among the various land use types, forest, cropland, and grassland, contributed the most of the carbon storage. Influenced by land use distribution and topography, the distribution of CS was high in the northwest and low in the southeast. At the same time, the change of CS was affected by the construction of ecological conservation area and urban construction, showed obvious spatial differences. The 2020 CS of BTH was 2536.88 Tg, and the predicted results under BAU, EDP, and EEB scenarios were 2525.86 Tg, 2684.89 Tg, and 2599.39 Tg, respectively, with CS increasing mostly in the western and northern regions. The EEB scenario balanced ecological services and economic rewards, increased the ecosystem carbon sink function, and is an efficient way to investigate “carbon neutrality”.