Next Article in Journal
Does Collective Forest Tenure Reform Improve Forest Carbon Sequestration Efficiency and Rural Household Income in China?
Previous Article in Journal
Species Differentiation of Two Endemic Montane Oaks in China: Population Genetics, Ecological Niche and Leaf Morphology Analyses
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Forests
Volume 16
Issue 3
10.3390/f16030550
forests-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Spatial Distribution Changes and Factor Analysis of Topsoil Organic Carbon Across Different Forest Types on Hainan Island

by
Xiang Zhang
1,
Zhongyi Sun
1,2,
Yinqi Zheng
1,
Lu Dong
3,
Peng Wang
1,
Jie Zhang
1,
Jingli Lu
1 and
Lan Wu
1,*
1
Center for Eco-Environment Restoration Engineering of Hainan Province, School of Ecology, Hainan University, Haikou 570228, China
2
Hainan Baoting Tropical Rainforest Ecosystem Observation and Research Station, School of Ecology, Hainan University, Haikou 570228, China
3
Hainan Research Academy of Environmental Sciences, Haikou 571101, China
*
Author to whom correspondence should be addressed.
Forests 2025, 16(3), 550; https://doi.org/10.3390/f16030550
Submission received: 25 February 2025 / Revised: 16 March 2025 / Accepted: 18 March 2025 / Published: 20 March 2025
(This article belongs to the Special Issue Carbon Dynamics of Forest Soils Under Climate Change)
Browse Figures
Versions Notes

Abstract

:
Topsoil organic carbon (SOC, 0–20 cm) is crucial for terrestrial carbon stocks and the global carbon cycle. This study integrated field survey data, re-analysis climatic data, and remote sensing-derived environmental factors to examine SOC distribution and its drivers across forest types on Hainan Island using machine learning models and statistical analysis. The results showed that univariate analysis had limited explanatory power for forest SOC, with terrestrial plantations exhibiting significantly lower SOC than mangroves and natural forests. For mangroves, vapor pressure deficit (VPD) was the most influential factor, followed by precipitation (PRE), the normalized difference vegetation index (NDVI), and forest age; meanwhile, for terrestrial forests, VPD, altitude, PRE, and NDVI were vital drivers. The optimal models demonstrated relatively stronger predictive performance (R2 = 0.71 for mangroves; R2 = 0.81 for terrestrial forests). Mangroves showed higher average SOC (27.91 g/kg) than terrestrial forests (15.82 g/kg), while higher concentrations in the central–western region were attributed to natural terrestrial forests. This study reveals the spatial variation patterns of forest SOC and its environmental regulation mechanisms on Hainan Island, providing important references for forest carbon stock management and environmental protection.

1. Introduction

Soil carbon represents the largest carbon pool within the global terrestrial ecosystem, holding three times the carbon stock of terrestrial vegetation and twice that of the atmospheric carbon pool [1]. Even minor fluctuations in soil carbon can significantly impact the global carbon balance and contribute to global climate change. Consequently, estimating the organic carbon (OC) stock in forest soils is a critical focus in the study of global forest ecosystem carbon cycling [2,3]. OC is predominantly found within the top 1 m of soil, with the highest concentrations located in the surface layer (0–20 cm), accounting for 42.6% of the total OC [4,5,6]. This layer is highly sensitive to environmental factors and human activities, and its variations directly influence atmospheric CO2 levels [7]. However, predicting topsoil organic carbon (SOC) remains fraught with significant uncertainty, and the processes and mechanisms driving SOC variation across different spatial scales are not yet fully understood. There is an urgent need to develop methods capable of accurately predicting large-scale spatial variations in SOC, making the exploration of SOC spatial distribution and its underlying driving mechanisms a pressing priority.
Previous studies have provided critical insights into the effects of precipitation (PRE), temperature (TEM), radiation, and other factors on SOC in forest ecosystems from a plant physiological perspective. However, most of the studies were conducted under controlled conditions, focusing on seedlings over short time periods and at leaf-scale or in localized experiments, which limits the generalizability of their findings to mature forests under natural conditions [8,9,10,11,12]. Extrapolating these conclusions from leaf or individual plant level to regional scales introduces considerable uncertainties [13]. Moreover, most research has primarily targeted a limited range of species, leaving the spatial heterogeneity of forests across diverse geographical regions underexplored. The relative importance of various driving factors across different forest types remains poorly characterized [14,15]. Furthermore, traditional methods for assessing SOC distribution, such as field sampling and laboratory analysis, while reliable, are often costly, time-consuming, and restricted to small-scale studies [16].
In contrast, recent advances in machine learning (ML), data mining techniques, and remote sensing (RS) technologies have significantly improved the efficiency and accessibility of collecting and processing surface data [17]. One promising approach in SOC research is the soil–landscape relationship model, which considers the spatial variability and interdependencies of SOC alongside the impact of environmental factors. This approach has advanced with interpolation techniques and ML models, allowing for more accurate predictions [18]. By integrating RS data with geostatistical methods, this model combines nearby observed variables with ML algorithms to forecast SOC distribution. It is effective in identifying the intricate relationships between soil properties and environmental factors, helping to create more precise estimation models [19,20]. The combination of spatial interpolation and RS techniques has presented new opportunities for incorporating local variations and conducting large-scale assessments of carbon dynamics [21,22].
In recent years, ML techniques have become increasingly popular in forest carbon research, owing to their flexibility and ability to generalize across various conditions [23,24]. For instance, Wang et al. used meteorological factors as drivers and ML methods to explore the important influencing factors of SOC in southeastern Australia, and constructed a model for predicting future SOC trends [25]. Similarly, Yigini et al. used ML models to project future SOC trends in Europe, highlighting potential changes under different climate scenarios [26]. Li et al. further employed the random forest (RF) approach to illustrate how SOC dynamics are influenced by the complex interactions between climate change and environmental changes [4]. The application of these modeling approaches has advanced the field of carbon dynamics research. However, studies on the spatial heterogeneity of SOC often rely on a single model, neglecting the varying magnitudes of influence that the same factors may have across different ecosystems [27,28]. To address this issue and improve prediction accuracy, our study specifically tackles two important scientific questions: (1) Are there significant differences in SOC between terrestrial forests and coastal mangrove forests on Hainan Island? (2) Do the environmental factors influencing SOC prediction differ significantly between these ecosystem types?
The factors driving SOC variations on Hainan Island are complex and multifaceted, involving intricate interactions among various environmental and ecological factors. Additionally, many vital predictors require the analysis of large datasets, further complicating the process. Traditional methods often struggle to address these challenges, making ML a valuable tool for such analyses. Our study addresses the uncertainty of previous large-scale regional predictions that relied on a single model. By constructing a unified methodological framework, we identified the importance of influencing factors based on different forest types, used the variance inflation factor (VIF) method to remove strongly correlated factors, and separately screened the best models for terrestrial forests and coastal forests. After extending the predictions, we merged the results to obtain the outcome. This study primarily investigates whether there are significant differences in SOC between terrestrial forests and coastal mangrove forests on Hainan Island and explores the main environmental predictors influencing SOC in different forest types. The specific research goals were as follows: (1) Construct a forest SOC estimation database for Hainan Island and combine measured SOC data with relevant environmental factors to create a robust dataset. (2) Analyze the driving factors of SOC and conduct a preliminary analysis of the impact of individual factors on SOC in different forest types through univariate analysis and grouped comparisons, further quantifying the contribution of each environmental factor to SOC in different forest types using the RF method. Based on the VIF method, screen the best estimation RF models for each forest type. (3) Conduct spatial pattern analysis and examine the spatial heterogeneity and error intervals of current forest SOC on Hainan Island and summarize the characteristics of SOC across different forest types. The research findings will provide important scientific support for formulating conservation strategies for different forest types and global climate change mitigation measures.

2. Materials and Methods

2.1. Site Description

Hainan Island (19°20′–20°10′ N, 108°21′–111°03′ E), situated in the southern sea area of China, is a geographically distinct region covering approximately 35,400 km2. The island experiences a tropical monsoon maritime climate, with an annual average TEM ranging from 22 °C to 26 °C, annual solar radiation between 4600 and 5800 MJ/m2, average annual PRE varying from 1000 to 2600 mm, and annual evapotranspiration of 1300 mm [29,30]. The island’s topography is dominated by the central uplift of Wuzhi Mountain and Yingge Ridge, gradually descending outward in a distinct step-like pattern. Hainan Island boasts a forest coverage rate of 62%, with a total forest area of about 22,000 km2, of which mangroves constitute approximately 2.5%. The soil horizons on Hainan Island are typically well-defined, with the surface soil layer (0–20 cm) rich in organic matter, gradually transitioning to weathered and parent material layers in the lower soil. Common soil types include red soil, yellow soil, lateritic soil, and sandy soil [31]. The island’s extensive forest cover, diverse soil types, and unique geographical advantages collectively contribute to its abundant forest SOC.

2.2. Data Sources

2.2.1. The Third National Land Survey Data

By utilizing data from the third national land survey as the foundation for extended analysis, the study ensured both the authority and accuracy of its findings. The third national land survey is based on both the classification of land use status and the classification of the third national land survey. It has employed advanced technologies such as 3S (geographic information system, RS, and global positioning system) and internet-based platforms to refine and enhance national land use data in accordance with unified standards. This initiative aims to provide detailed and accurate current land use information, thereby strengthening the national land survey, monitoring, and statistical systems. The survey meticulously documented multiple attributes, including the current use, quality, and management characteristics of cultivated land, orchards, forests, grasslands, and aquaculture water bodies.

2.2.2. Observed SOC Data

Soil samples were collected using a soil auger from the 0–20 cm layer within each 2 m × 2 m plot. For each soil layer, samples from five points were mixed to form one composite soil sample. After removing impurities such as litter and stones, the samples were air-dried naturally. Once completely dried, the samples were ground and stored in sealed bags. Finally, the specific SOC values were determined using the concentrated sulfuric acid-potassium dichromate oxidation method [32,33].
The forest SOC data used in this study were derived from measured forest SOC (≤20 cm) on Hainan Island. A total of 194 field data points were collected, comprising 117 sample plots surveyed and collected by our research team, which included 83 mangrove points and 34 terrestrial forest points. Additionally, 77 supplementary data points were obtained from 15 scientific publications published after 2010, including 12 mangrove points and 65 terrestrial forest points. These data points were consolidated to generate a comprehensive distribution map of forest SOC across Hainan Island (Figure 1).

2.2.3. Normalized Difference Vegetation Index (NDVI) Data

The Peking University (PKU) global inventory modeling and mapping studies (GIMMS) normalized difference vegetation index (NDVI) dataset is a newly developed global NDVI product covering the period 1982–2020 [34]. It provided a calibrated and extended version of GIMMS 3rd generation, integrating data from advanced very-high-resolution radiometer and moderate resolution imaging spectroradiometer sensors to minimize artifacts caused by sensor degradation and drift. Validation results indicated that its accuracy (R2 = 0.975) is higher than that of GIMMS 3rd generation (R2 = 0.942) [35]. Due to its wide coverage and high accuracy, we used the PKU GIMMS NDVI dataset as one of the important factors to train and estimate SOC.

2.2.4. Age Data

Age data were obtained from the global forest age dataset developed by Besnard et al. [36]. This study employed the ML method to estimate and provide the global distribution of forest age in 2010, using data from 40,000 sample plots, including forest inventories, biomass, and climate data. The results demonstrated relatively strong predictive capability (with R2 values of 0.81 for primary forests and 0.99 for secondary forests). This offers a new source of information related to disturbance history and forest regeneration, which is crucial for forest SOC research.

2.2.5. Environmental Data

Due to the limited availability of environmental factor data at some study sites, missing data were filled using reliable sources. Vapor pressure deficit (VPD) data were obtained from the TerraClimate dataset, which employed climate-driven aided interpolation. This method integrated high-spatial resolution climatological normals from the WorldClim dataset with coarser spatial resolution and time-varying data from the Climatic Research Unit Time-series version 4.0 and the Japanese 55-year Reanalysis [37]. Data for photosynthetically active radiation (PAR), PRE, TEM, cloud cover fraction (CCF), wind speed (WS), and digital elevation model (DEM) were sourced from the Modern-Era Retrospective Analysis for Research and Applications, Version 2 (MERRA-2). MERRA-2 is notable for being the first long-term global re-analysis product to assimilate space-based observations of aerosols and represent their interactions with other physical processes in the climate system. Tree species richness (TSR) data were derived from high-resolution imagery, developed by Liang et al. using the forest resources survey database [38]. The database included approximately 1.3 million individual tree records and local ecological characteristics. Detailed information on all supplementary data used for scaling up is found in Table 1.

2.3. Methods

2.3.1. Technical Approach

The research process is illustrated in Figure 2. First, we integrated forest ecosystem sample survey data from Hainan Island to construct a forest SOC observation dataset. Field measurement locations were paired with RS gridded environmental factor data to create a training dataset for ML model development. Next, a univariate regression analysis was performed to compare SOC group differences across various forest types. Environmental factors were then ranked based on their importance, and factor selection was conducted using the VIF to eliminate multicollinearity. The selected factors were incorporated into candidate models for evaluation. Finally, the best-performing model was used to analyze the spatial distribution patterns of forest SOC across Hainan Island and to explore the SOC characteristics of the soil under different forest types.

2.3.2. SOC Estimation Database Construction

Supplementary environmental factor data were extracted and aligned using the measured mangrove SOC distribution points. For climate-related data, the average values over the past 30 years were calculated to represent the climatic conditions. Using a spatial resolution of 500 m × 500 m as a reference, all environmental factors and the forest distribution data from the third national land survey of China (3rd NLSC) were standardized to ensure consistency. To address missing data in the spatiotemporal series, the empirical Bayesian Kriging method and extended interpolation techniques were applied. The final dataset included the following variables: longitude, latitude, forest age (AGE), NDVI, WS, CCF, TEM, PRE, PAR, VPD, TSR, DEM, forest type, and SOC. This comprehensive dataset integrated a complete set of measured forest SOC data, with detailed information provided in Table S1.

2.3.3. Univariate Analysis and Group Comparison

To examine the response relationship between individual environmental factors and forest SOC, forests were categorized into two groups: mangroves and terrestrial forests (plantations and natural forests). Linear regression analysis was applied to numerical environmental variables for each forest type to identify the primary factors influencing SOC. Furthermore, group comparisons were conducted to analyze the differences in SOC between these forest types. This approach helps clarify the distinct drivers and characteristics of SOC distribution in mangrove and terrestrial forest ecosystems.

2.3.4. Factor Selection and Model Validation

To enhance the model’s prediction accuracy, we first conducted a Shapiro–Wilk test on the data and applied a Box–Cox transformation to the SOC data that did not meet normality assumptions. The RF method was then used to assess the importance of different factors, with strongly correlated variables being removed based on the VIF threshold [39,40]. Next, we incorporated environmental factors into various models, including the RF model, linear mixed-effects model, linear model, and multiple linear regression. We also tested other methods such as ridge regression, least absolute shrinkage and selection operator, and stepwise regression. The simplest version of each model was evaluated using several metrics: R2, Akaike information criterion (AIC), mean absolute error (MAE), and root mean square error (RMSE). Ultimately, the final model was chosen based on a comparison of these statistical indicators, and the contribution of each factor was recorded.

2.3.5. Spatial Change Analysis

The final model, based on the RF algorithm, was then applied to predict the SOC across Hainan Island. The optimal λ value from the Box–Cox transformation was used to inverse transform the predicted SOC (Figure S1). The data were subsequently corrected using the least squares method to achieve a 1:1 linear regression between the predicted and observed SOC measurements. With the model’s accuracy improved, we conducted an extension analysis using forest data from the 3rd NLSC to determine the final spatial distribution of SOC across Hainan Island. Additionally, we performed an uncertainty analysis of the simulation results to assess the error and stability of the machine learning model [41]. This analysis has helped ensure the reliability of the predictions and has provided valuable insights into the robustness of the model.

3. Results

3.1. Univariate Analysis of Different Forest SOC

In this study, the results of univariate regression analysis for the 10 factors influencing SOC on Hainan Island revealed that individual factors had limited explanatory power for SOC in both mangroves and terrestrial forests (Figure 3a–j). This suggests that forest SOC is driven by a combination of multiple factors rather than a single variable. Among the factors analyzed, PAR had the greatest impact on mangroves (Figure 3h), while the DEM was the most influential factor for terrestrial forests (Figure 3j). Group comparisons of the measured data indicated that the average SOC for mangroves (32.73 g/kg) and terrestrial forests (31.31 g/kg) did not show significant differences (Figure 3k). However, scatter plot analysis revealed that SOC in terrestrial plantations was significantly lower than that in mangroves and natural terrestrial forests. Additionally, some natural terrestrial forests exhibited higher SOC compared to both terrestrial plantations and mangroves, which highlights the variability in SOC distribution across different forest types.

3.2. Construction and Evaluation of Different Forest SOC Model

Using the data-driven RF method, the importance of environmental factors was ranked for both mangroves and terrestrial forests (Figure 4), identifying the vital drivers of SOC in different forest types on Hainan Island. For mangroves, the environmental factors were ranked in descending order of importance as follows: VPD, PRE, NDVI, AGE, PAR, CCF, WS, TEM, TSR, and DEM. In terrestrial forests, the ranking of environmental factors by importance was as follows: VPD, DEM, PRE, NDVI, AGE, CCF, PAR, TSR, WS, and TEM. These rankings highlighted the distinct factors influencing SOC in mangroves and terrestrial forests, providing valuable insights into the drivers of SOC variability across Hainan Island’s forest ecosystems.
Based on the evaluation results of statistical indicators, the RF models were obtained according to the importance ranking from RF and the VIF threshold as follows:
Mangrove model: SOC ~ VPD + DEM + PRE + TEM + TSR + AGE + NDVI
Terrestrial forest model: SOC ~ VPD + PRE + AGE
The linear regression analysis of the predicted values and actual values of the corrected models is shown in Figure 5. The results from both models showed p < 0.001, indicating that they can predict forest SOC exceptionally well. The specific performance indicators for the mangrove model were R2 = 0.71, RMSE = 10.16 g/kg, and MAE = 7.15 g/kg. The specific performance indicators for the terrestrial forest model were R2 = 0.81, RMSE = 12.09 g/kg, and MAE = 6.87 g/kg.

3.3. Spatial Analysis of Forest SOC

This study employed the RF model to analyze the spatial distribution of forest SOC across Hainan Island. The spatial distribution characteristics, illustrated in Figure 6a, were derived from a total of 77,551 predicted SOC data points, with an average value of 15.84 g/kg. The mean SOC for mangroves was significantly higher at 27.91 g/kg compared to terrestrial forests, which averaged 15.82 g/kg (Figure 7). The spatial distribution of SOC on Hainan Island exhibited distinct patterns, with higher SOC concentrations observed in the central-western region and some northeastern coastal areas, ranging from 0.04 to 106.03 g/kg (Figure 6b and Figure S3a). Areas with elevated SOC in the central-western and northeastern coastal regions also displayed higher confidence interval values (≥10 g/kg) (Figure 6b). In the central mountainous region and the southeastern coastal area, the uncertainty confidence intervals for forests were notably significant, with values significantly higher than those of forests in other regions (Figure 6b and Figure S3b). These findings highlighted the significant spatial heterogeneity of forest SOC on Hainan Island, while also indicating some uncertainty in predictions for areas with high SOC concentrations.

4. Discussion

4.1. Impact of Important Factors on Different Forest SOC

SOC in Hainan’s forests is primarily influenced by various environmental factors, with specific drivers differing between forest types. In the RF model, VPD emerges as the most important factor for both forest types. As VPD rise from the coastal region to the inland areas in the central-western part of Hainan Island, the corresponding SOC also increases. This positive correlation could be explained by the fact that higher VPD enhances carbon stabilization in soils by reducing microbial decomposition rates under drier conditions [42].
This positive correlation between VPD and SOC in terrestrial forests contradicts the typically negative relationship observed in univariate analyses and previous studies suggesting that the relationship between VPD and SOC is highly context-dependent, influenced by local climate, soil properties, and vegetation characteristics [43,44]. The greater water use efficiency of plants in high-altitude areas, along with their carbon sequestration capacity, which remains unaffected within normal VPD variation thresholds, contributes to the accumulation of SOC [45,46]. Moreover, the interaction between VPD and other environmental factors, particularly soil texture and microbial activity, may also create favorable conditions for carbon stabilization in inland regions [47,48].
For terrestrial forests, higher carbon accumulation rates correspond with higher elevations [49], though DEM was excluded from our final model due to its high correlation with PRE and VPD. These latter two variables effectively capture elevation effects from an ecological perspective, reflecting topographic influences on TEM and humidity patterns, as demonstrated in Figure 3i and Figure S2 [50,51]. In mangrove ecosystems, VPD showed the expected negative relationship with SOC, as higher VPD increases plant water stress and reduces organic carbon inputs from roots [52,53]. These contrasting responses highlight the need for ecosystem-specific approaches when modeling SOC dynamics in response to environmental drivers in order to identify VPD thresholds for different ecosystems and to better understand the response mechanisms of SOC to VPD [54].
In terrestrial forests, PRE and forest age emerged as vital secondary drivers of SOC. PRE facilitates the transport of organic material to soil through root systems [55,56] and moves surface organic matter to deeper layers while maintaining soil moisture essential for microbial activity [57]. Forest age reflects ecosystem maturity, with older forests typically developing more stable SOC pools through prolonged ecological processes, forming decomposition-resistant organic–mineral complexes [58].
In the mangrove model, the results indicate that SOC is also influenced by various environmental factors, including PRE, TEM, and AGE. The input of freshwater from PRE reduces the salinity stress on mangrove forests, decreases soil respiration, and consequently reduces the loss of soil carbon, thereby enhancing the sequestration of OC in mangrove ecosystems [59]. The climatic value of TEM has a relatively weaker impact on mangrove SOC on Hainan Island, where latitude variation is minimal. Its primary effect is through regulating plant photosynthesis and respiration, indirectly influencing the carbon cycle [60]. Although mangrove productivity tends to decline with AGE, even established ecosystems continue to function as net carbon producers, with carbon density showing a clear increasing trend over time [61,62]. High NDVIs are positively correlated with plant biomass and health, demonstrating spatial consistency. Recent studies have shown significant results when using NDVI in combination with various environmental factors to estimate SOC [63,64]. These findings underscore the complex interplay of environmental factors in shaping mangrove SOC dynamics.

4.2. Spatial Distribution Pattern of Forest SOC

The spatial variation in SOC on Hainan Island shows strong regularity, with dense and uneven contour distribution in the central region, and the SOC in the central and western regions is generally higher than in other areas (Figure S3a). As higher SOC values were measured in the high DEM areas, and the spatial distribution is uneven, there are still different natural landscape spatial features at the local scale. Most traditional soil-forming factors show minor spatial changes, hindering their contribution to the soil–landscape model development, making the selection of suitable auxiliary variables challenging [65]. These uncertainties are also reasons for the large range in the predicted interval in the central part of Hainan Island. We also predicted that the SOC of terrestrial forests in the northeastern coastal region is higher, as this area has a higher AGE compared to other regions, which increases the predicted value. In the future, more field data points are urgently needed, especially in areas with concentrated contours in high DEM regions, to improve model prediction accuracy [66].
Although mangroves are generally considered to have a carbon storage capacity 3–5 times that of terrestrial forests [67], the measurements from this study show that the SOC of mangroves (32.73 g/kg) is only slightly higher than that of terrestrial forests (31.31 g/kg). This discrepancy may be related to the impacts of human activities [68,69]. The significantly lower SOC in plantation forests (Figure 3k) demonstrates the substantial influence of human activities on SOC [70,71]. In recent years, the mangroves on Hainan Island have been affected by human activities such as land reclamation and aquaculture, which have accelerated the decomposition and loss of SOC in mangroves, thereby reducing the carbon storage capacity of mangrove ecosystems [72,73]. Future research should consider broader sampling ranges and more detailed ecosystem analysis to comprehensively understand the differences in SOC between mangroves and other forest types.

4.3. Limitations and Uncertainties

SOC in Hainan Island forests is influenced by complex factors such as soil properties, climate, and terrestrial ecosystems [4,63]. We attempted to capture this time-varying nonlinear response using ML methods. However, the model’s predictions still have certain limitations and uncertainties. For example, atmospheric nitrogen deposition rates, fire frequency, abnormal PRE, typhoons, and land use changes may all affect the predictions. Given the positive effects of increased atmospheric CO2 concentrations and nitrogen deposition on plant carbon sequestration, our results may have underestimated the SOC [26,27]. Additionally, the environmental factors selected for different forest types are distinct. The environmental factors are considered to have some influence on all forests. Although more environmental factors were included in the mangrove model, the evaluation effect was not as strong as in the terrestrial forest model with fewer factors. For forest communities in different habitats, more targeted and strongly correlated environmental factors should be considered, such as tidal range and seawater TEM in mangroves [21,29]. Environmental factors, forest types, and models need further refinement to improve prediction accuracy.
For different soil types in various regions, there are significant differences in soil properties. Even within the same soil type, certain soil properties can differ across time and space, meaning SOC exhibits high spatial heterogeneity over time and space [74]. Simultaneously, the importance of environmental factors for SOC spatial distribution is scale-dependent. Although producers have used many methods to ensure the accuracy of these datasets, it must be acknowledged that resolution mismatches and uncertainties in modeling values are major reasons for the differences in environmental factor importance [21,75]. We used interpolation methods to fill missing values for environmental factors, and based on spatial autocorrelation, there are certain errors in the interpolation of factors with spatial heterogeneity, such as AGE and TSR. Additionally, due to the lower resolution of TSR, the tree species count in a pixel is much higher than the number of tree species surveyed in the field, leading to greater errors in higher-precision SOC statistics [76].

5. Conclusions

This study utilized data-driven ML methods to analyze the environmental control mechanisms and spatial distribution of SOC in different forest types on Hainan Island. The main findings are as follows: no significant factors were found in the univariate analysis for both mangrove and terrestrial forests, with the measured SOC of mangroves (32.73 g/kg) being slightly higher than that of terrestrial forests (31.31 g/kg). Integrated data analysis revealed that the most influential factor for mangroves is VPD, followed by PRE, NDVI, and AGE; for terrestrial forests, the most influential factor is also VPD, followed by DEM, PRE, and NDVI. The constructed RF model effectively predicted the spatial distribution of SOC on Hainan Island, with an average of 15.84 g/kg. High SOC concentrations were found in the high DEM regions in the central and western parts, where the predicted SOC also showed some uncertainty. The prediction results indicated that the average SOC of mangroves was 27.91 g/kg, significantly higher than that of terrestrial forests (15.82 g/kg), demonstrating their strong carbon sequestration capacity. Our study enhances the understanding of SOC dynamics in different forest types in tropical regions, providing essential scientific evidence for regional carbon management and global climate change mitigation efforts.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/f16030550/s1, Figure S1: Box–Cox normality transformation; Figure S2: Terrestrial forest matrix correlation plot; Figure S3: Hotspot analysis diagram: (a) distribution of forest SOC hotspots on Hainan Island and (b) distribution of Forest SOC error hotspots on Hainan Island; Table S1: Hainan Forest soil SOC data collection table; Table S2: Vocabulary abbreviations table.

Author Contributions

Conceptualization, methodology, and writing—original draft preparation, X.Z. and Z.S.; supervision, project administration, and funding acquisition, L.W. and Z.S.; data collection, Y.Z. and Z.S.; data processing and visualization, X.Z. and Z.S.; investigation and proofreading, J.L. and L.D.; visualization, resources, and investigation, P.W. and X.Z.; research design, manuscript review, and editing, J.L. and J.Z.; reviewing and editing the manuscript, Z.S. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the National Key Research and Development Program of China (Grant No. 2021YFD2200404), the National Natural Science Foundation of China (Grant No. 32460334), the Hainan Provincial Natural Science Foundation High-Level Talent Project (Grant No. 322RC580), and the Hainan Key Laboratory of Marine Geological Resources and Environment (Grant No. 23-HNHYDZZYHJKF035).

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Acknowledgments

The authors would like to thank the editors and the anonymous reviewers for their crucial comments, which improved the quality of this paper.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. Lal, R. Soil carbon sequestration impacts on global climate change and food security. Science 2004, 304, 1623–1627. [Google Scholar] [PubMed]
  2. Li, B.; Fang, X.; Li, Y.; Xiang, W.; Tian, D.; Chen, X.; Yan, W.; Deng, D. Dynamic properties of soil organic carbon in Hunan’s forests. Acta Ecol. Sin. 2015, 35, 4265–4278. [Google Scholar]
  3. IPCC. Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. Available online: https://www.ipcc.ch/report/ar6/wg1/downloads/report/IPCC_AR6_WGI_Full_Report.pdf (accessed on 22 November 2024).
  4. Li, H.; Wu, Y.; Liu, S.; Xiao, J.; Zhao, W.; Chen, J.; Alexandrov, G.; Cao, Y. Decipher soil organic carbon dynamics and driving forces across China using machine learning. Glob. Change Biol. 2022, 28, 3394–3410. [Google Scholar]
  5. Ding, J.; Chen, L.; Ji, C.; Hugelius, G.; Li, Y.; Liu, L.; Qin, S.; Zhang, B.; Yang, G.; Li, F.; et al. Decadal soil carbon accumulation across Tibetan permafrost regions. Nat. Geosci. 2017, 10, 420–424. [Google Scholar]
  6. Zhao, Y.; Wang, M.; Hu, S.; Zhang, X.; Ouyang, Z.; Zhang, G.; Huang, B.; Zhao, S.; Wu, J.; Xie, D.; et al. Economics- and policy-driven organic carbon input enhancement dominates soil organic carbon accumulation in Chinese croplands. Proc. Natl. Acad. Sci. USA 2018, 115, 4045–4050. [Google Scholar]
  7. Mohanty, B.P.; Famiglietti, J.S.; Skaggs, T.H. Evolution of soil moisture spatial structure in a mixed vegetation pixel during the Southern Great Plains 1997 (SGP97) Hydrology Experiment. Water Resour. Res. 2000, 36, 3675–3686. [Google Scholar]
  8. Andrews, T.J.; Muller, G.J. Photosynthetic Gas Exchange of the Mangrove, Rhizophora stylosa Griff., in Its Natural Environment. Oecologia 1985, 65, 449–455. [Google Scholar]
  9. Bjorkman, O.; Demmig, B.; Andrews, T.J. Mangrove Photosynthesis: Response to High-Irradiance Stress. Funct. Plant Biol. 1988, 15, 43–61. [Google Scholar]
  10. Ball, M.C. Interactive Effects of Salinity and Irradiance on Growth: Implications for Mangrove Forest Structure along Salinity Gradients. Trees 2002, 16, 126–139. [Google Scholar] [CrossRef]
  11. Parida, A.K.; Das, A.B.; Mittra, B. Effects of Salt on Growth, Ion Accumulation, Photosynthesis and Leaf Anatomy of the Mangrove, Bruguiera parviflora. Trees 2004, 18, 167–174. [Google Scholar] [CrossRef]
  12. Lopez-Hoffman, L.; DeNoyer, J.L.; Monroe, I.E.; Shaftel, R.; Anten, N.P.; Martínez-Ramos, M.; Ackerly, D.D. Mangrove Seedling Net Photosynthesis, Growth, and Survivorship Are Interactively Affected by Salinity and Light. Biotropica 2006, 38, 606–616. [Google Scholar]
  13. Raupach, M.R.; Rayner, P.J.; Barrett, D.J.; DeFries, R.S.; Heimann, M.; Ojima, D.S.; Schmullius, C.C. Model-Data Synthesis in Terrestrial Carbon Observation: Methods, Data Requirements and Data Uncertainty Specifications. Glob. Change Biol. 2005, 11, 378–397. [Google Scholar]
  14. Hagger, V.; Worthington, T.A.; Lovelock, C.E.; Adame, M.F.; Amano, T.; Brown, B.M.; Saunders, M.I. Drivers of Global Mangrove Loss and Gain in Social-Ecological Systems. Nat. Commun. 2022, 13, 6373. [Google Scholar] [PubMed]
  15. Arnaud, M.; Krause, S.; Norby, R.J.; Dang, T.H.; Acil, N.; Kettridge, N.; Ullah, S. Global Mangrove Root Production, Its Controls and Roles in the Blue Carbon Budget of Mangroves. Glob. Change Biol. 2023, 29, 3256–3270. [Google Scholar]
  16. Allen, D.E.; Pringle, M.J.; Page, K.L.; Dalal, R.C. A review of sampling designs for the measurement of soil organic carbon in Australian grazing lands. Rangel. J. 2010, 32, 227–246. [Google Scholar]
  17. Zhang, Z.; Ding, J.; Zhu, C.; Wang, J.; Ge, X.; Li, X.; Han, L.; Chen, X.; Wang, J. Historical and future variation of soil organic carbon in China. Geoderma 2023, 436, 116557. [Google Scholar]
  18. Thompson, J.A.; Kolka, R.K. Soil carbon storage estimation in a forested watershed using quantitative soil–landscape modeling. Soil Sci. Soc. Am. J. 2005, 69, 1086–1093. [Google Scholar]
  19. Tao, F.; Zhou, Z.; Huang, Y.; Li, Q.; Lu, X.; Ma, S.; Huang, X.; Liang, Y.; Hugelius, G.; Jiang, L.; et al. Deep learning optimizes data-driven representation of soil organic carbon in earth system model over the conterminous United States. Front. Big Data 2020, 3, 17. [Google Scholar]
  20. Yan, L.; Zhou, G.; Wang, Y.; Hu, T.; Sui, X. The spatial and temporal dynamics of carbon budget in the alpine grasslands on the Qinghai-Tibetan Plateau using the Terrestrial Ecosystem Model. J. Cleaner Prod. 2015, 107, 195–201. [Google Scholar]
  21. Sun, Z.; An, Y.; Kong, J.; Zhao, J.; Cui, W.; Nie, T.; Zhang, T.; Liu, W.; Wu, L. Exploring the spatio-temporal patterns of global mangrove gross primary production and quantifying the factors affecting its estimation, 1996–2020. Sci. Total Environ. 2024, 908, 168262. [Google Scholar]
  22. Hamilton, S.E.; Friess, D.A. Global carbon stocks and potential emissions due to mangrove deforestation from 2000 to 2012. Nat. Clim. Change 2018, 8, 240–244. [Google Scholar] [CrossRef]
  23. Pham, T.D.; Yokoya, N.; Bui, D.T.; Yoshino, K.; Friess, D.A. Remote sensing approaches for monitoring mangrove species, structure, and biomass: Opportunities and challenges. Remote Sens. 2019, 11, 230. [Google Scholar] [CrossRef]
  24. Worthington, T.A.; Andradi-Brown, D.A.; Bhargava, R.; Buelow, C.; Bunting, P.; Duncan, C.; Fatoyinbo, L.; Friess, D.A.; Goldberg, L.; Hilarides, L.; et al. Harnessing big data to support the conservation and rehabilitation of mangrove forests globally. One Earth 2020, 2, 429–443. [Google Scholar] [CrossRef]
  25. Wang, B.; Gray, J.M.; Waters, C.M.; Anwar, M.R.; Orgill, S.E.; Cowie, A.L.; Feng, P.; Liu, D.L. Modelling and mapping soil organic carbon stocks under future climate change in south-eastern Australia. Geoderma 2022, 405, 115442. [Google Scholar] [CrossRef]
  26. Yigini, Y.; Panagos, P. Assessment of soil organic carbon stocks under future climate and land cover changes in Europe. Sci. Total Environ. 2016, 557–558, 838–850. [Google Scholar] [CrossRef]
  27. Chen, Y.; Xiong, K.; Luo, L.; Chen, J. Ecosystem services supply-demand mismatch and their drivers in highly spatially heterogeneous South China karst ecoregions. Land Degrad. Dev. 2024, 35, 816–831. [Google Scholar] [CrossRef]
  28. Wang, Z. Geostatistics and its applications in ecology. Sci. Press 1999, 1, 123–145. [Google Scholar]
  29. Shi, X.; Nie, T.; Xiong, Q.; Liu, Z.; Zhang, J.; Liu, W.; Wu, L.; Cui, W.; Sun, Z. Assessment of carbon stock and sequestration of the mangrove ecosystems on Hainan Island based on InVEST and MaxEnt models. J. Trop. Biol. 2023, 14, 298–306. [Google Scholar]
  30. Li, W.; Zhang, J.; Liu, S.; Che, X.; Chen, X.; Zou, H. Meteorological Characteristics and Monitoring Indicators of Drought in Hainan Island. J. Trop. Biol. 2022, 13, 324–330. [Google Scholar]
  31. National Soil Survey Office of China. Soil Data of China; Resources and Environment Data Cloud Platform: Beijing, China, 1995; Available online: http://www.resdc.cn/data.aspx?DATAID=145 (accessed on 4 March 2025).
  32. Bao, S. Soil Agrochemical Analysis, 3rd ed; China Agriculture Press: Beijing, China, 2000; pp. 30–34. [Google Scholar]
  33. Zhang, S.Q.; Xu, Q.; Yang, Q.; Jiang, Y.M.; Wang, X.; Liu, W. Soil Carbon, Nitrogen, Phosphorus Content and Carbon-to-Nitrogen Ratio in Sandy Soils of the Coastal Zone of Hainan Island. J. For. Environ. 2019, 39, 398–403. [Google Scholar]
  34. Li, M.; Cao, S.; Zhu, Z. Spatiotemporally consistent global dataset of the GIMMS normalized difference vegetation index (PKU GIMMS NDVI) from 1982 to 2020. Earth Syst. Sci. Data Discuss. 2023, 15, 4181–4203. [Google Scholar]
  35. Pinzon, J.E.; Tucker, C.J. A non-stationary 1981–2012 AVHRR NDVI3g time series. Remote Sens. 2014, 6, 6929–6960. [Google Scholar] [CrossRef]
  36. Besnard, S.; Koirala, S.; Santoro, M.; Weber, U.; Nelson, J.; Gütter, J.; Herault, B.; Kassi, J.; N’Guessan, A.; Neigh, C.; et al. Mapping global forest age from forest inventories, biomass and climate data. Earth Syst. Sci. Data Discuss. 2021, 2021, 4881–4896. [Google Scholar]
  37. Abatzoglou, J.T.; Dobrowski, S.Z.; Parks, S.A. TerraClimate, a high-resolution global dataset of monthly climate and climatic water balance from 1958–2015. Sci. Data 2018, 5, 170191. [Google Scholar] [CrossRef] [PubMed]
  38. Liang, J.; Gamarra, J.G.; Picard, N.; Zhou, M.; Pijanowski, B.; Jacobs, D.F.; Reich, P.B.; Crowther, T.W.; Nabuurs, G.J.; De-Miguel, S.; et al. Co-limitation towards lower latitudes shapes global forest diversity gradients. Nat. Ecol. Evol. 2022, 6, 1423–1437. [Google Scholar]
  39. Fanin, N.; Bezaud, S.; Sarneel, J.M.; Cecchini, S.; Nicolas, M.; Augusto, L. Relative importance of climate, soil and plant functional traits during the early decomposition stage of standardized litter. Ecosystems 2020, 23, 1004–1018. [Google Scholar] [CrossRef]
  40. Kock, N. Common method bias in PLS-SEM: A full collinearity assessment approach. Int. J. e-Collab. 2015, 11, 1–10. [Google Scholar]
  41. Viscarra Rossel, R.A.; Webster, R.; Bui, E.N.; Baldock, J.A. Baseline map of organic carbon in Australian soil to support national carbon accounting and monitoring under climate change. Glob. Change Biol. 2014, 20, 2953–2970. [Google Scholar]
  42. Dubey, N.; Ghosh, S. The relative role of soil moisture and vapor pressure deficit in affecting the Indian vegetation productivity. Environ. Res. Lett. 2023, 18, 064012. [Google Scholar]
  43. Sulman, B.N.; Roman, D.T.; Yi, K.; Wang, L.; Phillips, R.P.; Novick, K.A. High atmospheric demand for water can limit forest carbon uptake and transpiration as severely as dry soil. Geophys. Res. Lett. 2016, 43, 9686–9695. [Google Scholar] [CrossRef]
  44. Eamus, D.; Boulain, N.; Cleverly, J.; Breshears, D.D. Global change-type drought-induced tree mortality: Vapor pressure deficit is more important than temperature per se in causing decline in tree health. Ecol. Evol. 2013, 3, 2711–2729. [Google Scholar] [CrossRef] [PubMed]
  45. Peñuelas, J.; Canadell, J.G.; Ogaya, R. Increased water-use efficiency during the 20th century did not translate into enhanced tree growth. Glob. Ecol. Biogeogr. 2011, 20, 597–608. [Google Scholar]
  46. Leonardi, S.; Gentilesca, T.; Guerrieri, R.; Ripullone, F.; Magnani, F.; Mencuccini, M.; Noije, T.V.; Borghetti, M. Assessing the effects of nitrogen deposition and climate on carbon isotope discrimination and intrinsic water-use efficiency of angiosperm and conifer trees under rising CO2 conditions. Glob. Change Biol. 2012, 18, 2925–2944. [Google Scholar]
  47. Leopold, A.; Marchand, C.; Renchon, A.; Deborde, J.; Quiniou, T.; Allenbach, M. Net ecosystem CO2 exchange in the “Coeur de Voh” mangrove, New Caledonia: Effects of water stress on mangrove productivity in a semi-arid climate. Agric. For. Meteorol. 2016, 223, 217–232. [Google Scholar]
  48. Leopold, A.; Marchand, C.; Allenbach, M.; Renchon, A.; Deborde, J.; Bourgeois, C.; Quiniou, T. Net ecosystem CO2 exchanges between a dwarf Avicennia marina var. Australasica mangrove and the atmosphere: Application of the eddy-covariance flux tower to “Le Coeur de Voh” mangrove (New Caledonia). In The Mangrove & Macrobenthos Meeting (MMM4); Elsevier: Amsterdam, The Netherlands, 2016; Volume 223. [Google Scholar]
  49. Davidson, E.A.; Janssens, I.A. Temperature sensitivity of soil carbon decomposition and feedbacks to climate change. Nature 2006, 440, 165–173. [Google Scholar]
  50. Xu, S.; Wu, C.; Wang, L.; Gonsamo, A.; Shen, Y.; Niu, Z. A new satellite-based monthly precipitation downscaling algorithm with non-stationary relationship between precipitation and land surface characteristics. Remote Sens. Environ. 2015, 162, 119–140. [Google Scholar]
  51. Zhang, H.; Wu, B.; Yan, N.; Zhu, W.; Feng, X. An improved satellite-based approach for estimating vapor pressure deficit from MODIS data. J. Geophys. Res. Atmos. 2014, 119, 12256. [Google Scholar]
  52. Gou, R.; Chi, J.; Liu, J. Atmospheric water demand constrains net ecosystem production in subtropical mangrove forests. J. Hydrol. 2024, 630, 130651. [Google Scholar]
  53. Wu, D.; Li, G.; Kang, Q.; Zhang, X.; Cao, Z.; Li, Z. Characteristics of CO2 flux and its influence factors over winter wheat agroecosystem in the North China Plain. J. Appl. Ecol. 2018, 29, 827–838. [Google Scholar]
  54. Green, J.K.; Berry, J.; Ciais, P.; Zhang, Y.; Gentine, P. Amazon rainforest photosynthesis increases in response to atmospheric dryness. Sci. Adv. 2020, 6, eabb7232. [Google Scholar]
  55. Pan, Y.; Wang, Y.; Xin, J.; Tang, G.; Song, T.; Wang, Y.; Li, X.; Wu, F. Study on dissolved organic carbon in precipitation in Northern China. Atmos. Environ. 2010, 44, 2350–2357. [Google Scholar] [CrossRef]
  56. Chen, S.; Huang, Y.; Zou, J.; Shi, Y. Mean residence time of global topsoil organic carbon depends on temperature, precipitation and soil nitrogen. Glob. Planet. Change 2013, 100, 99–108. [Google Scholar] [CrossRef]
  57. Pregitzer, K.S.; Euskirchen, E.S. Carbon cycling and storage in world forests: Biome patterns related to forest age. Glob. Change Biol. 2004, 10, 2052–2077. [Google Scholar] [CrossRef]
  58. Chen, H.Y.; Shrestha, B.M. Stand age, fire and clearcutting affect soil organic carbon and aggregation of mineral soils in boreal forests. Soil Biol. Biochem. 2012, 50, 149–157. [Google Scholar] [CrossRef]
  59. Sanders, C.J.; Maher, D.T.; Tait, D.R.; Williams, D.; Holloway, C.; Sippo, J.Z.; Santos, I.R. Are global mangrove carbon stocks driven by rainfall? J. Geophys. Res. Biogeosci. 2016, 121, 2600–2609. [Google Scholar] [CrossRef]
  60. Osland, M.J.; Gabler, C.A.; Grace, J.B.; Day, R.H.; McCoy, M.L.; McLeod, J.L.; From, A.S.; Enwright, N.M.; Feher, L.C.; Stagg, C.L.; et al. Climate and plant controls on soil organic matter in coastal wetlands. Glob. Change Biol. 2018, 24, 5361–5379. [Google Scholar] [CrossRef]
  61. Gao, T.; Guan, W.; Mao, J.; Jiang, Z.; Liao, B. Carbon storage and influence factors of major mangrove communities in Fucheng, Leizhou Peninsula, Guangdong Province. Ecol. Environ. Sci. 2017, 26, 985–990. [Google Scholar]
  62. Alongi, D.M. Carbon sequestration in mangrove forests. Carbon Manag. 2012, 3, 313–322. [Google Scholar] [CrossRef]
  63. Wang, S.; Fan, J.; Zhong, H.; Li, Y.; Zhu, H.; Qiao, Y.; Zhang, H. A multi-factor weighted regression approach for estimating the spatial distribution of soil organic carbon in grasslands. Catena 2019, 174, 248–258. [Google Scholar] [CrossRef]
  64. Khormali, F.; Ajami, M.; Ayoubi, S.; Srinivasarao, C.; Wani, S.P. Role of deforestation and hillslope position on soil quality attributes of loess-derived soils in Golestan province, Iran. Agric. Ecosyst. Environ. 2009, 134, 178–189. [Google Scholar] [CrossRef]
  65. Zhang, Y.; Guo, L.; Chen, Y.; Shi, T.; Luo, M.; Ju, Q.; Zhang, H.; Wang, S. Prediction of soil organic carbon based on Landsat 8 monthly NDVI data for the Jianghan Plain in Hubei Province, China. Remote Sens. 2019, 11, 1683. [Google Scholar] [CrossRef]
  66. Sun, Z.; Wang, X.; Zhang, X.; Tani, H.; Guo, E.; Yin, S.; Zhang, T. Evaluating and comparing remote sensing terrestrial GPP models for their response to climate variability and CO2 trends. Sci. Total Environ. 2019, 668, 696–713. [Google Scholar] [PubMed]
  67. Mcleod, E.; Chmura, G.L.; Bouillon, S.; Salm, R.; Björk, M.; Duarte, C.M.; Lovelock, C.E.; Schlesinger, W.H.; Silliman, B.R. A blueprint for blue carbon: Toward an improved understanding of the role of vegetated coastal habitats in sequestering CO2. Front. Ecol. Environ. 2011, 9, 552–560. [Google Scholar] [CrossRef]
  68. Sun, W.; Liu, X. Review on carbon storage estimation of forest ecosystem and applications in China. For. Ecosyst. 2020, 7, 4. [Google Scholar]
  69. Wang, X.; Feng, Z.; Ouyang, Z. The impact of human disturbance on vegetative carbon storage in forest ecosystems in China. For. Ecol. Manag. 2001, 148, 117–123. [Google Scholar]
  70. Zeng, Y.; Fang, N.; Shi, Z. Effects of human activities on soil organic carbon redistribution at an agricultural watershed scale on the Chinese Loess Plateau. Agric. Ecosyst. Environ. 2020, 303, 107112. [Google Scholar]
  71. Mohamed, E.S.; Abu-hashim, M.; AbdelRahman, M.A.; Schütt, B.; Lasaponara, R. Evaluating the effects of human activity over the last decades on the soil organic carbon pool using satellite imagery and GIS techniques in the Nile Delta Area, Egypt. Sustainability 2019, 11, 2644. [Google Scholar] [CrossRef]
  72. Wang, J.; Wang, H.; Ding, Y.; Zhang, Y.; Cong, W.; Zang, R.; Liu, S. Shifting cultivation and logging change soil organic carbon functional groups in tropical lowland rainforests on Hainan Island in China. For. Ecol. Manag. 2023, 549, 121447. [Google Scholar]
  73. Lin, J.; Tian, Y.; Zhang, Y.; Bai, X.; Zhang, Q.; Tao, J.; Yang, Y.; Yang, Y. Spatial distribution characteristics of soil organic carbon in mangrove wetlands at the Maoling River estuary, Beibu Gulf. Mar. Environ. Sci. 2024, 43, 339–347. [Google Scholar]
  74. Wang, Z.; Wang, Q. Study on the spatial heterogeneity of forest soil physical properties. Acta Ecol. Sin. 2000, 6, 945–950. [Google Scholar]
  75. Wiesmeier, M.; Urbanski, L.; Hobley, E.; Lang, B.; Lützow, M.; Marin-Spiotta, E.; Wesemael, B.; Rabot, E.; Ließ, M.; Garcia-Franco, N.; et al. Soil organic carbon storage as a key function of soils-A review of drivers and indicators at various scales. Geoderma 2019, 333, 149–162. [Google Scholar]
  76. Ren, Y.; Mao, D.; Wang, Z.; Yu, Z.; Xu, X.; Huang, Y.; Xi, Y.; Luo, L.; Jia, M.; Song, K.; et al. China’s wetland soil organic carbon pool: New estimation on pool size, change, and trajectory. Glob. Change Biol. 2023, 29, 6139–6156. [Google Scholar]
Forests 16 00550 g001
Figure 1. Distribution map of topsoil organic carbon (SOC) measurement points on Hainan Island. Circles represent terrestrial forests, squares represent mangroves, and different colors indicate varying SOC values.
Figure 1. Distribution map of topsoil organic carbon (SOC) measurement points on Hainan Island. Circles represent terrestrial forests, squares represent mangroves, and different colors indicate varying SOC values.
Forests 16 00550 g001
Forests 16 00550 g002
Figure 2. Study flow. SOC: topsoil organic carbon; ML: machine learning; VIF: variance inflation factor; 3rd NLSC: third national land survey of China.
Figure 2. Study flow. SOC: topsoil organic carbon; ML: machine learning; VIF: variance inflation factor; 3rd NLSC: third national land survey of China.
Forests 16 00550 g002
Forests 16 00550 g003
Figure 3. SOC univariate linear regression. The red section represents mangroves, and the cyan section represents terrestrial forests: (a) CCF (cloud cover fraction); (b) NDVI (normalized difference vegetation index); (c) PRE (precipitation); (d) AGE (forest age); (e) TEM (temperature); (f) VPD (vapor pressure deficit); (g) WS (wind speed); (h) PAR (photosynthetically active radiation); (i) DEM (digital elevation model); (j) TSR (tree species richness); (k) group comparison of different forest types.
Figure 3. SOC univariate linear regression. The red section represents mangroves, and the cyan section represents terrestrial forests: (a) CCF (cloud cover fraction); (b) NDVI (normalized difference vegetation index); (c) PRE (precipitation); (d) AGE (forest age); (e) TEM (temperature); (f) VPD (vapor pressure deficit); (g) WS (wind speed); (h) PAR (photosynthetically active radiation); (i) DEM (digital elevation model); (j) TSR (tree species richness); (k) group comparison of different forest types.
Forests 16 00550 g003
Forests 16 00550 g004
Figure 4. Ranking of environmental factor importance. The left side represents mangroves, and the right side represents terrestrial forests. Different colors indicate different environmental factor variables.
Figure 4. Ranking of environmental factor importance. The left side represents mangroves, and the right side represents terrestrial forests. Different colors indicate different environmental factor variables.
Forests 16 00550 g004
Forests 16 00550 g005
Figure 5. Measured–predicted SOC regression for different forest types. Orange represents mangroves, blue represents terrestrial forests, and the shaded area indicates the 95% confidence interval.
Figure 5. Measured–predicted SOC regression for different forest types. Orange represents mangroves, blue represents terrestrial forests, and the shaded area indicates the 95% confidence interval.
Forests 16 00550 g005
Forests 16 00550 g006
Figure 6. Spatial distribution of forest SOC on Hainan Island: (a) SOC distribution on Hainan Island; (b) spatial distribution of SOC error interval on Hainan Island.
Figure 6. Spatial distribution of forest SOC on Hainan Island: (a) SOC distribution on Hainan Island; (b) spatial distribution of SOC error interval on Hainan Island.
Forests 16 00550 g006
Forests 16 00550 g007
Figure 7. Statistical comparison of average SOC for all forest types on Hainan Island.
Figure 7. Statistical comparison of average SOC for all forest types on Hainan Island.
Forests 16 00550 g007
Table 1. Overview of spatial datasets.
Table 1. Overview of spatial datasets.
Variable TypeDataResolution (°)DatasetDetail
MeteorologyVPD0.0416 × 0.0416TerraClimateVapor pressure deficit
MeteorologyPAR0.5 × 0.625MERRA-2Photosynthetically active radiation
MeteorologyPRE0.5 × 0.625MERRA-2Precipitation
MeteorologyCCF0.5 × 0.625MERRA-2Cloud cover fraction
MeteorologyWS0.5 × 0.625MERRA-2Wind speed
MeteorologyTEM0.1 × 0.1MERRA-2Temperature
GeographyDEM0.00027 × 0.00027MERRA-2Digital elevation model
BotanyTSR0.025 × 0.025Science-iTotal species richness
BotanyNDVI0.0083 × 0.0083PKU GIMMS NDVINormalized difference vegetation index
BotanyAGE0.0083 × 0.0083MPI-BGC forest age datasetForest age
BotanyForest distribution-3rd NLSCThird national land survey of China
The following data sources were used: TerraClimate: https://www.climatologylab.org/ (accessed on 3 July 2024); MERRA-2: https://www.earthdata.nasa.gov/ (accessed on 18 July 2024); PKU GIMMS NDVI: https://zenodo.org/ (accessed on 20 August 2024) [34]; Science-i: https://science-i.org/ (accessed on 3 September 2024) [38]; and MPI-BGC forest age dataset: https://www.bgc-jena.mpg.de/ (accessed on 11 September 2024) [36].
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Zhang, X.; Sun, Z.; Zheng, Y.; Dong, L.; Wang, P.; Zhang, J.; Lu, J.; Wu, L. Spatial Distribution Changes and Factor Analysis of Topsoil Organic Carbon Across Different Forest Types on Hainan Island. Forests 2025, 16, 550. https://doi.org/10.3390/f16030550

AMA Style

Zhang X, Sun Z, Zheng Y, Dong L, Wang P, Zhang J, Lu J, Wu L. Spatial Distribution Changes and Factor Analysis of Topsoil Organic Carbon Across Different Forest Types on Hainan Island. Forests. 2025; 16(3):550. https://doi.org/10.3390/f16030550

Chicago/Turabian Style

Zhang, Xiang, Zhongyi Sun, Yinqi Zheng, Lu Dong, Peng Wang, Jie Zhang, Jingli Lu, and Lan Wu. 2025. "Spatial Distribution Changes and Factor Analysis of Topsoil Organic Carbon Across Different Forest Types on Hainan Island" Forests 16, no. 3: 550. https://doi.org/10.3390/f16030550

APA Style

Zhang, X., Sun, Z., Zheng, Y., Dong, L., Wang, P., Zhang, J., Lu, J., & Wu, L. (2025). Spatial Distribution Changes and Factor Analysis of Topsoil Organic Carbon Across Different Forest Types on Hainan Island. Forests, 16(3), 550. https://doi.org/10.3390/f16030550

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop