Topography, Diversity, and Forest Structure Attributes Drive Aboveground Carbon Storage in Different Forest Types in Northeast China

Abstract

Forests regulate air quality and respond to climate change by storing carbon. Assessing the driving factors of forest aboveground carbon (AGC) storage is of great importance for forest management. We assumed that different forest types would affect the relationship between species richness, stand density, individual tree size variation, and AGC. In order to test and verify it, we analyzed the inventory data of 206 fixed plots (20 m × 20 m) of Jingouling Forest Farm, taking advantage of the piecewise structural equation model (pSEM) to explore the effects of species diversity, stand structure attributes, and topography on the AGC storage in the Wangqing Forest in Jilin Province. In addition, in this study, we aimed to investigate whether the fixed factors (species diversity, stand structure attributes, and topography) influenced AGC storage more significantly than the random factor (forest type). According to the results of pSEM, the selected factors jointly explain the impact on 33% of AGC storage. The relationship between stand density and AGC is positive, and the impact of individual tree size variation on AGC storage is negative. Species richness has direct and indirect impacts on AGC storage, and the indirect impact is more significant through individual tree size variation. Both elevation and slope are significantly negatively associated with AGC storage. Forest type explains the impact on 12% of AGC storage, which means the relationship between AGC and predictors varies across forest types. The results provide a scientific basis for the protection and management decision of natural forests in northeastern China.

1. Introduction

The forest is the largest carbon pool on earth, and its importance in the carbon sink is considerable [1]. The aboveground part of the forest ecosystem carbon pool is particularly crucial, which can directly affect the carbon flux between the atmosphere and the forest ecosystem. Therefore, estimating aboveground carbon (AGC) storage accurately assists in mitigating the impact of climate change.

There are numerous driving factors of forest AGC storage, which depend on species diversity [2] as well as forest structure attributes, such as stand density [3] and individual tree size variation. Forest diversity is part of the influencing factors of aboveground biomass and AGC storage and is also affected by environmental factors [4]. The two main hypotheses explaining the positive correlation between species diversity and AGC storage are the niche complementarity hypothesis and mass ratio hypothesis [5,6]. The niche complementarity hypothesis holds that diversity can enhance AGB through maximizing resource utilization efficiency among co-occurring species or individuals through niche complementarity or facilitation. The mass ratio hypothesis predicts that AGB is mainly driven by the traits of dominant plant species [7,8]. Many researchers have considered the relationship between plant species diversity and AGC storage. Some research results are positive, which means the AGC storage of trees will increase with the increase in forest species diversity [9,10,11]. Some researchers believe that they are negatively correlated [12]. Therefore, extensive research is still necessary to explore the relationship between species diversity and AGC storage.

Individual tree size variation also affects the growth and development of forest communities [13,14,15]. Some researchers [16] concluded that the impact of species diversity on aboveground biomass in natural forests is achieved by increasing tree size inequality. Furthermore, individual tree size variation and stand density are extremely crucial to improve canopy filling and stratification. Aboveground biomass would increase by strengthening the interference of light, which can indirectly influence the AGC storage [8,17,18,19].

Abiotic factors also determine the growth of trees, directly altering the AGC storage [20,21]. Topographic factors play key roles in regulating forest structure, diversity, forest carbon storage [8,22,23]. Topography indirectly affects the growth of trees by causing heterogeneity in soil and light [24]. AGC storage varies along different elevation gradients [25], elevation changes the distribution of trees through the change of temperature, which indirectly affects the tree diversity. The slope has a certain impact on the soil, thus affecting the growth and development of trees [26,27,28,29].

Forest productivity and aboveground biomass showed significant differences among forest types in previous studies [30,31], and the difference exists in carbon storage among different forest types [32,33]. Although multiple driving factors affect AGC storage [30,34], few studies have proved whether the influence of these factors will change among different forest types. In general, more research is still required to explore the relationship between predictors and AGC storage in different types of forests.

The main purpose of this study was to examine the effects of diversity factor and stand structure attributes and topographic factors such as elevation and slope on AGC storage in different forest types. In this study, we sought to solve the following questions:

(1) How do topographic factors, diversity factors, and stand structure attributes affect AGC storage? (2) Would the driving factors of carbon storage change in different forest types? (3) Do fixed effects (topography, diversity, and stand structure attributes) contribute more to AGC storage than random effects (forest type)?

2. Materials and Methods

2.1. Site Description

The study area is located in Jingouling Forest Farm of Wangqing Forestry Bureau in Jilin Province (Figure 1), 130°05′–130°20′ E, 43°17′–43°25′ N. It belongs to the Changbai Mountain System in the eastern mountainous area of Jilin Province. The landform belongs to low mountains and hills, with an elevation of 300–1200 m and an average slope of 10°–25°. The region has a temperate continental monsoon climate, with an average annual temperature of 3.9 °C. The month with the lowest temperature in January, with an average minimum temperature of −32 °C. The temperature is the highest in July, with an average maximum temperature of 32 °C; the annual frost-free period is 138 days; the average annual precipitation is 600–700 mm, mainly in July [35]. The main soil type is dark brown soil, with an average thickness of about 40 cm.

The main tree species are Fraxinus mandshurica Rupr., Picea jezoensis var. microsperma (Lindl.) Cheng et L. K. Fu, Betula costata Trautv., Tilia amurensis Rupr., Pinus koraiensis Sieb. et Zucc., and Acer pictum subsp. mono (Maxim.) H. Ohashi. The main shrubs are Philadelphus incanus Koehne, Spiraea pubescens Turcz., Acer ukurunduense Trautv. et Mey., Acer tegmentosum Maxim., Lonicera japonica Thunb., and Corylus mandshurica Maxim. The main herbs are Deyeuxia langsdorffii (Link) Kunth, Aegopodium alpestre Ledeb., Urtica fissa E. Pritz. and Oxalis corniculata L.

Permanent sample plots in Jingouling Forest Farm were established in 1987–1988, with the size of 20 m × 20 m; in the tree layer (DBH ≥ 5.0 cm), the tree species name, diameter at breast height (DBH, cm), coordinates (x, y) were recorded. In addition, the elevation, aspect, slope, and other data of the sample plot were also measured and recorded. The aspect 0° represents the due north and increases to 360° clockwise. Since their establishment, the sample plots have been investigated every year.

We treated the inventory data of 2017 as the original data. After eliminating the abnormal data, the data of 206 fixed plots were ultimately utilized in this study.

2.2. Quantification of Variables Used in Analyses

Aboveground carbon storage was the response variable in this study. Based on the DBH of trees in fixed sample plots, we utilized the allometric equations [36,37] (

Table A1. Allometric equations for the calculation of aboveground biomass for each tree species.

Species	Equation	References
Abies holophylla	agb = 1000 × 0.0737 × (dbh)2.51264	Chen and Zhu 1989
Abies nephrolepis	agb = 1000 × 0.0737 × (dbh)2.51264	Chen and Zhu 1989
Acer pictum subsp. mono	agb = 10(1.930+2.535×log10(dbh))	Wang et al. 2006
Betula platyphylla	agb = 10(2.159+2.367×log10(dbh))	Wang et al. 2006
Betula costata	agb = 10(2.214+2.400×log10(dbh))	Wang et al. 2006
Carpinus cordata	agb = 1000 × 0.09802 × (dbh)2.2993	Chen and Zhu 1989
Fraxinus chinensis subsp. rhynchophylla	agb = 10(2.213+2.417×log10(dbh))	Wang et al. 2006
Fraxinus mandshurica	agb = 10(2.216+2.408×log10(dbh))	Wang et al. 2006
Juglans mandshurica	agb = 10(2.235+2.287×log10(dbh))	Wang et al. 2006
Larix gmelinii	agb = 10(1.997+2.451×log10(dbh))	Wang et al. 2006
Maackia amurensis	agb = 1000 × 0.0737 × (dbh)2.51264	Chen and Zhu 1989
Phellodendron amurense	agb = 10(1.942+2.232×log10(dbh))	Wang et al. 2006
Picea jezoensis var. komarovii	agb = 1000 × 0.0744 × (dbh)2.5411	Chen and Zhu 1989
Picea koraiensis	agb = 1000 × 0.0744 × (dbh)2.5411	Chen and Zhu 1989
Pinus koraiensis	agb = 10 (2.236+2.144×log10(dbh))	Wang et al. 2006
Populus davidiana	agb = 10(1.826+2.558×log10(dbh))	Wang et al. 2006
Quercus mongolica	agb = 10(2.002+2.456×log10(dbh))	Wang et al. 2006
Taxus cuspidata	agb = 1000 × 0.0737×(dbh)2.51264	Chen and Zhu 1989
Tilia amurensis	agb = 10(1.606+2.668×log10(dbh))	Wang et al. 2006
Tilia mandshurica	agb = 10(1.606+2.668×log10(dbh))	Wang et al.2006
Ulmus davidiana	agb = 1000 × 0.09802 × (dbh)2.2993	Chen and Zhu 1989
Ulmus laciniata	agb = 1000 × 0.09802 × (dbh)2.2993	Chen and Zhu 1989
Ulmus macrocarpa	agb = 1000 × 0.09802 × (dbh)2.2993	Chen and Zhu 1989
Other species	agb = 10(1.826+2.558×log10(dbh))	Chen and Zhu 1989

All allometric equations of the dominant species in our study were obtained from published references. For the missing species, we used the value of the same genera from (1) Chen, C.; Zhu, J. Biomass Manual of Main Trees in Northeastern China. China Forestry Press, Beijing, China, 1989. and (2) Wang, C. Biomass allometric equations for 10 co-occurring tree species in Chinese temperate forests. For. Ecol. Manag. 2006, 222, 9–16.

) to calculate the aboveground biomass of each tree. Then, we multiplied the carbon content of the main tree species [38] in the Changbai Mountain area (

Table A2. Carbon content for each tree species in Changbai Mountain forest area.

Species	Carbon Content
Acer pictum subsp. mono	0.4456
Populus davidiana	0.4877
Tilia amurensis	0.4944
Ulmus davidiana	0.3859
Betula costata	0.4976
Betula platyphylla	0.4508
Quercus mongolica	0.4840
Fraxinus mandshurica	0.4537
Maackia amurensis	0.4553
Pinus koraiensis	0.5339
Larix gmelinii	0.5079
Other coniferous tree species	0.5166
Other broad-leaved tree species	0.4609

) to obtain the aboveground carbon of each tree, summed the AGC storage of all individual trees in the plots, and converted the values to Mg/ha.

We selected some biotic and abiotic variables to describe their impact on AGC storage, including topography, stand structure attributes, and species diversity. As topography attributes, elevation and slope were used as predictors to test the impact on AGC storage.

In order to characterize the diversity attributes of plants, we used tree species richness to express the species diversity, i.e., the number of species contained in the sample plots.

The stand structure attribute adopted two indices: individual tree size variation and stand density. Individual tree size variation was quantified using the coefficient of variation of DBH, which can be calculated as the ratio of the standard deviation of all DBH to average DBH in each plot [39] (Equation (1)). Stand density was expressed as the number of trees per hectare.

$$C{V}_{DBH}=\frac{S_{DBH}}{{\overline{X}}_{DBH}}\times 100$$

(1)

where $C{V}_{DBH}$ represents the coefficient of variation of DBH in the sample plot, $S_{DBH}$ is the standard deviation of all DBH measurements in the sample plot, and ${\overline{X}}_{DBH}$ is the average DBH in each plot.

2.3. Statistical Analysis

Here, we regarded forest type as a random effect. In order to find out whether fixed factors (elevation, slope, stand density, species richness, and tree DBH variation) or the random factor (forest type) can explain the change in aboveground carbon storage, we further divided the collected sample plot data into four forest types.

At first, we used the volume table [40] (

Table A3. Volume equations for each tree species. Volume: $\mathrm{V}={\mathrm{a}\times \mathrm{D}}^{\mathrm{b}}{\times \mathrm{H}}^{\mathrm{C}}$; height: $\mathrm{H}={\mathrm{h}}_1-\frac{{\mathrm{h}}_2}{\mathrm{D}+{\mathrm{h}}_3}$.

Species	Parameters
	a	b	c	h1	h2	h3
Larix olgensis	8.47 × 10−5	1.97	0.75	34.59	650.53	18.0
Pinus koraiensis	7.62 × 10−5	1.90	0.86	21.84	309.16	14.0
Abies nephrolepis	5.79 × 10−5	1.89	0.99	46.40	2137.92	47.0
Picea jezoensis var. komarovii	5.79 × 10−5	1.89	0.99	46.40	2137.92	47.0
Acer pictum subsp. mono	4.88 × 10−5	1.84	1.05	24.82	402.09	16.3
Fraxinus mandshurica	5.33 × 10−5	1.88	1.00	29.44	468.93	15.7
Phellodendron amurense	5.33 × 10−5	1.88	1.00	29.44	468.93	15.7
Betula platyphylla	5.33 × 10−5	1.88	1.00	29.44	468.93	15.7
Tilia amurensis	5.33 × 10−5	1.88	1.00	29.44	468.93	15.7
Betula costata	5.33 × 10−5	1.88	1.00	29.44	468.93	15.7
Populus ussuriensis	5.33 × 10−5	1.88	1.00	29.44	468.93	15.7
Ulmus pumila	5.33 × 10−5	1.88	1.00	29.44	468.93	15.7
Ulmus davidiana	5.33 × 10−5	1.88	1.00	29.44	468.93	15.7
Taxus cuspidata	7.62 × 10−5	1.9	0.86	21.84	309.16	14
Abies fabri	7.62 × 10−5	1.9	0.86	21.84	309.16	14
Quercus mongolica	5.33 × 10−5	1.88	1	29.44	468.93	15.7
Populus davidiana	5.33 × 10−5	1.88	1	29.44	468.93	15.7
Other coniferous tree species	7.62 × 10−5	1.9	0.86	21.84	309.16	14
Other broad-leaved tree species	5.33 × 10−5	1.88	1	29.44	468.93	15.7

) to calculate the volume of trees and then divided the plots into four different types according to the conifer–broadleaf ratio: coniferous forest, coniferous mixed forest, broad-leaved mixed forest, and coniferous and broad-leaved mixed forest. The data of the forest type classification standard are listed in

Table 1. Forest type classification standard.

Forest Type	Classification Standard	Number
Coniferous forest	Single coniferous tree species ≥ 65% of total volume	11
Coniferous mixed forest	Coniferous species ≥ 65% of total volume	125
Broad-leaved mixed forest	Broad-leaved species ≥ 65% of total volume	24
Coniferous and broad-leaved mixed forest	Broad-leaved or coniferous species account for 35–65%	46

.

Multicollinearity affects the ability of explanatory variables to explain and predict response variables. The variance inflation factor (VIF) was used to test the multicollinearity between variables. The results show that the VIFs of the predictors in this study were less than 10, so there was no multicollinearity [41,42].

Piecewise structural equation model (pSEM) was used in this study for model simulation. In addition, in order to reflect the value of pSEM, we used boxplots to evaluate the differences of fixed factors (elevation, slope, stand density, species richness, and tree DBH variation) and AGC storage in different types of forests before pSEM analysis. pSEM [43] extends the traditional SEM, which allowed us to consider the effects of mixed (random and fixed) factors on response variables. For each response variable, pSEM decomposes the complex relationship into corresponding simple or multiple linear mixed-effect regression. Each regression can be evaluated separately and then combined to produce inferences about the entire SEM.

In particular, we tested the following hypotheses (Figure 2): (1) Species richness has a positive direct impact on AGC storage and indirect impact (through tree DBH variation); (2) Tree DBH variation has a direct impact on AGC storage; (3) Stand density has a direct and indirect impact on AGC storage (through its impact on species richness and tree DBH variation); (4) Elevation has a direct and indirect impact on AGC storage (through its impact on species richness); (5) The slope has a direct impact on AGC storage.

We evaluated the overall performance of pSEM by Fisher’s C statistics when the model has Fisher’s C statistics and p > 0.05. pSEM is considered to have a sufficient fit with the data. If the fitting degree did not meet the standard after modeling with pSEM, the least p value in the significant missing path (p < 0.05) could be further incorporated into the model through directional separation, and this process would be repeated until the model fitting degree met the standard, but the supplementary path had to have a theoretical basis. When multiple models passed the Fisher’s C test, the AIC values were compared to select the optimal model. For each dependent variable in pSEM, we calculated the conditional R² ($R_c^2$) and the marginal R² ($R_m^2$) to calculate the variance explained by fixed ($R_m^2$) and random factors ($R_c^2-R_m^2$). R v 4.0.2 software was used for all analyses. pSEM was employed by using the “piecewiseSEM” package.

3. Results

3.1. General Description

There are 206 plots with 6786 individuals, belonging to 11 families, 14 genera, and 15 species. The average DBH of the sample plot was 19.30 cm. The average AGC storage of forest was 72.44 ± 25.75 Mg/ha. Aboveground carbon storage ranged from 10.85 Mg/ha to 158.21 Mg/ha. In different types of forests, there were differences in the diameter class distribution of trees (Figure 3). The basic information of the sample plots is described in

Table 2. Summary of the variables across forest types.

Forest Types	Variable	Mean	SD	Maximum	Minimum
Coniferous forest (n = 11)	AGC (Mg/ha)	78.65	22.75	103.22	28.39
	Elevation(m)	681.15	34.75	722.49	635.47
	Slope (°)	9.08	3.85	15.23	4.66
	Stand density (n/ha)	618.18	186.44	800	200
	Species richness	5.36	1.69	9.00	4.00
	CVDBH (%)	37.87	6.50	45.14	25.47
Coniferous mixed forest (n = 125)	AGC(Mg/ha)	77.95	24.88	158.21	14.63
	Elevation (m)	665.74	39.67	757.10	603.17
	Slope (°)	9.56	3.50	21.00	2.03
	Stand density (n/ha)	671.64	184.32	1125.00	200.00
	Species richness	6.81	1.48	10.00	3.00
	CVDBH (%)	53.78	12.63	94.84	29.82
Broad-leaved mixed forest (n = 24)	AGC (Mg/ha)	62.83	27.67	132.85	17.90
	Elevation (m)	709.13	52.14	778.55	597.79
	Slope (°)	10.22	4.38	20.79	3.60
	Stand density (n/ha)	875.23	390.68	1650.00	200.00
	Species richness	7.63	2.08	12.00	3.00
	CVDBH (%)	57.69	9.85	74.59	33.84
Coniferous and broad-leaved mixed forest (n = 46)	AGC (Mg/ha)	61.11	23.42	123.64	10.85
	Elevation (m)	693.81	44.99	783.09	600.24
	Slope (°)	9.52	3.84	16.70	2.01
	Stand density (n/ha)	713.47	282.14	1975.0	200.0
	Species richness	7.43	1.57	11.00	5.00
	CVDBH (%)	52.44	11.71	86.37	30.76

DBH represents the diameter of breast height. ${\mathrm{CV}}_{\mathrm{DBH}}$ represents the coefficient of variation of DBH in the sample plot. AGC represents the aboveground carbon. SD represents the standard deviation.

. The boxplots showing the differences amongst the variables across four types of forests are provided in Figure 4.

3.2. Effects of Predictors on AGC Storage

The bivariate relationships between AGC and significant variables are described in Figure 5. AGC storage increased with increasing species richness and stand density. However, AGC storage significantly decreased with the increase in elevation and slope. Figure 6 shows a pSEM with the random factor (forest type). Solid and dotted lines indicate significant and non-significant paths. The numbers of $R_m^2 \mathrm{and} R_c^2$ close to the variables represent the variance explained by the fixed factors and all factors in the model, respectively. We found that pSEM explained the marginal and random changes of 21% and 12% of aboveground carbon storage, respectively. The difference in AGC growth was partially dependent on forest types. AGC storage increased with the increase in stand density ($\beta$= 0.21) and species richness ($\beta$= 0.16). AGC storage decreased with the increase in elevation ($\beta$=$-$0.17), slope ($\beta$=$-$0.28), and DBH variation ($\beta$=$-$0.15). Stand density promoted the increase in species richness and inhibited the variation in DBH. Stand density and the variation in DBH mainly depended on the random effect (i.e., forest type), whereas species richness mainly depended on fixed effects.

In terms of the relative contribution of predictors, we found that stand density, CV_DBH, and species richness accounted for more than 55% of total variations for AGC. The direct effects of elevation, density, and species richness were larger than the indirect effects. Slope and CV_DBH only had direct effects on AGC storage (Figure 7).

4. Discussion

This study shows that the AGC storage in different forest types is mainly caused by fixed factors (topography, forest structure attributes, species richness), with topography as the main reason explaining the AGC change.

Topography had a direct effect on AGC storage. Specifically, differences in topographic factors may result in microenvironment heterogeneity of resource availability (light, water, and soil nutrients) [24]. Higher elevation gradients receive lower temperature and soil moisture content [44]. Therefore, elevation had a negative impact on AGC storage, resulting in lower AGC storage. The slope had a negative impact on AGC storage, which is because slope plays an important role in determining soil properties, having a negative impact on forest growth by changing soil moisture change and soil erosion [45]. In addition, topography can indirectly affect AGC storage through stand density, which is consistent with previous studies [46]. We also assumed that elevation indirectly affected AGC storage by affecting species diversity and individual tree size variation. However, the results show that the relationship between elevation and diversity was not significant, and the relationship between elevation and diversity could not be described as a linear relationship but a complicated, nonlinear relationship caused by multiple factors.

The results of pSEM in this study show that species diversity had a strong, positive impact on AGC storage, which is consistent with previous studies [9,47,48,49]. This can be explained by the niche complementarity hypothesis [50] (i.e., a higher diversity allows a community to access a larger fraction of the total resource pool). Species diversity had a direct impact on AGC storage, and species diversity can also indirectly increase AGC storage through multiple factors such as stem density [19] and individual tree size variation [51]. We assumed that individual tree size variation positively affected AGC storage. However, contrary to our expectations, the results show that individual tree size variation had a strong negative effect on AGC storage. This may be due to the fact that the degree to which individual trees are affected by competition depends on their own and adjacent individual size, quantity, characteristics of utilized resources, and spatial influence area [52]. Changing the individual tree size variation will have a certain impact on competition and will then influence biomass accumulation. With the asymmetric competition mechanism, larger trees monopolize light resources more than proportionally to their size; therefore, the intraspecific and interspecific competition for the limited resources in forests may influence individual tree size variation—large-sized trees may eliminate small-sized trees, which consequently decreases AGB [53].

In this study, stand density promoted AGC storage. High stand density can optimize resources utilization through canopy packing, finally leading to higher AGB [54]. We also found a positive correlation between stand density and species richness, which supports the species energy hypothesis, indicating that higher energy availability supports higher species richness, resulting in higher AGB [5,6,55]. What supports this finding is that the negative effect of elevation on stand density is also derived from the species energy hypothesis.

As for the random effect (forest type), AGC storage varied among different forest types. The study forests are uneven-aged forest stands, and it has been reported that faster-growing forest stands may be dominated by species with shorter lifespans [56] and consequently lower AGC. On the other hand, the differences in productivity of different forest types may reflect the feedback effect of plant species composition [57]; some studies found that even minor changes in species composition can result in variation in forest biomass and carbon density among different forest types [58]. In addition, natural disturbances such as storms [59] and insect outbreaks [60] might diminish AGC storage; therefore, an increase in AGC storage can be achieved through sustainable forest management.

Moreover, it is important to note that we did not measure wood density directly, which is one of the variables modeling stand-level carbon storage [61]. Future studies should further investigate the performance of the above factors on AGC storage spanning different forest types.

5. Conclusions

In this study, the relationship between various variables and aboveground carbon storage in different types of the forest was described by using pSEM. We found that AGC is directly affected by the changes in topography, species diversity, and forest structure attributes. Among them, topography was the most important driving factor affecting AGC storage, and stand density and species richness were positively related to AGC storage, while CV_DBH was negatively related to AGC storage. Forest AGC storage varied among different forest types, which is in agreement with previous studies. In addition, consistent with the species energy hypothesis, we also found a positive correlation between stand density and species richness, indicating that higher energy availability supports stands with higher species richness. We should give more consideration to improving biodiversity, contributing to species renewal, and forest management in the future.