Temporal and Spatial Characteristics and Influencing Factors of Carbon Storage in Black Soil Area Under Topographic Gradient

Abstract

Exploring the characteristics and driving factors of carbon storage change in different terrain gradient variations can provide important insights for formulating the agricultural ecological protection policy for regional development. Previous studies have used the fixed value of carbon density to evaluate the change characteristics of carbon storage but ignored the spatio-temporal heterogeneity of carbon storage at the block scale and the impact of policy factors. Thus, this paper takes Sanjiang Plain, Heilongjiang Province, China, as a study area, and the spatio-temporal variation of carbon storage at different topographic gradients was revealed using hot and cold spot analysis and zonal statistics. Through the geographic detector and estimation of the soil carbon density model, the driving factors and intensity of carbon storage spatial distribution are revealed from 1990 to 2020. We conducted analyses on aboveground biomass, underground biomass, and soil carbon storage across three elevation levels (0–200 m, 200–500 m, 500–999 m) to reveal the quantitative distribution features of carbon storage. The study analysis finds that carbon storage indicates a sawtooth evolution during the study period. Carbon storage was dominant at elevation I (range is 0–200 m), slope I (range is 0–2°), and relief amplitude I (range is 0–30 m). Additionally, the carbon storage losses were severe at elevation II (range is 200–500 m), slope II (2–6°), and relief amplitude II (30–70 m). In contrast, the carbon storage losses at elevation III (500–999 m), slope III (6–15°), and relief amplitude III (70–186 m) were insignificant. The spatial pattern of carbon storage varies significantly under different topographic gradients from 1990 to 2020. The most critical driving factors influencing the spatial distribution pattern of carbon storage were land use and annual average temperature. Distance to urban centers and soil texture also moderately influence the distribution of carbon storage. As the topographic gradient increases, the dominant factors of carbon storage gradually change from annual mean temperature and the extent of land use to policy factors and other socio-economic factors. Therefore, this study emphasizes the importance of implementing policies that convert farmland to forests and wetlands and promote the green transformation of agriculture.

1. Introduction

The excessive emissions of greenhouse gases such as CO₂ and CH₄ from human activities have led to global climate change [1]. The Sixth Report of the Intergovernmental Panel on Climate Change highlighted a rise of 1.1 °C in global surface temperatures between the periods 1850–1900 and 2000–2020 [2]. Therefore, climate change has become an important challenge facing humanity [3], and land ecosystems significantly contribute to climate change mitigation by facilitating the exchange of greenhouse gases and influencing carbon storage within ecosystems. Land use change, as an important manifestation of human activities, directly changes the aboveground biomass, underground biomass, and soil organic matter sources of land ecosystems, thereby affecting ecosystem carbon storage [4].

Human activities, such as rapid socio-economic development, have highlighted the issues related to wetland degradation, soil erosion, and land salinization [5,6,7], which resulted in a significant loss of carbon storage. Moreover, numerous researchers have shed light on the impact of terrain and altitude on land use patterns and local climate patterns [8,9,10,11]. These factors directly affect aboveground and underground biomass, soil temperature and humidity, and soil organic matter sources through changes in temperature, rainfall, and human activity [12]. These changes affect various features of the land, including soil properties, ecological enzyme activity, organic matter decomposition rate, vegetation phenological changes, and stand type [13,14,15,16,17,18]. This process results in variations in carbon storage across different terrain gradients. Our study lends evidence to the support of knowledge for achieving Chinese dual carbon goals and optimizing land use patterns, which provides significant insight from theoretical and practical contributions.

Enormous research has been conducted on carbon storage at the macro, meso, and micro scales [1,3,8,9,10,13,15,19,20,21,22,23,24]. In essence, these studies explored the carbon storage issues on national and local scales such as county or farmland plot levels. Along with this, the main methods used in the research field for revealing the spatiotemporal variation features of carbon storage encompass the biomass conversion factor method, remote sensing prediction method, Multivariate Linear Regression (MLR) and Generalized Additive Model (GAM), Integrated Valuation of Ecosystem Services and Trade-offs (InVEST) model, and so on [25,26]. For example, Ippolito et al. [26] applied the field measurement method to extract 1600 samples from 103 wetlands in Victoria, southeastern Australia, in order to calculate the carbon storage of different wetlands. Furthermore, Sharma et al. [14] used the biomass conversion factor method to estimate the biomass and soil carbon storage of eight agroforestry systems in the northern part of the Himalayas. Along the same lines, Ding et al. [20] used the Denitrification–Decomposition (DNDC) model to reveal the spatiotemporal differentiation characteristics of soil organic carbon storage in Chinese farmland from 2003 to 2020. Last but not least, Wu et al. [3] analyzed the spatiotemporal variation characteristics of carbon storage in the Sanjiangyuan region, from 1990 to 2020, based on the InVEST model and combined it with the Patch-generating Land Use Simulation (PLUS) model in order to predict carbon storage in 2035.

The InVEST model, which is based on the constant carbon density of aboveground biomass, belowground biomass, litter and soil, and land use types for calculation, has shown interesting operability. It is common to combine it with land use prediction models, such as Markov, PLUS, and Future Land Use Simulation (FLUS), to predict future spatiotemporal changes in carbon storage [1,3,27,28]. As a result, these studies are limited to capturing the spatial heterogeneity of carbon storage. Essentially, while traditional methods demonstrate high accuracy, studies often supplement them with other techniques, resulting in intricate operations. In contrast, the biomass conversion factor method is suitable for calculating biochar storage but has low accuracy [3]. The assumption that the biomass conversion value is a constant is inaccurate as it is, in fact, dependent on a number of factors, including forest age, tree species composition, woodland conditions, and stand density. Furthermore, the average value of biomass conversion provided by forest inventory data and forest total area and stock data is unable to accurately reflect long-term series and large-scale spatial carbon storage changes. The model method often faces various issues [26]. The accuracy of the input data of the model method affects the prediction results, which requires a large amount of sample data. In addition, the model structure and computational complexity also determine the accuracy of the prediction results. The results of the same experiment will be different when using different models.

Previous studies have used various models to explore the factors affecting carbon storage. For instance, Wang et al. [29] used linear regression analysis to find that altitude is an important factor affecting carbon storage and carbon density. Through an enhanced regression tree model, Andreetta et al. [30] explored the effects of different environmental factors on carbon storage in forest surface soil (0–20 cm) and the entire soil profile (0–80 cm). It was found that climate directly controls soil organic carbon (SOC) storage. At the same time, geographic detectors were built to analyze the impact of different driving factors and their interactions on carbon storage. The result shows that average annual rainfall was the main control factor in carbon storage [31]. The combination of the Random Forest (RF) model with Pearson correlation analysis indicates that annual average temperature is one of the main factors affecting the spatiotemporal differentiation of soil organic carbon in Northeast China, especially in areas with an average yearly temperature below 8 °C [32]. In the Qinghai Tibet Plateau region, the use of the Random Forest Regressor (RFR) model to explore the driving factors of soil organic carbon spatial distribution found that altitude was related to soil organic carbon content among selected factors [33]. In a nutshell, on a macro scale, climate factors mainly influence carbon storage, and land use change is also an important factor in regional carbon storage variation [11]. Although the scale is important, factors such as differences in terrain variations, altitude, soil properties, climate, etc., determine the spatial distribution pattern of carbon storage [8,9,10,14].

Accordingly, many previous studies explored the spatiotemporal variation of carbon storage, focusing on carbon density as a fixed value and using the InVEST model. A few studies failed to analyze the spatiotemporal carbon density from terrain gradient heterogeneity at the micro-scale. In addition, there are few studies analyzing the factors that influence carbon storage on terrain gradients, along with the impact of the science and technology industry, energy conservation, and agricultural supply policies on carbon storage variation in black soil areas. In this region, the carbon storage status has shown a downward trend. However, a combination of factors, including climate change and human activities, have changed the physical and chemical properties of black soil, disrupting soil carbon balance and loss. Therefore, taking Sanjiang Plain, China’s black soil area, as an example, and using a mixed dataset, this paper calculated aboveground biomass, belowground biomass, and soil carbon storage from 1990 to 2020, explored the spatiotemporal variation of carbon storage for different elevations, slopes, and relief amplitudes. The influence factors and intensity of spatial and temporal differentiation of carbon storage were discussed by means of a geographical detector. Consequently, this study contributes to alleviating carbon storage losses, enhancing carbon storage measuring methods, and improving carbon sequestration policies in China’s black soil areas.

2. Materials and Methods

2.1. Study Area

The Sanjiang Plain is located in the northeast of Heilongjiang Province, China, between 45°01′00″~48°27′56″ N and 130°13′00″~135°05′26″ E (Figure 1). In essence, it is located in the black soil area, which is a ballast stone for Chinese grain production, ensuring Chinese food security. The black soil refers to the land marked by black or dark humus topsoil (0–30 cm), which is the most precious soil resource on Earth and one of the most fertile soils because its abundant humus provides a source of trace elements such as carbon, nitrogen, and phosphorus for crop growth. The administrative divisions encompass prefecture-level cities such as Jiamusi, Hegang, Shuangyashan, Qitaihe, and Jixi, along with Yilan County, which falls under Harbin City. Over the last three decades, under a continental monsoon climate, the average temperature has ranged between 2.5 and 4.5 °C, with rainfall ranging from 500 to 800 mm annually. Compared to the western, eastern, and southern areas, the northern area has flat terrain and is a plain formed by the Songhua, Heilongjiang, and Wusuli Rivers. In 2020, The population of the study area was 8.625 million, and the land area is 10,240 km². The land area at elevation I (0–200 m] accounts for 73.187% of the total area, elevation II (200–500 m] is 24.473%, and elevation III (500–999 m] is 2.340%. The land area for Slope I (0–2°] accounts for 69.319% of the total area, Slope II (2–6°] is 19.619%, Slope III (6–15°] is 10.754%, Slope IV (15–25°] is 0.307%, and Slope V (25–29°] is 0.001%. The land area for Relief amplitude I (0–30 m] accounts for 97.915% of the total area, Relief amplitude II (30–70 m] is 1.542%, and Relief amplitude III (70–186 m] is 0.543%.

2.2. Data Sources and Processing

This study collected data from various sources (

Table 1. Data types, sources, and treatment methods.

Data Types	Data Sources		Treatment Methods
	Name	Link
Aboveground biomass and underground biomass carbon density	National Laboratory of the National Aeronautics and Space Administration (NASA)	https://daac.ornl.gov/VEGETATION/guides/Global_Maps_C_Density_2010.html (accessed on 20 May 2024)	Extract By Mask in ArcGIS 10.7 was used to extract data from Sanjiang Plain
Average annual temperature and annual precipitation	Resource and Environmental Science Data Platform (China)	https://www.resdc.cn/DOI/doiList.aspx (accessed on 10 June 2024)
Soil texture, soil type, soil organic carbon content, and soil bulk density	National Glacier Frozen Desert Science Data Center	http://www.ncdc.ac.cn (accessed on 25 May 2024)
Elevation	Geospatial data cloud platform	http://www.gscloud.cn/ (accessed on 25 May 2024)
Land use type, vegetation coverage (NDVI)	Resource and Environmental Science Data Platform (China)	https://www.resdc.cn/ (accessed on 25 May 2024)
Distance to water area	Resource and Environmental Science Data Platform (China)	https://www.resdc.cn/ (accessed on 2 June 2024)	It is obtained by using Euclidean Distance in ArcGIS 10.7
Distance to roads	Global biodiversity model for policy support (GLOBIO)	https://www.globio.info/ (accessed on 12 June 2024)	It is obtained by using Euclidean Distance in ArcGIS 10.7
Urban center	National Center for Basic Geographic Information (NGCC)	http://www.ngcc.cn/ (accessed on 12 June 2024)	Transform into point elements using ArcGIS 10.7
GDP, population	Statistical Yearbook of Heilongjiang Province	http://tjj.hlj.gov.cn/ (accessed on 15 June 2024)	Inverse Distance Weighted(IDW) in ArcGIS 10.7 is used for spatial interpolation
Technology industry, emission reduction and energy conservation, and agricultural supply policy	Statistical Yearbook of Heilongjiang Province	http://tjj.hlj.gov.cn/ (accessed on 15 June 2024)	Assign 1 to counties with policy and 0 to those without. IDW in ArcGIS 10.7 is used for spatial interpolation

). This study selected 13 data indicators, including natural, socio-economic, and policy factors, to evaluate the driving factors of carbon storage in different terrain gradient variations. Firstly, natural factors include annual average temperature (X₁) and rainfall (X₂). Moreover, the soil texture (X₃), vegetation coverage (X₄), and distance from the water area (X₅) are added to the natural data for an understanding of the carbon variation features. In addition to natural data, the socio-economic factors include the extent of land use (X₆), average land GDP (X₇), population density (X₈), distance to roads (X₉), and distance to urban centers (X₁₀). Lastly, the policy factors include government technology industry policy (X₁₁), energy conservation and emission reduction policy (X₁₂), and agricultural supply-side structural reform policy (X₁₃). The extent of land use is obtained by its measurement model.

Considering the varying importance of elevation, slope, and relief amplitude, three topographic factors were selected. The data processing has gone through several steps. The slope data are extracted from the elevation data by using the spatial analysis tool in ArcGIS 10.7. In this context, the maximum and minimum values of the elevation data are obtained using the domain analysis tool in ArcGIS 10.7. Similarly, the relief amplitude is obtained using the grid calculator. Relief amplitude is the maximum relative elevation difference per unit area. It represents a quantitative index used to describe the geomorphic form. According to the terrain and land use characteristics of the study area and related references, the elevation, slope, and relief amplitude were classified. The elevation was divided into elevation I (0–200 m], elevation II (200–500 m] and elevation III (500–999 m]. The slope was divided into Slope I (0–2°], Slope II (2–6°], Slope III (6–15°], Slope IV (15–25°] and Slope V (25–29°]. The relief amplitude was divided into Relief amplitude I (0–30 m], Relief amplitude II (30–70 m] and Relief amplitude III (70–186 m]. The higher the levels, the greater the elevation, slope, and relief amplitude. Prior to the collection process, the data, namely land use type data, vegetation coverage (NDVI), distance to roads, GDP, population, technology industry policies, energy conservation and emission reduction policies, and agricultural supply, were reprojected using Alberts in ArcGIS 10.7 at a resolution of 30 m. Through the same software, the fishing net tool was built to divide the research area. After repeated debugging, a 1000 m × 1000 m square grid was used as the research unit, which was combined with the study area, research purpose, and computational efficiency. The required research data were transferred to the corresponding grid unit through multi-value extraction to points.

2.3. Research Methods

2.3.1. Soil Carbon Density Estimation

Estimating soil carbon density entails evaluating the quantity of carbon contained within a specified volume or area of soil. In the context of this study, the surface soil carbon density is estimated between 1990 and 2020 using the following formula [34]:

SOCD = 10TD ρ

(1)

where SOCD—the organic carbon density of surface soil (kg/m²), T—organic carbon content of surface soil (%), D—the soil thickness (0.3 m), ρ—soil bulk density (g/cm³), and 10–a Unit conversion factor.

The ArcGIS 10.7 grid calculator was used to input soil bulk density and organic carbon content into Formula (1), calculating the surface soil organic carbon density for the study area.

2.3.2. Carbon Storage Estimation

Carbon storage estimation involves determining the quantity of carbon held within a specific ecosystem or environmental reservoir. Referring to the work of Wang et al., and Jiang et al. [5,28], this study divides carbon storage into four basic carbon pools, namely aboveground, underground, soil, and litter carbon pools, as can be seen in Equation (2):

$$C_{total}=C_{above}+C_{below}+C_{soil}+C_{dead}$$

(2)

where C_total—total carbon storage of ecosystem, C_above—aboveground carbon density of land-use type, C_below—underground carbon density of land-use type, C_soil—soil organic carbon density of land-use type, and C_dead—dead organic carbon density of land-use type, C_dead = 0.

Based on the result of the above equations, the annual average rainfall and temperature are substituted into the carbon density modified formula according to the study period [35]. This correction allows us to match the carbon density data for the corresponding year.

C_BP = 6.798e ^0.0054M_P

(3)

C_BT = 28M_T + 398

(4)

C_SP = 3.3968M_P + 3996.1

(5)

where M_P—the annual rainfall (mm), M_T—the annual average temperature (°C), C_BP—the aboveground biomass carbon density corrected based on annual rainfall (t/hm²), C_B—aboveground biomass carbon density corrected based on the annual average temperature (t/hm²), and C_SP—the soil carbon density (t/hm²) corrected based on annual rainfall.

Using the ArcGIS 10.7 grid calculator, we corrected the extracted aboveground biomass carbon density and soil carbon density data. We combined the aboveground, underground, and soil carbon densities to calculate the total carbon storage in the study area.

2.3.3. Hot and Cold Spots Analysis

Hot and cold spots were analyzed using the Getis-Ord General G tool in ArcGIS 10.7 to identify the spatial clustering of carbon storage changes across different topographic gradients. The observed G value describes the distribution of hot and cold spots. A statistically significant positive value indicates a higher concentration of hot spot clusters, designating the area as a hot spot region. In contrast, a negative value identifies it as a cold spot region. The distribution of hot and cold spots is classified into six levels based on the confidence interval, with hot and cold spots categorized into 99%, 95%, and 90% confidence intervals. The calculation formula is as follows:

$$G={\displaystyle \frac{{\sum }_{i=1}^n{\sum }_{j=1}^n{W}_{ij}{M}_i{M}_j}{{\sum }_{i=1}^n{\sum }_{j=1}^n{M}_i{M}_j}}\left(i\ne j\right)$$

(6)

where $M_j$—the change in carbon storage of unit j, $W_{ij}$—the spatial weight matrix of elements i and j (a value of 1 means adjacent Spaces, and a value of 0 means non-adjacent Spaces).

2.3.4. Geographic Detector Model

The Geographic Detector Model is a spatial analysis tool used to understand the relationships between geographic factors and processes. In this study, the driving factors of carbon storage variations and their impact are estimated using Equation (7) [36]:

$$q=1-{\displaystyle \frac{{\sum }_{h=1}^L{N}_h{{\sigma }^2}_h}{N{\sigma }^2}}$$

(7)

where q is the explanatory power of the formation mechanism, the range is [0, 1]; L—the number of partitions; N_h—the number of units corresponding to the h layer carbon storage and factor; N—the number of units corresponding to carbon storage and factors in the whole region; σ²_h—variation variance of carbon emissions in layer h; and σ² is the variation variance of carbon storage in the whole region.

Interaction detection is mainly used to judge whether each factor has an independent influence on the dependent variable or has an influence after the interaction, and whether the influence force is weakened or strengthened. The relationship between the two factors can be divided into the following categories:

(1)

q(X₁∩X₂) < Min(q(X₁), q(X₂)), non-linear attenuation;

(2)

Min(q(X₁), q(X₂))q(X₁∩X₂) < Max(q(X₁), q(X₂)), single factor nonlinear weakening;

(3)

q(X₁∩X₂) > Max(q(X₁), q(X₂)), single factor nonlinear weakening;

(4)

q(X₁∩X₂) = q(X₁) + q(X₂), independent;

(5)

q(X₁∩X₂) > q (X₁) + q (X₂), non-linear enhancement.

3. Results

3.1. Terrain Gradient Characteristics of Carbon Storage

The spatial pattern of carbon storage in the study area from 1990 to 2020 reveals distinct trends (Figure 2). High carbon storage values are primarily concentrated in the northwest and eastern regions, which are characterized by higher elevations and mountainous terrain. Low-carbon storage areas are found in the northern and southeastern regions, where the elevation is lower, the terrain is mostly flat, and the land use is predominantly agricultural. Overall, carbon storage changes were more pronounced in the eastern, northwestern, and south-central areas compared to the lower-elevation regions.

Using ArcGIS 10.7’s zonal statistics tool, the study found that carbon storage progressively increased across topographic gradients from 1990 to 2020, with higher carbon storage corresponding to higher elevation gradients (Figure 3). Among the various topographic gradients, the largest increase in carbon storage occurred in 2000 compared to 2020. The carbon storage increase at elevation II was the most significant, rising by 229% compared to elevation I. In contrast, the increase at elevation III was only 40% higher than at elevation II in 2000. In terms of slope, Slope II showed the largest increase in carbon storage, with a 230% rise compared to Slope I, while Slope V experienced a 22% decrease compared to Slope IV in 2000. Relief amplitude II saw a 321% increase in carbon storage compared to relief amplitude I, while relief amplitude III showed only a 19% increase compared to relief amplitude II. Overall, within the same topographic gradient, carbon storage showed a downward trend, with the most significant declines observed at elevation III and relief amplitude III, which decreased by 41.43% and 41.28%, respectively. Slope I also experienced the largest decrease in carbon storage, declining by 41.15% from 1990 to 2020.

Spatial autocorrelation analysis using ArcGIS 10.7 showed that Moran’s I > 0.75, Z-score > 1.65, and p-value < 0.001, indicating a highly clustered spatial distribution of carbon storage in the study area. Based on this, with the help of the Getis-Ord General G tool from the ArcGIS 10.7 software, the cold hot spot analysis of carbon storage under different topographic gradients in the study area from 1990 to 2020 was carried out to reveal the clustering spatial distribution pattern of carbon storage. The results show that the distribution of cold hot spot regions of carbon storage performs small changes under different topographic gradients from 1990 to 2020, in which the hot spot regions of carbon storage are distributed in a banded and scattered way, while the cold spot regions are distributed in a scattered, banded and blocky way. In addition, the spatial pattern of carbon storage varies significantly under different topographic gradients in the study area from 1990 to 2020, and the distribution of cold hot spots is significantly consistent with land use types. The hot spots are mainly distributed in forest land, while the cold spots are mainly distributed in other land types such as construction land, water area, and cultivated land.

The cold hot spots of carbon storage in the “elevation I region” of the study area did not change significantly from 1990 to 2020 (Figure 4(a1–a4)). The spatial distribution at each time point is as follows: Hot spots are mainly distributed in the form of bands and scattered spots, and their ranges do little change. They are distributed in the north of Jixi City in the form of bands, while in the northeast part of Jiamusi City, the northeast part of Shuangyashan City, the middle part of Hegang City, and the east part of Jixi City in the form of scattered spots, and the cold spot area of carbon storage is distributed in the southeast part of Jixi City in a massive way. The hot spots of carbon storage in the “elevation I region” in the study area are mainly distributed in the Nadan Hada Ridge, the main vein of Wandashan Mountain in the east, and the fringe area of Qingheishan, the branch vein of Xiaoxingan Mountain in the northwest. The land use is woodland and cultivated land, the vegetation coverage is relatively high, and the carbon storage is relatively rich. In contrast, the vast plain area in the north is mainly cultivated land and construction land, and the carbon storage distribution is weak.

The distribution of cold hot spots of carbon storage in the “elevation II region” in the study area is stable from 1990 to 2020 (Figure 4(b1–b4)). The spatial distribution at each time point is as follows: Hot spots are mainly concentrated in the eastern region and northern region, and scattered distribution in the southwest and north of Jiamusi City and the northwest of Hegang City. The spatial distribution pattern of the cold spot area is banded, scattered, and fragmentary, mainly distributed in the southern region. It can be seen that the hot spots of carbon storage in the “elevation II region” in the study area are mainly distributed in the sloping land of Nadan Hada Ridge, the main vein of Wandashan Mountain in the east. The land type is forest land, and the carbon storage is rich. The main reason is that with the rise of elevation, natural vegetation such as forest land and grassland gradually increases, and the intensity of human activities gradually weakens. The cold spot area is mainly distributed in the cross-zone of cultivated land and forest land on the slope of the branch of Wandashan Mountain, which is due to the intensification of soil and water loss and serious loss of carbon storage in the sloping land caused by forest land reclamation.

From 1990 to 2020, the hot spots of carbon storage in the “elevation III region” are scattered in the western study area, especially in the northwest and southwest, while the cold spots can be almost ignored. In general, the hot spots of carbon storage at elevation III are significantly lower than those at other elevations, and the hot spots decrease considerably with the increase in elevation.

3.1.1. Distribution Characteristics of Carbon Storage at Elevation

To examine the quantitative distribution features of carbon storage, analyses were performed on aboveground biomass (AGC), belowground biomass (BGC), and soil organic carbon (SOC) storage across three elevation levels. The variations in AGC, BGC, and SOC exhibited similar patterns at different elevations (see

Table 2. Carbon storage variation at different elevation levels (Unit: 10³ ton).

Year	Carbon Pool	Elevation
		I	II	III
1990	AGC	14,124.62	9865.41	11,591.74
	BGC	9956.84	6615.99	7648.61
	SOC	8784.20	1055.09	795.02
	Total	32,865.66	17,536.49	20,035.37
2000	AGC	18,294.66	12,777.99	15,013.99
	BGC	10,024.63	6603.79	7612.27
	SOC	9004.80	1081.59	814.98
	Total	37,324.09	20,463.37	23,441.24
2010	AGC	9571.57	6685.31	7855.16
	BGC	9882.81	6629.31	7688.30
	SOC	8446.70	1014.56	764.47
	Total	27,901.08	14,329.18	16,307.93
2020	AGC	4054.64	2831.98	3327.55
	BGC	9793.11	6645.45	7736.39
	SOC	7793.13	936.05	705.32
	Total	21,640.88	10,413.48	11,769.26

Note: AGC means Aboveground Carbon storage; BGC means Belowground Carbon storage; SOC means Soil organic Carbon storage.

). For example, between 1990 and 2000, AGC at elevation I was the dominant form of carbon storage, reaching 18,294.66 × 10³ t by 2000. However, from 2010 to 2020, BGC at elevation I became the largest contributor, peaking at 9793.11 × 10³ t in 2020. This indicates a shift in carbon storage dominance from AGC to BGC over time at elevation I. Similarly, for elevations II and III, AGC was the largest contributor to carbon storage from 1990 to 2010, while BGC took the largest share by 2020. In contrast, SOC consistently accounted for the smallest proportion throughout the period from 1990 to 2020, showing a fluctuating decline, especially at elevation III. Overall, as elevation increased, AGC, BGC, and SOC all displayed a general downward trend, with SOC experiencing the most significant decrease between 1990 and 2020.

3.1.2. Distribution Characteristics of Carbon Storage on Slopes

Between 1990 and 2020, the total carbon storage decreased when slope levels were important (

Table 3. Carbon storage variation at different slope levels (Unit: 10³ ton).

Year	Carbon Pool	Slope
		I	II	III	IV	V
1990	AGC	111,150.93	138,605.3	101,366.7	3405.73	5.9
	BGC	8007.62	9249.32	6664.66	193.56	0.18
	SOC	8331.16	1494.36	755.83	21.24	0.05
	Total	127,489.71	149,348.98	108,787.19	3620.53	6.13
2000	AGC	14,396.62	17,952.6	13,129.34	441.12	0.76
	BGC	8091.29	9224.1	6628.68	187	0.14
	SOC	8540.38	1531.89	774.81	21.77	0.06
	Total	31,028.29	28,708.59	20,532.83	649.89	0.96
2010	AGC	7532.16	9392.61	6869.13	230.79	0.4
	BGC	7916.26	9276.85	6703.95	200.71	0.22
	SOC	8011.07	1436.95	726.79	20.42	0.05
	Total	23,459.49	20,106.41	14,299.87	451.92	0.67
2020	AGC	3190.72	3978.83	2909.85	97.77	0.17
	BGC	7805.56	9310.21	6751.55	209.38	0.27
	SOC	7391.2	1325.76	670.55	18.84	0.05
	Total	18,387.48	14,614.8	10,331.95	325.99	0.49

). The aboveground and underground biomass carbon storage increased and then decreased along with the slope degree. In addition, the changes in AGC, BGC, and SOC showed high consistency at different slope levels (Table 3). Among them, AGC and SOC showed “increasing first, then decreasing”. This situation is the same as that of BGC, which showed a sawtooth evolution at slopes I, II, and V. To summarize, the carbon storage in each slope level area has shifted from AGC to BGC-dominated.

3.1.3. Distribution Characteristics of Carbon Storage on Relief Amplitude

The carbon storage of both aboveground and underground organisms demonstrated an “inverted N-shaped” pattern. The amount of carbon stored gradually decreased with the topographic level (

Table 4. Carbon storage variation at different relief amplitude levels (Unit: 10³ ton).

Year	Carbon Pool	Relief Amplitude
		I	II	III
1990	AGC	16,148.21	9669.32	9544.17
	BGC	11,347.09	6449.52	6247.58
	SOC	9048.19	851.49	689.41
	Total	36,543.49	16,970.33	16,481.16
2000	AGC	20,915.68	12,524.01	12,361.91
	BGC	11,993.09	6485.5	6252.66
	SOC	8700.55	818.78	662.93
	Total	41,609.32	19,828.29	19,277.5
2010	AGC	10,942.86	6552.43	6467.62
	BGC	11,269.42	6469.29	6289.86
	SOC	8700.55	818.78	662.93
	Total	30,912.83	13,840.5	13,420.41
2020	AGC	4635.54	2775.7	2739.77
	BGC	11,175.31	6493.25	6341.09
	SOC	8027.34	755.42	611.63
	Total	23,838.19	10,024.37	9692.49

). There were significant differences in the changes of AGC, BGC, and SOC in the study area. From 1990 to 2020, AGC, BGC, and SOC decreased with the increase in relief amplitude, and SOC decreased by 92%. Under the same relief amplitude scale, AGC first increased and then decreased from 1990 to 2020. Relief amplitude I BGC showed a fluctuating downward trend, relief amplitude II and III BGC showed a fluctuating upward trend, while SOC showed a downward trend. In addition, the relief amplitude I from 1990 to 2020 is dominated by AGC, and the relief amplitude I from 2010 to 2020 is dominated by BGC. From 1990 to 2010, AGC of relief amplitude II and III dominated, while BGC dominated in 2020. It can be seen that relief amplitude II and III changed from AGC to BGC. In general, compared with relief amplitude III, the decrease in AGC, BGC, and SOC was the largest from 1990 to 2020, indicating that with the increase in relief amplitude, the reserves of the three will decrease.

3.2. Estimate the Carbon Storage Loss at Different Terrain Gradient Variations

Based on the carbon storage loss from 1990 to 2020, the natural breakpoint method divided the loss into five levels: Lost I (−89.19 to −54.60 tons per hectare), Lost II (−54.60 to −41.64 tons per hectare), Lost III (−41.64 to −26.96 tons per hectare), Lost IV (−26.96 to −11.47 tons per hectare), and Lost V (−11.47 to 0 tons per hectare). Using ArcGIS 10.7 software, the five levels of carbon storage loss were overlaid with terrain factors, and statistical analysis was performed. The results indicated that carbon storage loss varies depending on topographic factors. Lost I showed the most significant loss among the different terrain factors, with its proportion exceeding 70% regardless of elevation, slope, or relief amplitude. Specifically, the largest loss of Lost I occurred at elevation I, where it accounted for 79.48%, while Lost II-V showed the most significant loss at elevation III, with more than 68%. Lost I had the highest proportion of loss in Grades I and II, making up 87.70% and 32.92%, respectively. In contrast, Lost IV had the highest proportion of loss in Grades III to V, reaching over 31%, with Grade V alone contributing to 45.08%. Lost I exhibited a relatively high loss in Relief Amplitude Grade I, accounting for 69.81%, while Lost III and IV showed higher losses in Relief Amplitude Grade II, contributing 24.13% and 29.15%, respectively. Furthermore, Lost II–V displayed a much higher loss in Relief Amplitude Grade III, accounting for more than 62%. In summary, the proportion of distribution area in the lost area decreased as the terrain gradient increased, exhibiting a sawtooth evolution trend, suggesting a strong dependence of the carbon storage loss on the topography elevation (Figure 5).

3.3. Driving Factors of Carbon Storage on Terrain Gradients

The geodetector model was employed to elucidate the factors influencing carbon storage from 1990 to 2020. The findings are analogous. However, due to space limitations, only the results from 2010 to 2020 are presented. The extent of land use (0.5046), annual average temperature (0.2860), and soil texture (0.0046) all had a significant impact on carbon storage change. In addition to these factors, the following factors, in reverse order, have an impact on carbon storage: distance to roads (0.0951), annual rainfall (0.0840), vegetation coverage (0.0716), average fertilizer input per land (0.0685), and so on. In sum, land use and annual average temperature presented a significant effect on carbon storage variations, compared to the distance to the urban center and soil texture, which indicate a slight impact (Figure 6).

The correlation between different factors has been expressed as nonlinear enhancement, dual-factor enhancement, and single-factor nonlinear weakening. Accordingly, the correlation between the driving factors was mainly nonlinear. For instance, the extent of land use and the distance to the urban center (0.5312), the extent of land use, and the distance to the wetland (0.5259) depicted nonlinearly significant results. Also, a few factors, such as the extent of land use and rainfall, show a double-factor enhancement effect. Further, the correlation between annual average temperature and the extent of land use (0.5484) also indicates a double-factor enhancement correlation. Population density and the proportion of scientific research and technology service industry (0.0416), pollutant emissions (0.0416), and the proportion of scientific research and technology service industry and pollutant emissions (0.0416) all show a weakening effect. Compared to the nexus between annual average temperature and other factors, which was slight, the correlation between the extent of land use and other factors was more significant. In summary, while the correlation among natural factors, for example, was minimal, the link between socio-economic, political, and natural factors was highly significant (

Table 5. Correlation between influencing factors from 2010 to 2020.

	X1	X2	X3	X4	X5	X6	X7	X8	X9	X10	X11	X12	X13
X1	0.2860
X2	0.4423 *	0.0840
X3	0.3004 *	0.0985 *	0.0046
X4	0.3317 +	0.1581 *	0.0818 *	0.0716
X5	0.3130 +	0.1887 *	0.0789 *	0.1359 *	0.0581
X6	0.5484 +	0.5139 +	0.5207 *	0.5498 +	0.5259 +	0.5046
X7	0.3626 *	0.1865 *	0.0717 *	0.1346 *	0.1361 *	0.5233 +	0.0569
X8	0.3607 *	0.2236 *	0.0597 *	0.1213 *	0.1384 *	0.5158 +	0.0978 +	0.0413
X9	0.3486 +	0.1774 +	0.1109 *	0.1646 +	0.1654 *	0.5234 +	0.1619 *	0.1491 *	0.0951
X10	0.3218 *	0.1604 *	0.0237 *	0.0994 *	0.0886 *	0.5312 *	0.1202 *	0.1236 *	0.1171 *	0.0165
X11	0.3607 *	0.2236 *	0.0597 *	0.1213 *	0.1384 *	0.5158 +	0.0978 +	0.0416 −	0.1491 *	0.1236 *	0.0413
X12	0.3607 *	0.2236 *	0.0597 *	0.1213 *	0.1384 *	0.5158 +	0.0978 +	0.0416 −	0.1491 *	0.1236 *	0.0416 −	0.0413
X13	0.3699 *	0.2107 *	0.0798 *	0.1434 *	0.1521 *	0.5166 +	0.1161 +	0.1019 +	0.1795 *	0.1440 *	0.1019 +	0.1019 +	0.0685

Note: * means that the correlation between two factors is a nonlinear enhancement; + means double-factor enhancement; − means nonlinear irrelevant.

).

3.3.1. Evaluate the Impact of Driving Factors on Carbon Storage at Different Elevations

The influence intensity of carbon storage at different elevation levels was analyzed by using the geo detector model. The impact of driving factors on carbon storage varies significantly according to the degree of elevation. The results showed that land use degree (0.2502) was the main influencing factor in elevation I, followed by the distance from the urban center (0.1992), and the influence intensity of annual precipitation and soil texture was the smallest. In elevation II, the impact intensity of land use degree (0.2148) is relatively high, followed by the distance from the road (0.2050), and the impact of soil texture (0.0047) is the least. In terms of altitude class III, population density (0.0775) and energy conservation and emission reduction policies (0.0759) have the highest impact. On the whole, the altitude class increased, the impact of land use degree gradually decreased, and the impact of policies gradually increased (

Table 6. Intensity of influencing factors at different elevation levels in 2010–2020.

Level	X1	X2	X3	X4	X5	X6	X7	X8	X9	X10	X11	X12	X13
Elevation I	0.1780	0.0226	0.0268	0.0561	0.1041	0.2502	0.1299	0.1119	0.0477	0.1992	0.1119	0.1119	0.1091
Elevation II	0.1997	0.0999	0.0047	0.0750	0.0360	0.2148	0.1576	0.1403	0.2050	0.1179	0.1403	0.1403	0.0916
Elevation III	0.0687	0.0596	—	0.0289	—	0.0213	0.0593	0.0775	0.0097	0.0580	0.0731	0.0759	0.0203

Note: — means that it fails the significance test (p > 0.05).

).

3.3.2. Evaluate the Impact of Driving Factors on Carbon Storage at Different Slope Levels

The influence of different factors varied significantly across different slope grades. The study revealed that as the slope grade increased, the intensity of influencing factors also changed notably. In Grade I, land use intensity (0.2085) had the greatest impact, followed by annual average temperature (0.0796). In Grade II, the influence of average yearly temperature increased significantly (0.2852), but land use degree (0.3806) remained the dominant driving factor. Additionally, the influence intensity of other factors, except for soil texture, also increased. When the slope grade reached Grade III, the impact of both annual average temperature (0.2371) and land use degree (0.2574) decreased, though they remained the primary influencing factors. In Grade IV, average yearly temperature (0.1343) became the main influencing factor, followed by policy-related factors. Overall, as the slope grade increased, the influence of factors shifted from socio-economic influences to natural and policy-related factors (

Table 7. Intensity of influencing factors at different slope levels in 2010–2020.

Level	X1	X2	X3	X4	X5	X6	X7	X8	X9	X10	X11	X12	X13
Slope I	0.0796	0.0223	0.0203	0.0106	0.0096	0.2085	0.0108	0.0087	0.0232	0.0039	0.0101	0.0098	0.0191
Slope II	0.2852	—	0.0130	0.0887	0.0561	0.3806	0.1296	0.0737	0.0841	0.1168	0.0737	0.0737	0.0520
Slope III	0.2371	0.0695	0.0130	0.0759	0.0650	0.2574	0.1235	0.0763	0.0841	0.1221	0.0763	0.0763	0.0796
Slope IV	0.1343	0.0498	—	—	0.0923	—	—	0.0924	—	—	0.0967	0.0943	—

Note: — means that it fails the significance test (p > 0.05); The number of sample points in Slope V is too small to pass the significance test of the geographical detector.

).

3.3.3. Evaluate the Impact of Driving Factors on Carbon Storage at Different Levels of Relief

Land use intensity and annual average temperature are the primary influencing factors at the relief amplitude level, followed by average GDP, with soil texture having the least impact. Additionally, the intensity of these driving factors varied across different relief amplitude levels. In Relief Amplitude Level I, the influence of land use degree (0.2388) was relatively strong, while the influence of other factors was minimal. In Relief Amplitude Level II, both land use degree (0.2299) and annual average temperature (0.2141) saw a significant increase in their impact, followed by average GDP (0.1132). In Relief Amplitude Level III, the influence of land use degree (0.0819) and annual average temperature (0.1789) decreased, but they remained dominant, while the influence of average GDP (0.1086) increased. Overall, land use degree and annual average temperature were the main influencing factors at the relief amplitude level, though their impact showed a downward trend (

Table 8. Intensity of influencing factors at different relief levels from 2010 to 2020.

Level	X1	X2	X3	X4	X5	X6	X7	X8	X9	X10	X11	X12	X13
Relief amplitude I	0.1465	0.0489	0.0291	0.0827	0.0710	0.2388	0.0486	0.0315	0.0533	0.0360	0.0182	0.0182	0.0228
Relief amplitude II	0.2141	0.0474	0.0036	0.0648	0.0310	0.2299	0.1132	0.0796	0.0205	0.1091	0.0796	0.0796	0.0769
Relief amplitude III	0.1789	0.0306	—	0.0693	0.0505	0.0819	0.1086	0.0701	0.0085	0.0794	0.0673	0.0684	0.0840

Note: — means that it fails the significance test (p > 0.05).

).

4. Discussion

4.1. Distribution Characteristics of Carbon Storage on Terrain Gradients

Several previous studies reported the impact of elevation and mountain slopes on carbon storage [9,14,37]. In Qinghai Province, for instance, the area with the highest average carbon density is concentrated between 1200 and 1300 m altitude [5]. This carbon storage evolution is constant and indicates a sawtooth evolution according to the elevation level, while carbon storage showed a “W-shaped” change trend along with the slope degree [35].

Therefore, contrary to these previous results, the main finding for the carbon storage change in this study indicates an “N-shaped” trend in elevation, slope, and relief amplitude gradient. In essence, the relatively plain topography area, where cultivated land accounts for approximately 64% of the total area, concentrates carbon storage between 0 and 200 m. Moreover, carbon storage seems abundant between 200 and 500 m, where vegetation coverage is important with less human interference.

In contrast, the total carbon storage is relatively limited beyond 472 m, due to the small total area and poor hydrothermal conditions. This situation remains the same with slope, where carbon storage was concentrated between 0 and 6° due to the topography conditions. From the same perspective, carbon storage is scarce beyond a slope of 6°. The carbon storage loss in the southern and northwestern regions of the study area, such as the Wanda Mountains and the Xiaoxing’an Mountains, was the most severe, mainly due to human activities leading to severe forest destruction in the area. In summary, carbon storage was significant in flat areas with extensive forest land and minimal human activity.

4.2. Driving Factors and Governance Strategies for Carbon Storage

The terrain and elevation are important factors affecting soil organic carbon storage [8,9]. Further, vegetation type, climate, soil properties, land input intensity, population size, and policies have significant impacts on ecosystem carbon storage change [29,31,32,33,38]. This study examined the drivers of carbon storage changes from three perspectives: natural, socio-economic, and policy factors. The findings indicate that the extent of land use is a significant influencing factor on carbon storage in the study area. However, a single factor is inadequate to elucidate the spatial differentiation characteristics of carbon storage fully. The interplay between land use, socio-economic factors, and policy factors largely determines the spatial distribution pattern of carbon storage in the study area. Rapid urbanization, economic growth, and the population density in urban areas were the main drivers of this serious change in land use. Aboveground biomass, belowground biomass, and soil carbon density varied considerably among different land types in the study area, with land use changes, such as expansion of construction land, encroachment of cultivated and forested land, and increasing extent of land use, due to socio-economic and policy factors. These changes have led to a significant decrease in carbon storage, which is consistent with the results of previous studies [27,31,35,39]. In addition, climate change impacts the variation in carbon storage through increasing rainfall and decreasing temperature, contributing to afforestation [5]. Previous studies [32] have highlighted the negative correlation between average yearly temperature and carbon storage. Our analysis also found that annual average temperature and its link with land use were important driving factors in carbon storage spatial distribution on different terrain gradients. In addition, the impact of annual rainfall on carbon storage was relatively small.

It can be attributed to the fact that wetlands and forests regulate the local climate. Land use change in the context of different policies and socio-economic development has led to the shrinkage of wetlands and forests, which have affected the local climate and resulted in the loss of carbon storage. The interaction between land use, socio-economic factors, and policy factors determines the spatial distribution pattern of carbon storage at different terrain gradients in the study area. Socio-economic development and policy factors can cause severe changes in land use, such as the expansion of construction land areas occupying arable land and forest land, leading to a sharp loss of carbon storage. Accordingly, climate change and socio-economic change lead to rapid carbon storage loss on different terrain gradients.

The factors that drive carbon storage vary across different terrain gradients, including population density, the proportion of scientific research and technology services, pollutant emissions, and GDP. This indicates that controlling population density, promoting energy saving and emission reduction policies, science and technology industry policies, and agricultural supply-side structural reforms in high-elevation areas in the south and the northwest of the research area can effectively alleviate carbon storage loss. In addition, the carbon storage loss in elevation I, slope I, and relief amplitude I were relatively small, accounting for the main factors influencing it, namely land use and annual average temperature. Therefore, in low-elevation farmland, the implementation of the policy of returning farmland to forests and lakes should be ensured, alleviating the degradation of grasslands and promoting green agriculture.

4.3. Shortcomings and Prospects

Despite the significant findings, this research has a few limitations. These factors include policies, socio-economic factors, and natural factors from 2010 and 2020, but they do not reveal the impact of factors such as land management practices on carbon storage. Furthermore, this study does not include an analysis of the impact of land use, specifically sowing, on carbon storage. In addition, numerous studies have employed the InVEST model to analyze variations in carbon storage. The principle lies in simulating regional carbon storage changes based on two modules, namely land use types and carbon density. However, this model often considers the carbon density within the region as a fixed value, disregarding the spatial heterogeneity of carbon density [25]. Existing research indicates that the carbon density corresponding to different land use types at the plot scale varies and is influenced by climatic factors, etc. Thus, treating the carbon density of the entire region as a fixed value will overestimate the regional carbon storage value [12,40]. In contrast, our study fully considers the spatial heterogeneity of carbon density, calculates the carbon density values corresponding to different land use types based on the plot scale (The carbon density of the same land type is different in different plots), and adjusts the carbon density in accordance with climatic factors, thereby reducing potential errors in the results [35]. Compared with the InVEST model, the method in this study enhances the accuracy of carbon storage measurement [40,41]. By comparing the carbon density data of the study area with the “Carbon Density Dataset of Terrestrial Ecosystems in China in the 2010s” (experiment measurement) created by Xu Li et al. (2019) and the data summarized in the literature, it was found that the carbon density values of various land use types in the study area were within their respective ranges, with a normalized root mean square error (NRMSE) of less than 10%. This indicates that the corrected carbon density data of various land use types in the study area is in line with reality, which is consistent with the research results of Zhang. et al. (2022) and Gai. et al. (2024) [35,40]. Therefore, the NRMSE between the estimated and measured carbon reserves is also less than 10%. Although this approach improves accuracy, it must be admitted that our study did not directly measure all the carbon density data on-site, which might introduce a certain degree of error. Therefore, future research on carbon storage should address these deficiencies and provide references for policy adjustments.

5. Conclusions

In this study, we compiled an innovative database from 1990 to 2020 to provide an estimate of the carbon storage in Sanjiang Plain, China’s black soil area. In essence, through zoning statistics, the quantitative distribution method of carbon storage, and the geographic detector model, the influencing factors of carbon storage, particularly natural, socio-economic, and policy factors, have been explored to determine the carbon storage features at different terrain gradients.

The main finding based on terrain gradients is that carbon storage was concentrated in elevation I, slope I, and relief amplitude I, followed by elevation II, slope II, and relief amplitude II. The smallest amount of carbon storage was observed in elevation III, slope III, and relief amplitude III. Additionally, the spatial pattern of carbon storage remained largely unchanged from 1990 to 2020.

The carbon storage loss was most severe at elevation II, slope II, and relief amplitude II, compared to those at elevation III, slope III, and relief amplitude III. The extent of land use and annual average temperature were the dominant factors affecting the spatial distribution pattern of carbon storage. In addition, factors such as the distance to the urban center and soil texture have a tiny impact on carbon storage. In summary, the study highlights the importance of addressing issues, such as deforestation in the middle elevation areas of the south and northwest regions, for mitigating carbon storage losses. Additionally, promoting conventional tillage practices in the low-elevation areas of the north could help alleviate these losses. Consequently, the study’s contribution lies in its focus on terrain gradient variations, consideration of spatial heterogeneity, integration of climate factors, and provision of empirical evidence, all of which advance understanding in the field of carbon storage change.