Spatio-Temporal Pattern, Evolution and Influencing Factors of Forest Carbon Sinks in Zhejiang Province, China

Abstract

Zhejiang Province, charged with the task of opening “an important window”, should take the lead in the implementation of the national “carbon peak, carbon neutral” strategy. The intention of this study is to start from the intersection of economics and geography, and on the basis of GIS analysis, use the spatial Dubin model to study the spatio-temporal evolution trend, key influencing factors, and spatial spillover effects of forest carbon sequestration in 72 districts and counties of Zhejiang Province from 2000 to 2020, to provide a theoretical and practical reference for how to formulate regional synergistic forest sink enhancement measures and help Zhejiang achieve the goal of carbon neutrality. The conclusions are as follows: (1) From 2000 to 2020, the overall growth of forest carbon sequestration in all regions and counties of Zhejiang Province was relatively high, but the regional differences were large. The concentration center and range moved to the southwest. (2) If the spatial effect is considered, the increase of per capita GDP, road density, and deforestation will reduce the forest carbon sequestration, and the conservation of ecological public welfare forests will have a significant positive relationship with the forest carbon sequestration.

1. Introduction

In order to respond to climate change and demonstrate its role as a great power, China formally made a commitment to the international community in 2020 to achieve “carbon peak” and “carbon neutrality” (hereinafter referred to as “double carbon”) and elevated it to a national development strategy [1]. The scientific formulation of the “double carbon” action plan has become one of the focuses of governments and academics at all levels. Zhejiang, as an “Important Window”, is a major economic province in China, so how is it to ensure sustained and stable economic growth while scientifically formulating the “double carbon” strategy, action plan, and realization path, in order to take the lead in achieving the “double carbon” goal? It is a major issue and political task for Zhejiang’s future economic and social development to set an example for the whole country by formulating a scientific “double carbon” strategy, action plan, and path to achieve the goal of “double carbon”, while ensuring sustained and stable economic growth.

According to statistics, the GDP per capita of Zhejiang province has crossed the threshold of 80,000 Yuan in 2020, entering the ranks of medium developed countries; carbon emissions have increased from 120 million tons in 1997 to 417 million tons in 2019, with an average annual growth rate of 6.7%, much higher than the national average growth rate of 5.70%. Emissions data are derived from the latest revision of energy data from the China Bureau of Statistics (2020). For Zhejiang, it is difficult to achieve carbon neutrality by relying entirely on emissions reduction, and it must vigorously develop ecological carbon sinks, including forest carbon sinks. The forest resources inventory data show that as of 2021, Zhejiang’s forest coverage rate was 61.15%, much higher than the national average of 22.96%; among them, the proportion of young and middle-aged forest stock was 58.49%, higher than the national average of 17.69%. However, the forest unit stock in Zhejiang province is only 63 cubic meters/ha, a large gap considering the national average of 80 cubic meters/ha; this also indicates that in 2021, Zhejiang province promulgated the “Zhejiang Province Carbon Peak and Carbon Neutral Science and Technology Innovation Action Plan”, and took forest sink as an important way to achieve the “double carbon” goal. The forestry department, in addition has started to formulate the medium and long-term development plan for forestry sink.

In recent years, there have been many studies on the spatial distribution, influencing factors and spillover effects of forest carbon sequestration and enhancement, but there is still room for improvement. The initial research on forest carbon sinks mainly focused on the measurement of above-ground biomass, but with the expansion of research on forest carbon sequestration and sink enhancement, including the enrichment of carbon pools and the exploration of the mechanism model of sink enhancement, the research on the theory of influencing factors, spatial and temporal evolution patterns and spillover effects of forest carbon sinks has expanded. In order to promote future inter-regional synergistic sinks and reduce the cost of achieving the “double carbon” target, the spatial distribution and influencing factors of forest carbon sequestration and sinks need to be sorted out and summarized. The spatial distribution and influencing factors of forest carbon sequestration and sink need to be reviewed and summarized. Regarding the spatial distribution of forest carbon sinks, Zhao et al. [2] simulated and analyzed the spatial pattern of forest carbon sinks in China, and Luo et al. [3] analyzed the spatial distribution of forest carbon sinks from a provincial perspective. Most of the existing studies are based on regional and even national and global macroscopic perspectives, while studies on smaller-scale districts and counties are rare, and the counties are the foundation and important link for realizing the modernization of governance capacity. Regarding the influence factors and spillover effects of forest carbon sinks, Xue et al. [4] showed that there are significant spatial spillover effects of precipitation and forest harvesting. Forest disaster also has a significant impact on forest carbon sinks, while Du et al. [5] showed that the global forest carbon sinks have a spillover effect on neighboring countries through the study of NGO activity, forest area growth, and fuelwood use. Existing studies focus on the independent description of natural or anthropogenic factors in the influence of forest carbon sequestration capacity, lacking a systematic perspective to explore the interaction and association between natural and anthropogenic factors. Thus, the findings of existing studies cannot provide a more scientific basis for policy-making.

Based on the ArcGIS platform, this paper uses forest carbon stocks to characterize forest carbon sinks at the county scale and objectively reflect the spatial and temporal evolution patterns of forest carbon sequestration and sinks in Zhejiang province from 2000 to 2020ZZX. In order to provide theoretical and practical references for the formulation of regional synergistic forest sink, enhancement measures are needed to help Zhejiang achieve its carbon neutrality target.

2. Mechanism Analysis

2.1. Analysis of the Mechanism of Carbon Sequestration and Sink Enhancement in Forests

Forestry is a process in which natural and economic production are intertwined, and the mechanism of forest carbon sequestration and carbon sink enhancement is shown in Figure 1. First, natural factors such as slope, elevation, temperature, and precipitation influence the difficulty of afforestation, species selection, and the difficulty and frequency of nurturing. Second, with rapid economic development and urbanization, the spatial pattern of land is constantly adjusted, which further influences the selection of afforestation, economic investment in nurturing, and the area and frequency of harvesting. Third, relevant forestry policies, such as afforestation subsidy policies, nurturing subsidy policies, and forest harvesting quota directly affect afforestation, nurturing, and harvesting. Eventually, afforestation, nurturing, and harvesting in turn affect forest area, forest stock, and ultimately forest carbon sinks [6]. As a result, forest carbon sinks are influenced by a combination of natural, economic and social, and policy factors.

2.1.1. Natural Factors

Based on forest life cycle theory, carbon sequestration by forest land is mainly influenced by natural factors such as temperature, precipitation, and elevation. First, the rise in temperature will increase the soil carbon content, thus promoting the growth of vegetation biomass [7]. Second, regarding precipitation, according to Brown et al. there is a positive correlation between precipitation and forest vegetation carbon density at 400~3200 mm and a negative correlation above 3200 mm [8]. Third, altitude affects temperature, and the higher the altitude, the less suitable conditions are for human survival and development; thus, the degree of human intervention is reduced, which is conducive to vegetation “rest and recuperation” [9].

2.1.2. Socio-Economic Factors

The impact of economic and social factors on forest resources has two sides. From a macro perspective, on the one hand, population growth has led to urban expansion, and the phenomenon of “forest to construction land” and “forest to farmland” has emerged, leading to a decrease in carbon stocks [10,11]. On the other hand, with the improvement of living standards with economic development, residents’ vision of a better ecological environment gradually increases, and a series of afforestation and reforestation, as well as forest management and care, are carried out, which is conducive to the accumulation of forest carbon sinks [12,13,14]. From the perspective of farmers’ personal wishes, with economic and social development, the main source of their income is non-farm income, thus reducing the possibility of deforestation and creating conditions for the “rest and recuperation” of forests; however, it is also due to the long-term engagement in non-farm business that the management of forests is reduced, which affects forest quality [15]. Finally, the increase in farmers’ income while timber prices remain constant over time means that the relative price of timber decreases and farmers no longer cut down forests for economic gain, which to some extent facilitates the restoration of forest ecosystems.

2.1.3. Forestry Policy Factors

Ecological public welfare forests and logging quotas have a double effect on forest management. In the short term, these can maintain forest area and increase carbon sink, but in the long term, they may also reduce afforestation and enthusiasm for its management. For one thing, ecological public welfare forests increase farmers’ enthusiasm for forest management and nurturing through financial subsidies, thus enhancing the growth and carbon sequestration capacity of forest trees; in addition, ecological public welfare forests improve the quality of forests by prohibiting logging, forbidding changes in forest land use, and reducing human intervention, thus promoting an increase in carbon sinks, but not in the long term, considering that carbon sinks may become carbon emissions after excessive forest maturation [16]. Second, forest harvesting leads to the reduction of forest area and unit accumulation, which leads to the decrease of carbon stock; it is worth noting that, as a means of forest management, limiting the harvesting index or reasonably planning the harvesting quota for each region according to the natural growth optimization of forest accumulation, although it can effectively improve the growth environment of forest trees, the replacement of mature forests by young forests in the short term will inevitably lead to the decrease of carbon stock [4].

In summary, the mechanistic model of forest carbon sequestration and sink enhancement is constructed as shown in Function (1).

$$C=f\left(Nature,Economy Society,Policy\right)$$

(1)

where $C$ denotes the amount of forest carbon sink, and $f\left(·\right)$ denotes the function of forest carbon sink with respect to natural (N), economic and social (ES), and policy (P) factors, and the economic and social factors are represented by GDP per capita, population density, road density, and urbanization rate, while the policy factors are represented by forest harvesting quota and ecological public welfare forest holdings.

2.2. Analysis of the Spillover Effects of Forest Sink Enhancement

In January 2022, General Secretary Xi Jinping proposed to achieve “double carbon” under the unified leadership and layout of the central government, and to formulate scientific and reasonable roadmaps and timetables based on the resource endowments and development levels of different regions, in order to promote the achievement of the “double carbon” target. In this process, to achieve the double carbon goal, it is necessary to promote synergy, including the synergy of carbon reduction and sink increase, the synergy between regions, the synergy between departments, and the synergy between policies, etc. Therefore, for Zhejiang, in order to take the lead in achieving the “double carbon” target, it is also necessary to scientifically formulate the “double carbon” action plan and formulate the corresponding policy system according to the principle of collaborative promotion. In terms of forest sink, it is necessary to identify its spatial and temporal patterns in different regions and to clarify the potential space of forest sink, key influencing factors, and spillover effects between different regions, in order to provide theoretical support and decision-making references for the achievement of forest carbon sink growth and carbon neutrality in Zhejiang province.

According to “the first law of geography”, all things are related, but nearby things are more related than distant things [17]. The spatial spillover effect of forest carbon sinks is reflected in the fact that the factors influencing the growth of forest carbon sequestration and sinks in a certain region will have positive or negative effects on the growth of forest carbon sinks in other regions. First, there is an inter-regional spillover effect of forest carbon sinks because the physical geography, the basic condition for forest growth, and the continuity of the geographic distribution of forest resources determining its forestry development and its carbon sink growth are correlated between regions [4]. Secondly, due to the existence of the land indicator occupancy balance mechanism, this system stipulates that the area of construction land, agricultural land, and other types of land in various cities of China each year is under the control of a certain target. Specifically, if an area (especially a developed area) in China uses excess land as construction land due to development needs, the agricultural land area in this area may already be occupied. Currently, it is necessary to purchase agricultural land indicators as in other areas with redundant agricultural land, so that China’s construction land, agricultural land, and other types of land area are consistent with the overall goal. In fact, the area of forestry land between different regions will inevitably be affected through inter-regional transactions. Finally, as human beings pay more attention to forest ecological protection and implement the concept of “Clear waters and green mountains are as good as mountains of gold and silver” proposed by General Secretary Xi, it is increasingly urgent for the provincial government to guide by macro policies and formulate overall planning, so that forest carbon sinks show a trend of agglomeration development, which would indicate that external policies promote the positive spillover of forest carbon sinks [18].

3. Materials and Methods

3.1. Research Methodology

In this paper, we use a panel regression method with fixed time and region, which can solve the problem of endogeneity caused by factors that do not vary over time and are unobservable, such as natural factors with regional characteristics included in the fixed effects.

If the spatial element is not considered, the fixed effects model is used as the baseline model and set as follows.

$$Y_{it}=\alpha +\beta X_{it}+{\delta }_i+{\lambda }_t+{\epsilon }_{it}$$

(2)

where the subscripts $i$ and $t$ denote the region and time, respectively, and $Y_{it}$ is $i$ area, $t$ the forest carbon sink in the time period, and $X_{it}$ are the relevant natural, economic, social, and policy influences in the text, and ${\delta }_i$ and ${\lambda }_t$ are the area and time fixed effects, respectively. $\alpha$ and $\beta$ are parameters to be estimated, and ${\epsilon }_{it}$ are the error terms.

Since each neighboring county has similarities in natural conditions, economic and social conditions, and forestry policy implementation, and these conditions affect carbon sinks, thus making the carbon sinks between neighboring counties possibly have a spillover effect, a spatial Durbin model was used to analyze whether there is a spillover effect. Since the spatial Durbin model analysis must be conducted on the basis of the existence of spatial correlation of forest carbon sinks in neighboring counties, Moran’s I test for spatial correlation was used first to confirm the existence of spatial correlation before the spatial Durbin model was used for the study. Natural factors such as temperature, precipitation, and elevation were included in the fixed effects due to their regional characteristics, and therefore, natural factors were not considered in the empirical part.

The global Moran’s I formula is as follows.

$$I=\frac{n{{\displaystyle \sum }}_{i=1}^n{{\displaystyle \sum }}_{i=1}^n{w}_{ij}\left(x_i-\overline{x}\right)\left(x_j-\overline{x}\right)}{{{\displaystyle \sum }}_{i=1}^n\left(x_i-\overline{x}\right){{\displaystyle \sum }}_{i=1}^n{{\displaystyle \sum }}_{i=1}^n{w}_{ij}}$$

(3)

Of which $n$ is the total number of county units. $x_i$ and $x_j$ denote the amount of carbon sequestered in county $i$ and the amount of carbon sequestered in county j. $\overline{x}$ is the average amount of carbon sequestered in each county in the study area. $w_{ij}$ is the spatial weight matrix, and this paper uses the Rook neighborhood weight matrix after row normalization Rook adjacency weight matrix: When two regions are not adjacent, the spatial weight is 0; when two regions are adjacent, the spatial weight is 1 [19,20].

If the spatial element is considered, the spatial Durbin model is set as follows [21,22].

$$Y_{it}=\rho W{Y}_{it}+\beta X_{it}+\phi W{X}_{it}+{\delta }_i+{\lambda }_t+{\epsilon }_{it}$$

(4)

where $i$, $t$, $Y_{it}$, $X_{it}$, ${\delta }_i$, ${\lambda }_t$, $\beta$ and ${\epsilon }_{it}$ consistent with the underlying regression implications, this paper $W$ is the row-standardized Rook neighborhood weight matrix, and the robustness tests are performed using the row-standardized K-nearest neighborhood weight matrix K-nearest neighbor weight matrix: set each region has K neighboring regions, the spatial weight of neighboring regions is the inverse of the distance, and the spatial weight of non-neighboring regions is 0. According to the rule of thumb, the value of K in this paper is set to 5 and the distance-based spatial weight matrix (Distance-based spatial weight matrix: the spatial weight of neighboring regions within the threshold of each region is the inverse of the distance between the two regions; the spatial weight of regions outside the threshold is 0). The threshold of this paper is 49.4815 km, and this threshold ensures that each region has at least 1 neighboring region and is the spatial lag coefficient of the dependent and independent variables, respectively.

3.2. Data sources

The research sample of this paper is 72 districts and counties in Zhejiang province. (Municipal districts are combined in this paper due to the small number of carbon sinks and the geographical proximity). The forest carbon sink data were obtained from the forest carbon sink data of Zhejiang Forest Resource Center in 1999, 2004, 2009, 2014, and 2019.

Data on economic and social factors: GDP per capita, population density, road density, and urbanized population data were obtained from the statistical yearbooks of Zhejiang cities in 2000, 2005, 2010, 2015, and 2020; urbanized area data were obtained from the 30 m precision China Land Cover Map of the Chinese Academy of Sciences. Among them, population density is characterized by the ratio of resident population to administrative area, road density is characterized by the ratio of road mileage in the territory to administrative area, and urbanization rate is selected as an explanatory variable in this paper.

Data on forestry policy factors: Data on forest harvesting quota were obtained from the announcement of annual forest harvesting quota during the Ninth to Thirteenth Five-Year Plan; data on public welfare forest area were obtained from the area of public welfare forest above the provincial level in each district and county of Zhejiang province in 2004, 2009, 2012, 2015, 2019, and 2021 published by the General Office of the People’s Government of Zhejiang Province, and the data of missing years were processed by the interpolation method.

4. Results

4.1. Spatio-Temporal Pattern, Evolution of Forest Carbon Sink in Zhejiang Province

Based on the data of forest carbon sinks in Zhejiang province in 2000, 2005, 2010, 2015, and 2020, the spatial and temporal evolution pattern of forest carbon sinks in Zhejiang province was analyzed, and the results are shown in Figure 2. The overall change trend of the province is as follows: From the time change, the total carbon sink in Zhejiang province increased from 128,452,900 tons in 2000 to 280,704,400 tons in 2020, an increase of 118.53%. During 2000–2005, 2005–2010, 2010–2015, and 2015–2020, their increases were 22.37%, 15.86%, 23.31%, and 25.00%, respectively. The growth rate of forest carbon sink decreases and then increases, and the largest increase is from 2015–2020. Spatially, western and southwestern Zhejiang are the high-value gathering areas of forest carbon sinks in Zhejiang, while the carbon sinks in eastern regions are relatively low.

The temporal and spatial changes in counties are also large, as shown in Figure 3. In terms of increase, most of the districts and counties increased their carbon sinks to various degrees from 2000 to 2020, among which the most significant increase was in Yiwu, Dongyang, Wuyi, Yongkang, and Pujiang counties under Jinhua City, with an increase of more than 160%. Jiaxing, Huzhou, Zhoushan’s subordinate districts and counties, however, rose less than 60%, or even had negative growth, and the average level was 118.53% in Zhejiang province; the reason for that is that the plains and islands had less forest land planning, and their carbon storage often depends on urban green land. Through the comparison of GDP per capita and carbon sink of each district and county in Zhejiang province from 2000 to 2020, a preliminary pattern can be found that the level of economic development and forest carbon sink shows a negative correlation. In terms of stock, among them, Chun’an County ranks first in the province in terms of forest carbon sinks, with the annual average forest carbon sinks accounting for 5.42% of the province’s total, while Lin’an District and Longquan City rank second and third, respectively. Most of the important forest carbon sink areas in Zhejiang province belong to the mountainous areas in western and southern Zhejiang. Jiashan County, Jiaxing City District, and Tongxiang City are in the Hangjiahu Plain, and Shengsi County and Dongtou District are in the smaller islands, where forest resources are scarce and carbon sinks are low, and the sum of annual average carbon sinks in these five areas from 2000 to 2020 only accounts for 0.15% of the province’s total, with a lowest annual average carbon sink of 1654.99 tons in Jiashan County.

4.2. Descriptive Statistics

In this paper, data from 72 regions (districts and counties) in Zhejiang province were used as samples for analysis, and the study spanned every five years from 2000 to 2020 due to the more serious missing data before 2000. Descriptive statistics of each variable are shown in

Table 1. Descriptive statistics of forest carbon storge, socio-economic variables and policy variables.

Variables			Unit	Mean	S.D.	Min	Max
Explained variables		Forest carbon storge (carbon)	Hundred tons	27,028.38	24,649.96	12.64	158,431.90
Explanatory variables	Socio-economic variables	GDP per capita (gdp)	Yuan/person	54,066.76	50,710.55	3250.00	345,617.60
		Population density (popu)	People/km2	583.30	488.65	87.18	4383.83
		Road density (road)	m/km2	936.87	489.45	166.67	2784.42
		Urbanization rate (urban)	‰	31.40	45.70	0.40	378.42
	Policy Variables	Forest harvesting quota (harv)	m3	17,040.51	21,900.88	28.00	132,762.00
		Ecological public welfare forest holdings (ecol)	hectares	33,725.79	29,964.61	209.00	160,556.00

.

4.3. Regression Analysis

The results of the region and time fixed effects models for the panel data without spatial effects are shown in column 2 of

Table 2. Results of empirical analysis, including disregarding spatial effects and considering spatial effects.

	Disregarding Spatial Effects	Consider Spatial Effects
		Main	W
Gdp	−0.036 *	−0.029 **	0.040 **
	(−1.92)	(−2.26)	(2.42)
Popu	5.765 *	4.019	−5.329
	(1.67)	(1.43)	(−0.87)
Road	−3.672 ***	−3.178 ***	3.318 **
	(−3.38)	(−2.68)	(2.23)
Urban	−42.152 *	−15.666	−76.857 *
	(−1.74)	(−0.68)	(−1.87)
Harv	−1.104 ***	−0.997 ***	−0.048
	(−4.79)	(−12.30)	(−0.31)
Ecol	0.281 ***	0.247 ***	−0.045
	(9.47)	(9.05)	(−1.02)
Carbon			0.323 ***
			(4.30)
Regional fixed effects	Yes	Yes
Time fixed effects	Yes	Yes
R2	0.879	0.824

Note: ***, **, * denote 1%, 5%, and 10% significance levels, respectively. Numbers in parentheses are t-values. Same as in the following tables.

. According to the results of area and time fixed effects, all variables are significant. Among the economic and social factors, GDP per capita, road density, and urbanization rates have negative and significant effects on forest carbon sinks with coefficients of −0.036, −3.672, and −42.152, respectively; population density has a positive and significant effect on forest carbon sinks with a coefficient of 5.765; among the policy factors, forest harvesting quota has a negative and significant effect on forest carbon sinks with a coefficient of −1.104; ecological public welfare forest. Among these factors, the coefficient of population density is the opposite of the expected direction. The possible explanation is that at the beginning of the 21st century, the concept of “Two Mountains”, lucid water and lush mountains, are invaluable assets, was put forward to provide new ideas to solve the contradiction between economic development and ecological protection in the new era, and the concept of ecological protection gradually gained popularity. Therefore, despite the 115.7% increase in population density from 2000 to 2020, the amount of forest carbon sink in Zhejiang further increased [23].

If there is a spatial spillover effect on forest carbon sinks in Zhejiang province districts and counties, the setting of the non-spatial model will be erroneous. To further test whether there is a spatial effect and select a suitable spatial model, global Moran’s I analysis, the LM test, the robust LM test, and the LR test were conducted in this study. Through the global Moran’s I statistical analysis of forest carbon sink data in Zhejiang province for five periods, the results obtained were 0.463, 0.500, 0.497, 0.474, and 0.470, which all passed the significance test at the significant level of 1%, indicating that the forest carbon sink in Zhejiang province showed a positive correlation in spatial distribution and satisfied the higher regional clustering characteristics required for spatial econometric analysis. This indicates that the spatial distribution of forest carbon sinks in Zhejiang province shows a positive correlation and meets the higher regional concentration characteristics required for spatial econometric analysis. Through the LM test and robust LM test, it is concluded that the choices of both the SEM model and the SAR model are appropriate, and the SDM model combining both is selected at this time; through the LR test, it is concluded that the spatial model cannot be simplified to the SAR or SEM model, therefore, the spatial Durbin model with two-way fixed effects is selected for the study of spatial effects for this paper.

The results of the spatial Durbin model considering spatial effects are shown in columns 3 and 4 of Table 2.

Regarding the socio-economic factors, the increase of GDP per capita and road density will reduce the forest carbon sink. Economic and social development are at the expense of certain forest ecosystem development, and an increase in GDP per capita will lead to a decrease in carbon sinks. If road density increases, it will lead to non-forest land encroaching on forest land or cultivated land, thus negatively affecting the carbon sink in the region.

In terms of policy factors, the increase in deforestation will directly reduce the existing carbon sink of the forest. Although the mature forest will be replaced by young forest after deforestation, the carbon sequestration capacity of young forest is much less than that of mature forest, thus negatively affecting the carbon sink of the region. The relevant national laws and regulations explicitly prohibit cutting down ecological public welfare forests; they also strictly control the harvesting and renewal of public welfare forests, and prohibit harvesting, nurturing, and transformation activities for the main purpose of producing timber. Therefore, the increase in ecological public welfare forests is actually the implementation of the “rest and recuperation” policy for forests, thus improving the carbon sequestration capacity of forests.

It is noteworthy that the factors of population density and urbanization rate show insignificant characteristics after considering spatial effects. The reason is that the population-related variables, i.e., population density and urbanization rate, have a strong agglomeration effect, and the high spatial similarity leads to multicollinearity, making the two variables less significant.

4.4. Further Analysis of the Spatial Spillover Effect

In the spatial Durbin model, the coefficients do not fully reflect the effects of the explanatory variables on the explained variables. To further understand the difference in the degree of influence of influencing factors after considering the spatial effects, how much of the impact is caused by the influencing factors in the region, and how much of the impact is caused by influencing factors in other regions, the direct and indirect (spillover) effects of the influencing factors need to be further discussed [24]. The estimation results are presented in

Table 3. Direct, indirect and total effects, using the spatial Dubin model.

Dependent variable	Direct Effect		Indirect Effects		Total Effect
	Coefficient	t-Value	Coefficient	t-Value	Coefficient	t-Value
Rook Neighborhood Weight Matrix
Gdp	−0.026 **	−1.98	0.040 **	1.97	0.014	0.55
Popu	3.603	1.31	−4.822	−0.62	−1.219	−0.14
Road	−2.876 **	−2.57	3.099 *	1.71	0.223	0.11
Urban	−21.888	−0.98	−110.297 **	−2.28	−132.185 **	−2.51
Harv	−1.022 ***	−12.70	−0.486 ***	−2.97	−1.508 ***	−7.59
Ecol	0.250 ***	8.94	0.047	0.95	0.297 ***	5.48

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

, where the direct effect is the extent to which the explanatory variables in the region affect the explanatory variables in the region, the indirect effect is the extent to which the explanatory variables in the surrounding region affect the explanatory variables in the region, and the overall effect is the extent to which changes in the explanatory variables in all regions affect the explanatory variables in the region [25].

Regarding the direct effect, per capita GDP and road density are negatively related to forest carbon sequestration in this region. An increase of 1 yuan per capita GDP in this region will reduce the carbon sequestration of this region by 2.6 tonnes; adding 1 m of roads per square kilometre would reduce the region’s carbon stock by 287.6 tonnes; a higher number of public welfare forests in the region has a significant positive effect on its forest carbon sinks; and deforestation has a negative significant relationship with forest carbon sequestration in this region. An additional hectare of ecologically beneficial forest in the region would add 25 tonnes of carbon storage, and an additional cubic metre of forest felling quota would reduce the region’s carbon storage by 102.2 tonnes. The comparison of the parameter values in Table 2 with the direct effect parameter values in Table 3 shows that there are some differences between the parameters in the two tables due to the feedback effect of the influencing factors in the region on the carbon sinks in other regions. For example, the direct effect of GDP per capita on the forest carbon sink in the region is −0.026, while the spatial Durbin model estimates −0.036, which indicates that for every 1 Yuan of GDP per capita in the region, the forest carbon sink will increase by 0.01 tons due to the feedback effect, after having an impact on the neighboring regions.

Regarding the indirect effects, the per capita GDP and road density have significant positive spillover effects, which are the opposite of the local effects, and the urbanization rate has a significant negative spillover effect. An increase of 1 yuan in per capita GDP in neighboring regions will add 4 tonnes of carbon storage in this region. Adding one metre of roads per square kilometre in adjacent areas would increase carbon stocks by 310 tonnes in the region; policy factors, such as a significant negative spillover effect of the felling quota and an increase in the felling quota of one cubic metre in surrounding areas will save the region 48.6 tonnes of carbon.

Robustness Tests

In order to examine the robustness of the spatial Durbin influence factor model and the subsequent spatial effect decomposition model, this study uses the K-nearest neighbor weight matrix and the distance-based spatial weight matrix to test the robustness of the spatial Durbin model and its direct, indirect, and total effects [20,26]. It can be found that the significance of using different weight matrices is basically the same, and the direction is basically the same, which generally verifies the regression results on robustness, as shown in

Table 4. Robustness tests of spatial Dubin model and its direct and indirect effects.

Dependent Variable	Main		W		Direct Effect		Indirect Effects		Total Effect
	Coefficient	t-Value	Coefficient	t-Value	Coefficient	t-Value	Coefficient	t-Value	Coefficient	t-Value
K-nearest neighbor weight matrix
gdp	−0.019	−2.26	−0.002	2.42	−0.019	−1.45	−0.011	−0.48	−0.030	−1.22
popu	4.152	1.43	−0.89	−0.87	4.070	1.60	0.923	0.14	4.993	0.68
road	−3.317 ***	−2.68	2.355	2.23	−3.104 ***	−2.90	1.886	0.79	−1.219	−0.51
urban	−2.666	−0.68	−121.717 ***	−1.87	−10.939	−0.53	−171.168 ***	−3.44	−182.107 ***	−3.35
harv	−0.971 ***	−12.30	−0.139	−0.31	−0.998 ***	−12.97	−0.599 ***	−2.80	−1.596 ***	−6.68
ecol	0.267 ***	9.05	−0.156 ***	−1.02	0.262 ***	9.63	−0.101*	−1.78	0.161 ***	2.65
Distance-based spatial weight matrix
gdp	−0.020 *	−1.46	−0.01	−0.10	−0.023 *	−1.89	−0.031	−1.38	−0.055 **	−2.02
popu	3.56	1.59	−3.142	−0.18	3.228	1.28	−2.106	−0.32	1.122	0.15
road	−2.723 **	−2.88	0.332	1.27	−2.742 ***	−2.70	−1.299	−0.53	−4.041	−1.54
urban	0.518	−0.13	−62.999 **	−3.05	−9.769	−0.47	−103.314 **	−2.25	−113.083 **	−2.10
harv	−0.924 ***	−12.39	0.177	−0.78	−0.956 ***	−12.33	−0.348 **	−2.11	−1.304 ***	−6.10
ecol	0.261 ***	10.03	−0.157 ***	−3.28	0.254 ***	9.79	−0.071	−1.21	0.184 ***	2.78

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

.

5. Discussion

In this paper, the spatial Doberman model is used to study the impact of economic and social factors and policy factors on forest carbon sequestration. Some results of the direct and indirect effects of the spatial Doberman model are generally recognized differently, so this chapter will further discuss them.

Regarding direct effects, among the socio-economic factors, GDP per capita and road density have a negative significant relationship with forest carbon sequestration in the region. The growth of GDP per capita means that more land space will be used for industrial development in the region, which compresses the space for forest growth. According to the calculation formula of forest carbon sink, forest area and forest stock are the main determinants of forest carbon sink. The increase of road density and deforestation will negatively affect forest area and forest stock, respectively, resulting in the reduction of forest carbon sequestration. Among the policy factors, public interest forest holdings have a significant positive effect on forest carbon sequestration in the region, and deforestation has a negative significant relationship with forest carbon sequestration in the region. Due to the inferiority of forests in terms of public goods and externalities, they are prone to market failure and undermine the incentive of forestry producers. Public benefit forests play a key role in maintaining ecological balance, protecting biodiversity and improving the ecological environment, etc. Since the implementation of public benefit forest protection subsidies in Zhejiang in 2001, farmers have been highly motivated to protect forests, which directly contributes to the increase of public benefit forest holdings above the provincial level, and thus the increase of forest carbon sinks [27].

Regarding indirect effects, among the socio-economic factors, GDP per capita and road density have a significant positive spillover effect, which is the opposite of that of the region, while urbanization rate has a significant negative spillover effect. This indicates that the development of forest carbon sinks in Zhejiang should be complementary and synergistic and cannot be achieved alone. According to Xia and Yang [28], the amount of forest carbon sink is related to the main functional zoning of the region. The main function zoning is based on the theory of regional function, which considers the natural ecological system and economic and social system and divides the land space into specific function types to promote economic and social development according to local conditions. The places with high per capita GDP are often optimized development zones and key development zones, and the surrounding areas are restricted development zones and prohibited development zones, according to the principle of coordination, so the per capita GDP has a positive spillover effect. According to Bai et al. [29], roads are essential infrastructure engineering facilities for forestry production and consumption and forest management and maintenance, and they are necessary for humans to enter the forest, know the forest, contact the forest, and protect the forest; therefore, the increase of road density plays a positive role in the growth of carbon sink in neighboring areas. The urbanization rate reflects the area of urban land; the higher the urbanization rate, the smaller the area of forest land, and the urbanization of Zhejiang has a clustering effect; so there is also a negative effect on the neighboring areas [30]. Among the policy factors, forest harvesting quotas have a significant negative spillover effect. According to Dai et al. [31], from the perspective of forest ecological product consumption, forest ecological products have the property of public goods and are inseparable as a whole; although they have administrative boundaries, this does not imply the existence of natural geographical boundaries, and the results of this paper are consistent with this view—there is a significant negative spillover effect of deforestation on nearby areas.

6. Conclusions

This paper uses GIS methods and the spatial Durbin model to study the spatial and temporal evolution patterns of forest carbon sinks in Zhejiang province and their influencing factors, which leads to the following conclusions.

(1) The overall change trend of the province is as follows: From the time change point of view, the total carbon sink in Zhejiang increased from 128,452,900 tonnes in 2000 to 280,704,400 tonnes in 2020, which is a 118.53% increase. From the spatial point of view, western and southwestern Zhejiang are the high-value gathering areas of forest carbon sink in Zhejiang, while the carbon sink in the eastern region is relatively low; the spatial and temporal changes in the county are also large. In terms of increase, the most significant increases are in Yiwu City, Dongyang City, Wuyi County, Yongkang City, and Pujiang County under Jinhua City, which are all up by more than 160%. Jiaxing, Huzhou, and Zhoushan under the jurisdiction of the district and county rose less than 60%. In terms of stock, Chun’an County ranks first in the province in forest carbon sink, with the annual average forest carbon sink accounting for 5.42% of the province’s total. Jiashan County, Jiaxing City District, and Tongxiang City have scarce forest resources and low carbon sinks.

(2) If the spatial spillover effect is not considered, economic and social factors such as GDP per capita, road density, and urbanization rate will have negative and significant effects on forest carbon sinks; population density has positive and significant effects on forest carbon sinks. Among the policy factors, forest harvesting quota has a negative and significant effect on forest carbon sink; ecological public welfare forest retention has a positive and significant effect on forest carbon sink.

If spatial effects are considered, regarding the economic and social factors, the increase of GDP per capita and road density will reduce forest carbon sink. Regarding the policy factors, the increase of deforestation will reduce the forest carbon sink, and there is a significant positive relationship between ecological public interest forest holdings and forest carbon sink.

To further discuss the spillover effects of each influencing factor—regarding direct effects, among the socio-economic factors, GDP per capita and road density have a negative significant relationship with forest carbon sequestration in the region; among the policy factors, deforestation has a negative significant relationship with forest carbon sequestration in the region, and forest public interest forest holdings have a significant positive effect on forest carbon sink in the region. Regarding the spillover effect, among the socio-economic factors, GDP per capita and road density have significant positive spillover effects, which is the opposite of the direction of the effect on this region, and urbanization rate has a significant negative spillover effect; among the policy factors, urbanization rate and forest harvesting quotas have significant negative spillover effects.