Spillover Effects and Influencing Factors of Forest Carbon Storage in the Context of Regional Coordinated Development: A Case Study in Guangdong Province

Abstract

Clarifying the spatial relationships and impact mechanisms of forest carbon storage is essential for designing carbon sink policies and promoting coordinated regional and sustainable development. Using panel data from 21 cities in Guangdong Province between 2012 and 2021, this study employs the forest accumulation expansion method, exploratory spatial data analysis (ESDA), and spatial econometric models to investigate the distribution, spillover effects, and impact mechanisms of forest carbon storage. The results show the following: (1) During the study period, forest carbon storage in Guangdong Province exhibited a fluctuating upward trend and notable regional disparities, with the highest levels observed in the northern region. (2) Forest carbon storage exhibits spatial correlation characteristics and a positive spillover effect, with a value of 0.2394. (3) Temperature has a negative spillover effect on forest carbon storage, while gross regional product demonstrates a negative direct effect. In contrast, labor and afforestation are key factors that possess significant positive direct and spillover effects. Therefore, in developing forest carbon sinks, it is recommended that the government implement adaptation strategies and strengthen inter-city cooperation to promote sustainable development.

1. Introduction

Climate change stands as a major problem facing humanity in the 21st century [1]. Economic development leads to higher resource consumption, which in turn drives up carbon emissions and causes global warming [2,3]. In response, the world’s major economies have announced voluntary carbon reduction targets, aiming for carbon peak and neutrality [4]. In particular, China has committed to achieving a carbon peak by 2030 and carbon neutrality by 2060 [5]. Achievement of the “dual carbon” target requires both decreasing carbon emissions and enhancing carbon sequestration [6].

As major components of terrestrial ecosystems, forests play a central role in regulating the global carbon balance [7,8]. Forest carbon sequestration currently represents the most environmentally friendly and efficient way to mitigate climate change, and the promotion of carbon storage in forests can offset anthropogenic carbon emissions [9,10]. However, the degradation and deforestation of forests due to human activities pose a significant threat to carbon storage in forest ecosystems [11]. The tensions between ecological, environmental protection, and economic development have become increasingly clear. The 20th National Congress of the Communist Party of China emphasized coordinated regional development and highlighted comprehensive and balanced progress in the human, economic, and ecological fields [12]. With the implementation of this concept, the coordinated development of resources, environment, and economic growth has become a focus in all areas. Given this context, enhancing forest carbon storage is crucial for coordinating economic development and ecological construction, as well as promoting coordinated regional development and sustainable development.

Guangdong Province is one of the largest provinces in China in terms of population and economy. In 2019, the urbanization rate in Guangdong Province reached 71.4% [13]. However, the accelerated economic expansion has also led to a gradual decline in forest carbon storage [11]. As a major economic powerhouse in China, it is vital for Guangdong to coordinate economic growth and environmental protection [14,15]. Ensuring sustainable economic growth while increasing forest carbon storage, spearheading efforts for coordinated regional development, and serving as a role model at the national level are important challenges for Guangdong’s future development. Since the distribution of forest resources may have a regional correlation [16]. Therefore, to enhance the forest carbon storage in Guangdong Province, it is essential to explore its spatial relationships and underlying impact mechanisms.

Research on forest carbon storage originated in the 1990s [17]. After three decades of development, these studies have comprehensively addressed various topics, including aboveground biomass [18], economic valuation [19], spatial and temporal evolution patterns [20], and potential carbon storage [21]. Nevertheless, there are fewer studies on the regional correlations and spillover effects of forest carbon storage [22]. At the macro scale, Du et al. [23] reported the spillover effects in the density of forest carbon storage between 139 countries. Lü et al. [24] identified spillover effects of carbon storage in forests between the 31 provinces in China, while Zhou et al. [25] observed a spatial correlation in the cost of forest carbon sequestration between neighboring provinces. At the micro-scale, Liu et al. [26] demonstrated spatial spillover effects between counties in Shaanxi Province. Dai et al. [27] analyzed 838 forest samples in subtropical Zhejiang Province and observed significant spatial autocorrelation in the carbon densities of vegetation, soils, and litter. These research studies highlight that the spatial relationships of forest carbon storage should not be ignored.

Regarding the influencing factors, the research results indicate that forest carbon sequestration is determined by a combination of various factors. Alongi [28] found that climatic factors play an important role in carbon storage in mangroves. Meng et al. [29] found that geography affects mangrove carbon stocks through adjustments in sediment nitrogen content and plant species diversity. Yang et al. [30] demonstrated that forest land and industrial structure were major socio-economic factors affecting forest carbon storage. Furthermore, anthropogenic disturbances, such as timber harvesting [31], population growth [32], and air pollution [33], contribute to further changes in carbon storage. It is obvious that forest carbon stocks are not determined by a single factor. On the other hand, the development of forest carbon storage may also be influenced by other regional factors. Lü et al. [24] and Du et al. [23] noted that there are spillover effects of natural and economic factors, which indirectly affect forest carbon storage in neighboring regions.

While research to date has made some progress in examining the spatial relationships and impact mechanisms of forest carbon storage, the majority of studies focus on large-scale analyses, such as at the national or provincial levels. Furthermore, these studies often examine isolated factors in specific regions, lacking a systematic perspective to explore the interactions and linkages between different areas. The spillover effects of forest carbon storage at the city level and the role of various influencing factors remain underexplored. With the advancement of regional coordinated development, it is important to understand the spatial relationships and impact mechanisms of forest carbon storage at the city level. Given this, this study collects panel data for 21 cities in Guangdong Province for the period from 2012 to 2021 and investigates the spatial characteristics of forest carbon storage using the forest accumulation expansion method and ESDA. By establishing spatial models, we explore the spillover effects and impact mechanisms of forest carbon storage. We try to address the following questions: (1) Is there a spatial spillover of forest carbon storage in Guangdong, and if so, what kind of spillover exists? (2) What are the key factors influencing the development of forest carbon storage in Guangdong? These findings elucidate the spatial relationships and influencing factors of forest carbon storage in urban areas, which can provide theoretical references for the design of forest carbon sink strategies, and facilitate coordinated regional and sustainable development. The contributions of this study include two aspects. First, the spatial econometric models using panel data are developed to analyze the spillover effects of forest carbon storage between cities from the perspective of spatial correlation, rather than considering only simple linear relationships, which helps to better capture the characteristics of the development of forest carbon storage. Second, although there are many studies that focus on the driving factors of forest carbon storage, most of them only consider a single aspect of the factors. In this paper, the role of natural, socio-economic, and anthropogenic factors on forest carbon storage is examined from a systemic perspective and categorized into direct and indirect effects, providing a new perspective for promoting sustainable development.

2. Materials and Methods

2.1. Research Area

Guangdong Province (Figure 1), is located in southern China and comprises 21 prefecture-level cities. It is known to be the prominent economic engine and populous province in China, contributing 11.6% to the national GDP [34]. In addition, Guangdong Province is an important element within the Guangdong–Hong Kong–Macao Greater Bay Area, and it serves an essential role in the national development strategy and “dual carbon” goals. However, the acceleration of economic growth has caused a reduction in forest carbon storage, alongside other environmental issues. Addressing these issues requires consolidation and improvement of forest carbon storage.

For a more detailed analysis, we refer to the Guangdong Provincial Bureau of Statistics and divide the province into four regions, including northern, Pearl River Delta (PRD), eastern, and western. This allows a more systematic analysis of forest carbon storage from a regional and urban perspective.

2.2. Methods

2.2.1. Forest Accumulation Expansion Method

The calculation of carbon storage from forest stock volume is practical, easy to operate, and very accurate [36,37,38]. The formula is as follows:

$$C=V_1\delta \rho \gamma +\alpha V_1\delta \rho \gamma +\beta V_1\delta \rho \gamma$$

(1)

where C denotes the forest carbon storage (tonnes), and V₁ represents the forest stock volume (m³). The parameters δ, ρ, γ, α, and β represent the biomass expansion factor, bulk density, carbon content rate, and coefficients of carbon sequestration by understory plants and forest land, respectively. The values of these parameters are 1.9, 0.5, 0.5, 0.195, and 1.244, respectively [39].

2.2.2. ESDA

Exploratory spatial data analysis, an essential aspect of spatial econometrics, explains phenomena such as spatial correlation. The Moran’s I index is utilized to evaluate the spatial autocorrelation, incorporating local and global indexes [40]. The formula is as follows:

$$I_G={\displaystyle \frac{\left[{\sum }_{i=1}^N{\sum }_{j=1}^N{w}_{ij}\left(x_i-\overline{x}\right)\left(x_j-\overline{x}\right)\right]}{(S^2{\sum }_{i=1}^N{\sum }_{j=1}^N{w}_{ij})}}$$

(2)

where $\overline{x}={\displaystyle \frac{1}{N}}{\sum }_{i=1}^N{x}_i$, $S^2={\displaystyle \frac{1}{N}}{\sum }_{i=1}^N{\left(x_i-\overline{x}\right)}^2$; I_G denotes the global Moran’s I index; N denotes the number of regions; x_i and x_j denote the specific attribute values of regions i and j, respectively; $\overline{x}$ denotes the mean of the attribute values of each region; S² denotes the variance of the variable; and w_ij denotes the weight matrix.

The localized Moran index can further indicate the degree of correlation of variables within spatial neighborhoods. The formula is as follows:

$$I_L={\displaystyle \frac{Z_i{\sum }_{i\ne j}{w}_{ij}{Z}_j}{S^2}}$$

(3)

where $Z_i=x_i-\overline{x}$, $Z_j=x_j-\overline{x}$; I_L represents the local Moran’s I index.

2.2.3. Estimation Models

The traditional ordinary least squares regression (OLS) model explains the multivariate linear function of the relationship between the dependent and independent variables, which is as follows:

$$y_i={\sum }_{i=1}^n{\beta }_i{x}_i+{\beta }_0+{\epsilon }_i$$

(4)

where y_i denotes the forest carbon storage (tonnes), and x_i is the independent variable. β₀ is the constant term, and β_i represents the regression coefficient. ε_i is the random error term.

To further investigate the potential spillover effects of forest carbon storage, we used the spatial econometric models. In addition, we use the Lagrange Multiplier (LM) test, Likelihood Ratio (LR) test, and Wald test for model selection, ensuring the rationality of the selected model. Based on the models of Zhao et al. [22] and Du et al. [23], the model is formulated as follows:

$$y=\rho Wy+X\beta +WX\delta +{\mu }_i+{\gamma }_t+u$$

(5)

where y denotes the forest carbon storage (tonnes). X denotes the independent variables, and β represents the regression coefficient. ρ and δ represent the spatial regression coefficient. μ and γ represent the individual and time effects, respectively. W represents the spatial weight matrix. In this study, we use geographic distance matrix to analyze and use adjacent weight matrix to perform robustness tests. u represents a random perturbation term. If λ = 0, it is a spatial Durbin model (SDM), if λ = 0 and δ = 0, it is a spatial lag model (SLM), and if ρ = 0 and δ = 0, it is a spatial error model (SEM).

2.3. Theoretical Mechanism

2.3.1. Spatial Spillover Effects of Forest Carbon Storage in Urban Area

The variability of forest resources in different regions and the mobility of materials in nature make it difficult to study the forest carbon storage in isolation. Various factors not only exert direct effects on forest carbon stocks but also indirectly influence it through regional correlations. These interactions can either reinforce or offset each other, leading to a combined effect on forest carbon storage. Therefore, this paper constructs a research framework for forest carbon storage based on spatial economic theory which posits that spatial correlations are inherent [41]. For forest carbon storage, on one hand, the flow of factors such as population and capital facilitate economic and industrial agglomeration. Coupled with political diffusion and competition effects among neighboring regions, policies implemented in one area often extend to surrounding areas, transcending spatiotemporal boundaries and exerting broader impacts. On the other hand, due to the gaseous nature of carbon dioxide, it readily diffuses to neighboring areas, resulting in spatial spillover. Consequently, this may lead to significant spatial spillover effects in regional forest carbon storage.

2.3.2. Impact Mechanisms of Forest Carbon Storage

Forestry represents a dynamic process, in which the natural and economic production are intertwined, with the carbon sequestration function of forests being determined by a variety of factors. First, natural factors play a crucial role in determining the spatial differentiation of forest vegetation carbon storage. Variations in the natural environment significantly influence the growth conditions for vegetation. Among these, climatic factors are particularly impactful, as they dictate afforestation feasibility and tree species selection. These factors directly affect the spatial distribution of forest vegetation, thereby exerting a pronounced influence on forest carbon storage. Second, the rapid pace of economic development and urbanization drives continuous adjustments in land use patterns, which affects the forestry industry structure and the investments in forestry activities. This eventually leads to fluctuations in forest area and standing stock, affecting the carbon sequestration capacity of forests. Finally, anthropogenic forest management strategies including afforestation, logging, and conservation, have a direct impact on forest structure and area. These practices modify the composition and health of forests, thereby influencing the levels of carbon storage. In conclusion, changes in the forest carbon storage result from an interplay of natural, economic, and anthropogenic factors. Combined with the above analyses, the impact mechanism of forest carbon storage is shown in Figure 2.

2.4. Variable Selection and Description

Based on the previous mechanism analyses, we used forest carbon storage, calculated according to Formula (1), as the explained variable and selected natural, socio-economic, and anthropogenic factors as explanatory variables to comprehensively explore the impact mechanisms of forest carbon storage.

Precipitation (Pre, unit: mm) and temperature (Tem, unit: degree Celsius). Natural factors like precipitation and temperature are essential in forest carbon storage [42]. Therefore, we selected annual precipitation and mean annual temperature as indicators.

Gross regional product (Grp, unit: ten billion CNY). The level of economic development significantly influences carbon sinks and sources [8,43]. Economic growth, which often accompanies industrialization and urbanization, alters forest areas and consequently affects forest carbon storage. Thus, we chose the urban gross regional product to reflect the impact of economic development.

Labor (Lab, unit: one hundred persons). The number of forestry practitioners can also influence forest carbon storage [44]. Hence, we used the number of people working in forestry units in each city to represent labor.

Afforestation (Aff, unit: thousand hectares). It has been shown that afforestation significantly increases vegetation carbon storage [45]. Nevertheless, Wu et al. [46] suggested a lag period in the effect of afforestation. Garten [47] observed that carbon storage in plantation forests was higher in older trees than younger ones. Following the approach of Fu et al. [48], we used the afforestation area from ten years ago as a proxy. The sum of afforestation, site renewal, and low-yield forest renovation areas in the current year represented afforestation.

Forest management (Mgt, unit: thousand hectares). The intensity of anthropogenic forest management correlates significantly with variations in carbon stocks [49]. Therefore, we selected the area of cultivated middle-aged and young forests as a measure of forest management. Since the effect of forest management on carbon storage also exhibits a lag [46], data from ten years ago were used as a proxy.

Harvesting (Harv, unit: ten thousand m³). Forest timber harvesting has a considerable influence on forest carbon storage [50]. Thus, we select the volume of commercial timber output to represent timber harvesting. The descriptive statistics for each variable are presented in

Table 1. Descriptive statistics of variables.

Variable		Symbol	Mean	Std. Dev.	Min	Max
Explained variables	Forest carbon storage	-	30.8667	30.3953	2.0252	116.6280
Explanatory variables	Precipitation	Pre	1828	333.9	1078.4	2743.3
	Temperature	Tem	22.9298	0.8953	20.0818	24.6513
	Gross regional product	Grp	42.8782	60.2863	5.2428	306.6485
	Labor	Lab	11.0765	8.2672	0.66	36.66
	Afforestation	Aff	6.9603	6.7429	0.211	32.543
	Forest management	Mgt	5.0340	5.3435	0.012	27.339
	Harvesting	Harv	39.6981	44.6671	0.0027	221.3387

.

2.5. Data Sources

The data on forest stock volume, afforestation data, and forest timber harvesting data are derived from the main statistics of forest resources in Guangdong Province and forest inventory data. The climate data of precipitation and temperature are derived from the China Meteorological Data Network. Additional socio-economic data are derived from the Guangdong Rural Statistical Yearbook, city yearbooks within Guangdong Province, and the China Economic Net Statistics Database (https://ceidata.cei.cn/, accessed on 11 January 2023). Any missing data were supplemented using linear interpolation.

3. Results

3.1. Spatial and Temporal Distribution of Forest Carbon Storage

Figure 3 shows the development of forest carbon storage across Guangdong Province. Furthermore, to further understand the spatial distribution of forest carbon storage in Guangdong Province, we used the natural breaks method to categorize the average annual forest carbon storage in each city, which was divided into four tiers (Figure 4) from 2012 to 2021. These results demonstrate the variation and discrepancy in forest carbon storage on an urban scale. Shaoguan is the city with the highest forest carbon storage, while Shantou is the lowest. Moreover, the discrepancy in forest carbon storage is increasing year on year, indicating that cities with abundant forest resources and adequate ecological protection possess higher levels of forest carbon storage, which also experience a more pronounced rise.

From a regional perspective, forest carbon storage in northern Guangdong is significantly higher than in other areas. The overall distribution follows the pattern of “highest in the north, followed by the west, and lower in the PRD and eastern regions”. The northern region possesses extensive forested areas due to favorable geographic conditions. In June 2018, the Guangdong Provincial Party Committee proposed transforming northern Guangdong into an ecological development zone, prioritizing green development. As a result, the northern region exhibits the highest forest carbon storage. In the western region, cities have higher forest carbon storage due to slower industrialization and urbanization. There is a significant disparity between core and peripheral cities in the PRD. Core areas with high urbanization rates result in lower carbon storage. This phenomenon is also evident in eastern Guangdong. The forest carbon storage in the core cities of the PRD and eastern Guangdong consistently lags behind, indicating that there is still significant development potential for forest carbon storage in these cities.

Regarding development trends, cities with medium to high forest carbon storage have expanded prominently. However, there remains a substantial gap, with a polarized distribution pattern. Since 2012, most cities have experienced increased forest carbon storage, except for Zhanjiang and Zhuhai. In Zhanjiang, forest carbon storage declined until 2018, likely due to the high proportion of commercial forests, short rotation periods, and excessive timber harvesting. In Zhuhai, forest carbon storage sharply decreased from 2015 to 2018, possibly due to rapid urban development leading to overexploitation of forest resources and extreme weather events, such as typhoons and heavy rainfall.

3.2. Spatial Agglomeration Analysis of Forest Carbon Storage

As shown in

Table 2. Global Moran’s I test of forest carbon storage in Guangdong Province from 2012 to 2021.

Index	2012	2013	2014	2015	2016	2017	2018	2019	2020	2021
Moran’s I	0.260	0.255	0.247	0.239	0.240	0.243	0.244	0.242	0.242	0.242
Z value	2.59	2.55	2.49	2.42	2.43	2.45	2.46	2.44	2.44	2.44
p value	0.005	0.005	0.007	0.008	0.008	0.007	0.007	0.007	0.007	0.007

, all Moran’s I indices are greater than zero and pass the significant test. It suggests that the forest carbon storage exhibited significant spatial agglomeration characteristics in Guangdong during the study period. Additionally, the spatial clustering of forest carbon storage demonstrates dynamic changes over time. Between 2012 and 2015, the index declined from 0.260 to 0.239. Subsequently, it fluctuated between 0.240 and 0.244, remaining relatively stable.

These changes are related to the relevant policies implemented by the Guangdong Provincial Government during this period. In 2012, Guangdong Province formulated the “Guangdong Province Forest Carbon Sinks Key Ecological Project Construction Plan (2012–2015)”, which provides policy support for the development of forest carbon storage. However, varying implementation in different cities led to a decline in spatial correlation. Since 2015, most cities in Guangdong Province have promoted the creation of national forest cities, which effectively supports the coordinated development of forest carbon storage in all regions. Consequently, Moran’s I increased and remained relatively stable after 2015.

In our study, we use the Moran scatterplot to illustrate the clustering of forest carbon storage among cities. In the Moran scatterplot, the first quadrant (H-H) represents the city and its neighboring cities with high forest carbon storage. The second quadrant (L-H) demonstrates that the city has low forest carbon storage while the surrounding cities have high. The third quadrant (L-L) represents the city and its neighboring cities with low forest carbon storage. The fourth quadrant (H-L) shows that the city exhibits high forest carbon storage while the surrounding cities display low. Simultaneously, “high” and “low” refer to relative values, indicating the magnitude of forest carbon storage in each city.

Moran scatter plots were conducted for 2012, 2015, 2018, and 2021, as shown in Figure 5. Most cities are located in the first and third quadrants during the study period. These results refute the random distribution of forest carbon storage in Guangdong and further confirm significant spatial agglomeration. The first quadrant predominantly consists of northern cities, where rich forest resources and ecological protection policies have fostered the development of forest carbon storage clusters. The third quadrant primarily includes cities in the PRD and eastern regions. In these regions, extensive urban development has reduced forest area, leading to an agglomeration of low forest carbon stock cities. Fewer cities are in the second and fourth quadrants, highlighting a notable disparity in forest carbon stocks between these cities and their neighbors.

There were no obvious changes in most cities during the entire study period. Although Yunfu has transitioned from the second to the first quadrant, the majority of cities remain in the third quadrant. The results demonstrate that there is considerable scope for enhancement of forest carbon storage in Guangdong.

3.3. Regression Analysis

3.3.1. Results of OLS Analysis

As shown in

Table 3. The regression results of OLS.

Variable	Coefficient
Pre	−4.3809 (−1.4318)
Tem	−7.6485 *** (−5.4835)
Grp	−0.0399 ** (−2.1911)
Lab	1.5673 *** (9.7427)
Aff	1.0739 *** (5.5437)
Mgt	−0.0399 (−0.2031)
Harv	0.1495 *** (5.0853)
Intercept	185.3934 *** (5.5461)
R2	0.7834

Note: **, and *** denote significance levels at 10%, 5%, and 1%, respectively. T statistic values in parentheses.

, all other factors except precipitation and forest management have a significant impact on forest carbon storage. Specifically, temperature and gross regional product exhibit a negative effect on forest carbon storage, while labor, afforestation, and harvesting have a positive effect.

3.3.2. The Results of the Model Selection Test

The spatial correlation analysis demonstrates a significant spatial agglomeration in forest carbon storage across various cities. Omitting the spatial relationship may result in inaccurate estimates. Thus, this paper introduces spatial econometric models and selects the appropriate model using the tests.

Table 4. The results of diagnosis and selection test of spatial measurement model.

Testing Method	Coefficient
LM test (no spatial lag)	7.7058 ***
Robust test (no spatial lag)	2.8207 *
LM test (no spatial error)	30.0383 ***
Robust test (no spatial error)	25.1533 ***
Wald test spatial lag	14.7427 **
LR test spatial lag	16.6606 **
Wald test spatial error	16.0446 **
LR test spatial error	15.8275 **

Note: *, **, and *** denote significance levels at 10%, 5%, and 1%, respectively.

shows that all four LM tests reject the null hypothesis. Given that the SDM addresses the limitations of both the SLM and the SEM, it is initially concluded that the SDM is appropriate. Additionally, the LR and Wald tests evaluate whether it is possible to simplify the SDM to either an SEM or an SLM. Both tests reject the hypothesis, confirming the appropriateness of the SDM.

As shown in

Table 5. The results of the spatial econometric models.

Variable	Model (1)	Model (2)	Model (3)	Model (4)	Model (5)
Pre	2.1216 (1.4968)	2.2601 (1.4924)	5.9689 (1.4141)	0.7631 (0.4462)	2.6084 (1.5687)
Tem	0.4621 (0.5116)	0.5784 (0.6079)	−17.3415 *** (−12.0314)	2.1082 ** (2.147)	0.8906 (0.9127)
Grp	−0.0301 ** (−1.9819)	−0.033 ** (−2.1773)	−0.0703 *** (−4.8678)	−0.0299 * (−1.865)	−0.0304 ** (−2.0047)
Lab	0.1471 *** (2.6986)	0.1314 ** (2.4908)	1.0804 *** (8.443)	0.1172 ** (2.1132)	0.1516 *** (2.8191)
Aff	0.14 ** (2.4206)	0.1206 ** (2.1189)	0.7406 *** (5.0451)	0.1273 ** (2.2088)	0.1278 ** (2.2683)
Mgt	0.0316 (0.6394)	0.0358 (0.7483)	0.0739 (0.5008)	0.0506 (1.0173)	0.0069 (0.1408)
Harv	−0.0128 *** (−1.5262)	−0.0132 (−1.5714)	0.0647 *** (2.8483)	−0.0086 (−0.9493)	−0.011 (−1.2894)
W × Pre	-	-	16.4867 ** (2.4289)	0.6988 (0.3506)	−1.001 (−0.3815)
W × Tem	-	-	−2.5154 *** (−4.4466)	−1.3171 (−1.0834)	−4.8733 ** (−2.4311)
W × Grp	-	-	0.0588 (1.159)	0.1258 *** (3.5754)	0.013 (0.27)
W × Lab	-	-	0.724 * (1.6702)	−0.0404 (−0.2844)	0.2561 (1.6383)
W × Aff	-	-	1.1655 ** (2.367)	0.2253 (1.4882)	0.386 ** (2.0824)
W × Mgt	-	-	−0.6902 (−1.6205)	−0.0966 (−0.6747)	−0.1807 (−1.2121)
W × Harv	-	-	0.0071 (0.0893)	−0.0028 (−0.0868)	−0.0146 (−0.4452)
ρ	0.2851 *** (2.9319)	-	−0.123 (−1.2249)	0.468 *** (5.839)	0.2394 ** (2.4054)
λ	-	0.42 *** (4.8518)	-	-	-
R2	0.9915	0.9914	0.8944	0.9908	0.9922

Note: Model (1) represents the SLM with double-fixed effects, Model (2) represents the SEM with double-fixed effects, Model (3) represents the SDM with time-fixed effects, Model (4) represents the SDM with spatial-fixed effects, and Model (5) represents the SDM with double-fixed effects. *, **, and *** denote significance levels at 10%, 5%, and 1%, respectively. T statistic values in parentheses.

, the R² value of the SDM with double-fixed effects is higher than that of the other models. Consequently, we utilize it for further estimation and analysis.

3.3.3. Spatial Econometric Model Results

The results of the selected model showed that the coefficient (ρ) passed the significant test at the 5% level, with a value of 0.2394. This finding suggests that the forest carbon storage in cities across Guangdong Province exhibits a notable spatial spillover effect. Specifically, every increase of 1 million tonnes in forest carbon storage in surrounding cities will boost forest carbon storage by 0.2394 million tonnes in the city. Therefore, the spatial correlation of forest carbon storage between cities is caused by the spillover effect.

3.3.4. Decomposition of Spillover Effects

Since the coefficients of SDM cannot directly indicate the marginal effects of independent variables on the dependent variable, this study proceeds to decompose the spatial spillover effect of each independent variable [22]. As shown in

Table 6. The results of decomposing spillover effects.

Variable	Direct	Indirect	Total
Pre	2.5656 (1.5624)	−0.4856 (−0.1544)	2.08 (0.7064)
Tem	0.7026 (0.706)	−6.0466 ** (−2.4713)	−5.344 ** (−2.1591)
Grp	−0.0296 * (−1.9994)	0.008 (0.1308)	−0.0216 ** (−0.3454)
Lab	0.1665 *** (2.9724)	0.3762 * (1.8397)	0.5427 ** (2.3765)
Aff	0.1496 ** (2.5828)	0.542 ** (2.204)	0.6916 ** (2.6334)
Mgt	−0.0032 (−0.0608)	−0.2385 (−1.1989)	−0.2417 (−1.0847)
Harv	−0.0122 (−1.4751)	−0.0235 (−0.5563)	−0.0356 (−0.8411)

Note: *, **, and *** denote significance levels at 10%, 5%, and 1%, respectively. T statistic values in parentheses.

, the direct effect represents the influence of variables within this city, the indirect effect represents the influence of adjacent urban variables on this city, and the total effect denotes the total of the direct and indirect effects [51].

For direct effects, gross regional product, labor, and afforestation are significant, while other variables are not. Specifically, the gross regional product negatively impacts forest carbon storage, indicating that economic development inhibits its growth, which is consistent with the findings of Wu et al. [43]. Economic development accompanied by urbanization often destroys forests, resulting in a decrease in forest carbon storage. Conversely, labor and afforestation positively affect forest carbon storage. More forestry workers can address issues like pests, diseases, forest fires, and illegal logging more effectively, boosting the carbon sequestration capacity of forests. Increased afforestation can expand forest areas. Thus, enhancing the afforestation contributes to the growth of forest carbon storage.

Regarding indirect effects, temperature, labor, and afforestation pass the significance tests, while the remaining variables do not. Specifically, the temperature exhibits a negative spillover effect, which suggests that the temperature fluctuations in neighboring cities will adversely affect forest carbon storage in the local city. Generally, the forests thrive in favorable water conditions, and rising temperatures reduce water supply, limiting forest growth, and carbon sequestration capacity [52]. Due to similar temperature changes in adjacent areas, a rise in temperature in neighboring cities often indicates a corresponding increase in the local city [23]. Consequently, the rise in temperature in neighboring cities leads to a reduction in the local city’s forest carbon storage. In comparison, labor and forest afforestation exhibit positive spillover effects. Optimization of forestry labor and afforestation efforts in one region will set a positive example for neighboring areas. As regions imitate these practices, their forest quality, biomass, and area covered improved. Therefore, labor and afforestation in neighboring cities positively impact the forest carbon reserves of the local city.

3.3.5. Robustness Test

For the spatial model, the spatial weight matrix has a large impact on the regression results, so we use different spatial weight matrices for the robustness test. The selected matrix is the adjacent weight matrix.

The estimation results (

Table 7. The regression results based on the adjacency weight matrix.

Variable	Direct	Indirect	Total
Pre	2.3391 (1.3242)	2.0646 (0.7321)	4.4037 * (1.8355)
Tem	−0.488 (−0.3783)	0.8406 (0.2983)	0.3526 (0.1589)
Grp	0.0301 * (−1.9598)	0.0071 (0.2305)	−0.023 (−0.6797)
Lab	0.1836 * (3.5093)	0.1726 (1.254)	0.3562 ** (2.2553)
Aff	0.1496 ** (2.7139)	0.5409 *** (4.2411)	0.69 *** (4.6339)
Mgt	−0.0011 (−0.0218)	−0.2989 ** (−2.2593)	−0.3 * (−1.8936)
Harv	−0.011 ** (−1.3454)	−0.002 (−0.1092)	−0.013 (−0.6728)

Note: *, **, and *** denote significance levels at 10%, 5%, and 1%, respectively. T statistic values in parentheses.

) show that the results of the decomposition of the effects of each explanatory variable are basically consistent with the previous paper. The coefficients and significance of the direct effects show little difference, while the differences in coefficients and significance for the indirect effects are more pronounced. This is mainly because the weights between a larger number of cities in the adjacent weight matrix are set to 0, so the estimation of the spillover effect is changed.

4. Discussion

4.1. Spatial and Temporal Distribution Characteristics of Forest Carbon Storage

This study has found a fluctuating upward tendency in forest carbon storage across Guangdong Province. This is related to various forestry projects implemented in Guangdong Province since 2012. The implementation of forestry policies can promote the development of forest carbon sinks [43]. Spatially, the forest carbon storage in northern Guangdong is significantly higher than that in other regions. This is consistent with the research findings of Xu et al. [53]. In addition to being influenced by resource endowment, this is also closely related to the development of urbanization. Generally, economic development in highly urbanized areas often comes at the cost of environmental sustainability [54]. This results in the forest carbon storage in economically developed regions being significantly lower than that in less developed regions. Therefore, in the process of developing carbon sinks, it is essential for regional governments to consider their specific circumstances and balance economic development with ecological protection.

4.2. Spatial Spillover Effects and Driving Factors of Forest Carbon Storage

We identified a positive spillover effect in forest carbon storage within Guangdong Province, which is consistent with the findings of previous scholars at global and national scales [23,24]. This evidence demonstrates that both local factors and the spillover effects of adjacent cities influence the development of forest carbon storage in urban areas. The value of the spillover effect calculated in this paper was 0.2394. Compared with previous studies, the value is slightly smaller than that of Du et al. [23]’s study (0.678). This may be related to the different research scales. In this paper, we focus on the city level as our research scale; thus, the spillover effect of forest carbon storage may not be as significant as that observed between countries. Furthermore, Fu et al. [48] have studied the spillover effect of forest carbon sinks in 31 provinces and municipalities in China, and found that the value of the spillover effect was 0.238. This further confirms that the spillover effect of forest carbon storage may vary across different spatial scales. This finding suggests that the interplay of forest carbon storage among cities should not be ignored, and development strategies that solely focus on the particular circumstances of an individual city are inherently limited. Considering the above results, it is recommended that the government enhance intercity exchanges and capitalize on the radiation effect of high-carbon storage areas to increase the forest carbon stocks more effectively.

Regarding influencing factors, temperature, gross regional product, labor, and afforestation are closely related to forest carbon storage in Guangdong Province. Notably, temperature, labor, and afforestation exhibit spatial spillover effects exceeding their direct effects. Piao et al. [55] pointed out that the increase in carbon sinks is primarily attributed to changes in the environment. Therefore, the spillover effects of various factors on regional forest carbon storage may occur through their impacts on the environment. Compared with previous studies, in this research, the effects of temperature [23], labor [16], and afforestation [48] on forest carbon storage are relatively consistent with the conclusions of earlier research. This result further validates that the development of urban forest carbon storage requires coordination among cities. Conversely, the direct and spillover effects of precipitation, forest management, and harvesting are insignificant. Although these findings diverge from some existing studies, there is no consensus on the relationship between these factors and forest carbon storage in the academic community. For instance, Xue et al. [16] found that reasonable harvesting facilitates forest carbon storage growth, while Du et al. [23] argued that timber harvesting reduces forest stock volume, negatively impacting carbon storage. Therefore, the relationship between these factors and forest carbon storage in Guangdong Province requires further investigation.

Moreover, combining the coefficients of direct and spillover effects, it is evident that natural factors contribute more significantly to forest carbon storage than human and socioeconomic factors. While natural factors have a greater influence, they are generally less amenable to change than anthropogenic factors in the actual development process. This implies that targeted adaptation strategies should be implemented for natural factors, such as “matching species with the site” and optimizing forest structure. Regarding human factors, effective regulatory strategies should be formulated, including afforestation, tree species selection, and nursery thinning. Similarly, the impact of economic development is less pronounced than human factors, as economic factors typically do not directly affect forest carbon stocks. Concerning the significance of influencing factors, only the direct and spillover effects of labor and afforestation pass the statistical significance tests. This is consistent with previous research conclusions, namely that forest carbon storage is more significantly influenced by human factors [56]. These findings indicate that human factors, such as labor and afforestation, are key to affecting the development of urban forest carbon storage. Consequently, it is suggested that the government prioritize human factors in policymaking.

4.3. Limitations and Further Research

This paper presents an in-depth study of the spatial spillover effects and impact mechanisms of forest carbon storage using spatial Durbin models, thereby extending the theory in forest carbon sink development and innovation management. However, our study has some limitations. Firstly, due to limited data availability, we employed linear interpolation to supplement some missing data, introducing a degree of error. Secondly, although this study includes several factors, additional unquantifiable factors may impact forest carbon storage.

In the future, researchers can further refine the calculation methods by incorporating remote sensing data to calculate forest carbon storage more accurately. Furthermore, innovations in other statistical methods can also mitigate potential biases in the results. Future scholars can also investigate the interaction between influencing factors and their spatial and temporal effects to further identify the impact mechanisms of forest carbon storage.

5. Conclusions and Policy Recommendations

5.1. Conclusions

Exploring the spatial relationships and influencing factors of forest carbon storage is crucial for formulating effective carbon sink strategies and fostering coordinated regional development. This study utilizes panel data for 21 cities in Guangdong Province between 2012 and 2021. We employ the forest accumulation expansion and ESDA to examine the evolution characteristics of forest carbon storage. On this basis, we construct spatial models to investigate the spillover effects and impact mechanisms of forest carbon storage. The main findings are as follows.

First, forest carbon storage in most cities in Guangdong Province demonstrates a fluctuating upward trajectory. The spatial pattern reveals that northern cities have the highest levels of forest carbon storage, followed by western cities, with central and eastern cities having the lowest levels. Considerable discrepancies exist between cities, with Shaoguan showing the highest, while Shantou shows the lowest.

Second, forest carbon storage exhibits spatial agglomeration in Guangdong Province. The local pattern is characterized by high forest carbon storage showing agglomeration in the north, whereas cities with low exhibit clustering along the PRD and eastern regions.

Third, forest carbon storage exhibits a demonstrably positive spillover effect in Guangdong Province. For each 1 million tonnes improvement in forest carbon storage in adjacent cities, the city’s own will increase by 0.2394 million tonnes. Regarding influencing factors, temperature has a negative spillover effect, while gross regional product has a negative direct effect. In contrast, labor and afforestation are key factors affecting forest carbon storage, with significant positive direct and spillover effects.

5.2. Policy Recommendations

Based on the research findings, this paper proposes the following policy recommendations:

Firstly, for cities in northern Guangdong, priority should be given to strengthening the management and protection of existing forests and safeguarding the stability of the forest environment. For cities in the eastern and western Guangdong Province, and the Pearl River Delta, priority should be given to exploring forestry actions aimed at increasing carbon sinks to enhance the carbon sequestration capacity of forests.

Secondly, the forest carbon storage of cities in Guangdong Province exhibits significant spatial correlation and spillover effects. It is necessary to fully leverage the radiation effect of high carbon sink areas, promote the development of carbon sink projects in neighboring cities, establish cross-regional carbon sink information sharing and development platforms, strengthen inter-city connections, achieve inter-regional resource sharing, and facilitate the coordinated development of the region.

Finally, considering the different influences of various factors on forest carbon storage, the government should formulate differentiated development strategies. The government should take into account the direct and indirect impacts of climate, economic conditions, human interference, and other factors on forest carbon storage based on the actual situation of each region. Furthermore, it should focus on the role of key factors, such as afforestation and labor, in order to promote the enhancement of the overall forest carbon sequestration capacity.