Trade-Offs in Aboveground and Soil Mangrove Carbon Stocks Under Species Introduction: Remote Sensing Reveals Temporal Divergence in Restoration Trajectories

Abstract

Mangrove ecosystems play a critical role in global carbon cycling, serving as significant carbon sinks by storing carbon in both aboveground biomass (ACG) and soil carbon stock (SOC). However, the temporal dynamics of ACG and SOC, as well as their spatial variations across different mangrove age stages, remain poorly understood, particularly under the influence of introduced species such as Sonneratia apetala Buch.-Ham. To address these gaps, our study used a long-term series of NDVI from Landsat (from 1990 to 2024) and the mangrove product of China (1990, 2000, 2010, and 2018) to estimate the mangrove age stage ($\mathrm{Stage}\mathrm{I}$ 10–24 years, $\mathrm{Stage}\mathrm{II}$ 24–34 years, and $\mathrm{Stage}\mathrm{III}$ > 34 years). UAV-LiDAR and in-situ surveys were applied to measure mangrove canopy height to calculate ACG and measure the belowground soil carbon stock, respectively. Combined with the mangrove age stage, ACG, and SOC, our results reveal that ACG accumulates rapidly in younger mangroves dominated by Sonneratia apetala, peaking early (<20 years) and then stabilizing as mangroves, indicating that the introduction of Sonneratia apetala changed the increase in ACG with age. In contrast, SOC increases more gradually over time, with only older mangroves (over 30 years) storing significantly higher SOC. Root structure, TN, and TP were sensitive to the SOC. The different root structures (pneumatophore, plank, pop, and knee root) had different SOC results, and the pneumatophore had the lowest SOC. Remote sensing data revealed that the introduction of Sonneratia apetala altered the species composition of younger mangroves, leading to its predominance within these ecosystems. This shift in species composition not only altered the temporal dynamics of aboveground carbon (ACG) but also favored pneumatophore-dominated root structures, which were associated with the lowest soil organic carbon (SOC). Consequently, younger stands may require more time to accumulate SOC to levels comparable to older mangrove forests. These results suggest that restoration targets for vegetation carbon and soil carbon should be set on different timelines, explicitly accounting for stand age, species composition, and root functional types.

1. Introduction

Mangrove ecosystems are critical carbon sinks that store substantial amounts of carbon in both aboveground biomass (ACG) and belowground soil carbon stocks (SOC), playing a vital role in mitigating climate change [1,2,3]. However, the temporal dynamics of ACG and SOC—i.e., how these carbon pools change over time and interact with forest age, species traits, and environmental factors—remain insufficiently understood. Understanding these dynamics is crucial for optimizing mangrove restoration strategies, particularly in light of increasing restoration efforts globally [4,5]. For instance, fast-growing introduced species such as Sonneratia apetala Buch.-Ham are widely used in restoration projects due to their rapid aboveground biomass accumulation [6,7,8]; however, their long-term effects on ACG and SOC trajectories remain unclear. Without a clear understanding of the mechanisms driving ACG and SOC changes, restoration policies may fail to maximize carbon sequestration or, worse, disrupt natural carbon cycles.

Recent technological advancements, including Landsat satellite imagery and UAV-LiDAR, have provided powerful tools for large-scale and long-term monitoring of mangrove carbon stocks [9,10]. UAV-LiDAR captures high-resolution structural data essential for accurately quantifying ACG [11,12]. Previous studies have demonstrated that structural metrics derived from UAV-LiDAR can retrieve vegetation biomass with higher accuracy than satellite-based remote sensing [12,13]. Landsat, with its multi-decadal archive since the 1980s, enables the estimation of forest age and land cover changes, and numerous studies have used these datasets to assess forest age and evaluate the potential of forest carbon stocks for mitigating climate change [13,14]. Integrating these technologies with in situ measurements offers a robust approach to calculating ACG and dynamics with mangrove age increasing and identifying their driving mechanisms [15,16]. By combining remote sensing data with field surveys, this study investigates how mangrove carbon pools evolve over three forest age stages, focusing on the temporal divergence between ACG and SOC, thereby offering insights to support government efforts in monitoring the spatial and temporal dynamics of mangrove carbon storage.

Aboveground carbon stock (ACG) generally increases with forest age due to biomass accumulation, but this growth is not uniform [17,18]. Young mangroves often exhibit rapid biomass accumulation due to high growth rates, while older forests tend to stabilize as they approach their biomass saturation point [14,19]. Species-specific traits—such as the exceptionally rapid growth of Sonneratia apetala—can substantially modify ACG accumulation patterns [6,7]. As this fast-growing exotic species has been widely replanted, it has gradually become dominant in many restoration areas [6,20], thereby accelerating aboveground carbon (ACG) accumulation and altering the temporal trajectory of ACG compared to older, native mangrove stands. However, it remains unclear whether these species-level differences influence the temporal distribution and dynamics of ACG across different forest stages.

In contrast, soil carbon stock (SOC), which is primarily stored in sediments, is more strongly influenced by soil environmental factors and root traits than by species level differences [21]. Mangrove root systems—by trapping sediments and organic matter—serve as key drivers of SOC accumulation [22,23]. Root architecture varies significantly among species, with shallow, spreading roots and deep, vertical roots influencing both the rate and depth of SOC accumulation [21]. Furthermore, environmental factors such as soil salinity, water-table depth, and microbial activity regulate root growth and organic matter decomposition, thereby shaping SOC dynamics [24,25]. Nevertheless, further research is needed to clarify whether root structural traits and environmental conditions play a more decisive role in SOC regulation than species-level differences.

Recent studies have shown that the sequestration and turnover of ACG and SOC are fundamentally different, resulting in distinct temporal trajectories for these two carbon pools [1,26]. For instance, while ACG may peak and plateau within decades, SOC can accumulate over much longer timescales [24,27,28]. Moreover, the introduction of fast-growing species such as Sonneratia apetala may further amplify this divergence by rapidly altering ACG dynamics while exerting minimal or delayed effects on SOC [6]. These discrepancies underscore the importance of examining the asymmetrical and asynchronous temporal changes in ACG and SOC and their underlying drivers. The asynchronous behavior between aboveground and belowground carbon also suggests that SOC in restored or disturbed mangroves requires substantially longer recovery periods to reach levels comparable to those in natural, mature forests.

In this study, we aimed to reveal the differences in the temporal dynamics of SOC and ACG, as well as the key factors influencing them. We integrated time-series Landsat NDVI data (1990–2024) with national mangrove maps (1990, 2000, 2010, and 2018) to classify mangroves into three age stages (Stage I: 10–24 years; Stage II: 24–34 years; Stage III: >34 years). Aboveground carbon (ACG) was estimated from UAV-LiDAR-derived canopy height, while SOC was quantified through field sampling. We then analyzed the temporal patterns of both ACG and SOC and identified their dominant driving factors, providing insights to guide restoration planning and determine appropriate timescales for achieving ACG and SOC recovery.

We hypothesized:

(1)

The temporal changes in ACG and SOC are asymmetric and asynchronous.

(2)

Species-level differences play a dominant role in determining ACG and root structure, whereas soil environmental factors are the primary drivers of SOC dynamics

2. Methods

2.1. Study Area

Zhanjiang, located in the southernmost region of mainland China, spans from 109°40′ E to 110°58′ E and 20°13′ N to 21°57′ N. The region experiences a tropical monsoon climate with considerable rainfall, receiving between 1500 and 2000 mm of precipitation annually and with an average temperature of about 21 °C [6,26]. Zhanjiang, renowned for its abundant marine resources, features a coastline stretching over 2023.6 km and encompasses a mangrove forest area surpassing 6000 hectares. Within Zhanjiang’s mangrove ecosystems, there exist 28 distinct mangrove species, among which the dominant ones include Avicennia marina (Forssk.) Vierh., Aegiceras corniculatum (L.) Blanco, Sonneratia apetala Buch.-Ham., Kandelia obovata Sheue, H.Y. Liu & J.W.H. Yong, and Rhizophora stylosa Griff. [10].

2.2. Field Sampling

Two field surveys were conducted, in May 2022 and September 2022, covering a total of 60 plots (Figure 1), each measuring 10 × 10 m (

Table 1. In situ survey calculation of mangrove aboveground (ACG) and soil carbon stocks (SOC) for each sampling site in Zhanjiang, China.

Location	Latitude	Longitude	Primary Species	Vegetation			Soil (0–100 cm)
				Plots	Canopy Height (m)	ACG (Mg C/ha)	Plots	Soil Type	SOC (Mg C/ha)
Gaoqiao (GQ)	21°29′–21°35′	109°43′–109°48′	Kandelia obovat, Aegiceras corniculatum, Bruguiera gymnorhiza	10	3.02 ± 1.39	31.73 ± 23.46	12	muddy tidal soils	193.42 ± 61.17
Suixi (SX)	21°24′–21°25′	109°53′–109°56′	Aegiceras corniculatum, Sonneratia apetala, Avicennia marina	3	4.57 ± 2.81	39.6 ± 30.67	6	muddy tidal soils	142.54 ± 32.81
Xiadoulun (XDL)	20°45′–20°48′	109°43′–109°44′	Avicennia marina, Rhizophora stylosa	4	1.27 ± 0.5	22.85 ± 18.94	2	muddy tidal soils	129.84
Luishawan (LS)	20°26′–20°28′	109°56′–109°59′	Avicennia marina, Aegiceras corniculatum, Rhizophora stylosa	4	1.07 ± 0.30	21.83 ± 17.93	7	muddy tidal soils	103.66 ± 16.91
Mazhang (MZ)	20°57′–21°5′	110°8′–110°18′	Aegiceras corniculatum, Sonneratia apetala, Kandelia obovata, Rhizophora stylosa	6	4.92 ± 4.78	41.38 ± 40.66	5	muddy tidal soils	79.22 ± 27.70
Techengdao (TCD)	21°8′–21°9′	110°25′–110°26′	Avicennia marina, Kandelia obovata	3	2.69 ± 0.27	30.06 ± 17.77	6	silty–sand	136.61 ± 39.05
Lengzhou (LZ)	20°49′–20°58′	110°8′–110°18′	Aegiceras corniculatum, Sonneratia apetala	3	4.71 ± 5.26	40.31 ± 43.10	5	muddy tidal soils	45.25 ± 21.57
Wailuowan (WLW)	20°33′–20°43′	110°18′–110°28′	Avicennia marina, Kandelia obovata, Aegiceras corniculatum, Sonneratia apetala	27	2.36 ± 1.88	28.38 ± 25.94	12	muddy tidal soils	125.12 ± 53.90

), distributed across eight areas in Zhanjiang. During these surveys, detailed data on mangrove species and tree height were collected. Although the soil survey was conducted at the same time and in the same locations as the vegetation survey, the number of vegetation quadrats exceeded that of soil quadrats (55 samples) due to the labor-intensive process of collecting soil samples, which followed procedures outlined in previous studies [5,14]. Within each area, 2 to 4 plots, each measuring 5 m × 5 m, were selected to represent the primary mangrove communities in that specific area. A total of 55 plots were chosen across the 8 areas, with their respective geographical coordinates (longitude and latitude) and species names meticulously recorded. The primary mangrove communities identified included Avicennia marina, Aegiceras corniculatum, Sonneratia apetala, Kandelia obovata, Rhizophora stylosa, and Bruguiera gymnorhiza. All the geographical coordinates of soil and vegetation plots were monitored using the handheld GPS (UniStrong).

During low tide, we collected three 1 m soil cores per plot using a 5 cm (diameter) stainless-steel corer. We avoided respiratory roots, crab burrows, and trampling traces and removed the surface apoplastic layer and living tissue prior to coring. To correct for compaction, we recorded both the core length and penetration depth, calculated a compression coefficient (core length ÷ penetration depth), and applied this correction factor to each depth interval [29]. In the field, soil cores were sectioned into eight depth intervals: 0–5, 5–10, 10–20, 20–30, 30–45, 45–60, 60–80, and 80–100 cm [30]. All soil cores were used as an undisturbed profile sample. Samples were placed in zip-top polyethylene bags, kept at −4 °C, and transferred to a −20 °C laboratory freezer within four hours of collection. Samples were sealed in zip-lock bags, chilled at −4 °C, and transferred to a −20 °C freezer within four hours.

2.3. Data Preparation

2.3.1. Collecting the UAV-LiDAR and Extracting the Canopy Height

The LiDAR data were collected during the two-field surveys encompassing 60 sampling plots. To ensure data quality, all flights were conducted during low tide under clear and windless weather conditions. A Zenmuse L1 LiDAR sensor (DJI, Shenzhen, China) mounted on a DJI UAV was used, with a flight altitude of approximately 50 m. The raw LiDAR data were processed in DJI Terra (DJI, Shenzhen, China) to generate point clouds, with the point cloud density set to High and the elevation metric defined as the 95th percentile. All outputs were projected to the WGS-84 coordinate system. Canopy height was subsequently derived from the LiDAR point clouds following standard procedures described in previous studies [10,16,31,32]. Previous studies have demonstrated that canopy height derived from UAV-LiDAR data is highly consistent with field-measured canopy height, with strong correlations and low estimation errors [10,31]. To improve data quality, a hierarchical denoising workflow was applied before subsequent analysis using LiDAR 360 (version 5.4). This process combined statistical outlier removal (1σ threshold within a 1 m neighborhood) with morphological filtering. After preprocessing, a 10 m Digital Surface Model (DSM) was generated by extracting the highest LiDAR returns [33]. A Digital Elevation Model (DEM) was then derived by separating ground and non-ground points with the improved progressive TIN densification algorithm [34], supplemented with manual adjustments, also at 10 m resolution. Finally, the Canopy Height Model (CHM) was obtained by subtracting the DEM from the DSM, where each pixel represents mangrove canopy height [16]. To assess accuracy, we compared UAV-LiDAR-derived canopy height (CH) with in situ measurements, confirming the robustness of the UAV-LiDAR estimates (Figure A1).

2.3.2. Auxiliary Data

For analysis of the relationship between climate factors and mangrove carbon stock, ECMWF/Copernicus Climate Change Service (2 km) was used to obtain climate factors. According to Meng et al. [35] and Lu et al. [36], air temperature (Tem), precipitation (Preci), and vapor pressure deficit (VPD) were important factors in the carbon fluxes of China’s coastal wetlands. Those climate factors were chosen and calculated for the mean value of 2000–2023 (Tem, Preci, and VPD) on the Google Earth Engine platform.

HSL_MangroveChina_LASAC_share provides the spatial distribution of China’s mangrove resources in Shapefile format across five epochs (1978, 1990, 2000, 2013, and 2018). The 2018 mangrove map was produced from domestic 2 m satellite imagery (including ZY-3 01/02, GF-1, and GF-1 B/C/D) using a combined workflow of automated computer interpretation, expert refinement, and field verification, achieving an overall accuracy of 98% [37,38].

The Landsat NDVI time series (1990–2024) and the China mangrove dataset (HSL_MangroveChina_LASAC_share) were used to estimate mangrove age stage (

Table 2. Datasets used in this study.

Metadata	Type	Description	Reference
HSL_MangroveChina_LASAC_share	ESRI shapefile	The China mangrove distribution of 1978, 1990, 2000, 2010, 2018 (www.sasclouds.com/chinese/platform/newsList/notic/detail/618cc900fd423278867c5dda (accessed on 5 May 2025))	[37,38]
The classifiction of speceies mangrove in China	ESRI shapefile	Using Synthesized Sentinel-2 High-Separability Images and mapping mangrove species	[39,40]
UAV-LiDAR	Raster (resampled 10 m)	Using majority for resampling 1 m to 10 m
NDVI Time series from Landsat	raster (30 m)	1990–2024 for screening and estimation mangrove age stage.
ECMWF/Copernicus Climate Change Service	Raster (2 km)	For extracting climate factors

). We utilized the Google Earth Engine (GEE) platform’s full archive to compile stacks of Surface Reflectance Tier 1 Landsat TM/ETM+/OLI images spanning from 1990 to 2024. To reduce phenological and solar-geometry effects, images were restricted to summer (1 June–1 September) across Zhanjiang. We then generated annual median composites with minimal cloud contamination and computed NDVI from each composite to track vegetation dynamics from 1990 to 2024.

To understand the composition of mangrove species across different age stages, a spatial map of mangrove species is essential (Table 2). Spatial distribution data of mangrove species in China (10 m) were utilized to support the analysis, which was extracted from synthesized Sentinel-2 images [20,39,40]. The spatial distribution of mangroves in China was originally classified into ten types, five of which—Kandelia obovata, Aegiceras corniculatum, Sonneratia apetala, Rhizophora stylosa, and Avicennia marina—were consistent with our study. For subsequent analysis, we extracted those five mangrove species: Kandelia obovata, Aegiceras corniculatum, Sonneratia apetala, Rhizophora stylosa, Avicennia marina.

2.4. Calculations of the Aboveground Vegetation and Soil Carbon Stocks

2.4.1. Calculation of the Aboveground Vegetation Carbon Stocks

To estimate aboveground carbon stock (ACG), we derived canopy height (CH) from UAV-LiDAR data, following the methodology described by Maimaitijiang et al. [9] and Lu et al. [31], and we applied a commonly used allometric growth equation [41,42]. This involved calculations of aboveground biomass for each plant within the sampled quadrats. Subsequently, the total aboveground biomass for each quadrat was determined by aggregating the individual plant biomass values. This approach enabled us to account for the varied characteristics of mangrove species and conduct a comprehensive assessment of biomass across the study area.

$$ACG=\left(10.8\times h_i+34.9\right)\times 0.47$$

(1)

$ACG$ is the vegetation carbon stock, $h_i$ is the canopy height from UAV-LiDAR, and 0.47 is the transform parameter from aboveground vegetation biomass to the vegetation carbon stock, according to Meng et al. (2022) [35]. Within each area, a total of 60 plots were chosen across the 8 areas.

2.4.2. Laboratory Analysis

The soil samples were subjected to oven-drying at 65 °C until reaching a constant weight [43,44]. Following the removal of plant debris and shells, the soil samples were ground and sieved through a 0.2 mm sieve for the determination of Total Nitrogen (TN), Total Phosphorus (TP), salinity, and organic carbon content. Soil salinity was assessed using a YSI-proplus multi-probe sensor (YSI Incorporated, Yellow Springs, OH, USA) with a soil-to-deionized water ratio of 1:5 (w/v). TP was determined through H₂SO₄-HClO₄ digestion and analyzed using a UV-1900i spectrophotometer (Shimadzu (Hong Kong) Ltd., Hong Kong, China). TN was determined via Kjeldahl digestion using an A AA3 continuous flow analyzer (SEAL Analytical, Norderstedt, Germany) [7]. Soil bulk density (BD) was calculated based on the mass of the completely dried sample and the soil volume using the formula:

BD (g cm⁻³) = Dry weight of soil (g)/Soil volume (cm³)

(2)

Soil volume = 3.14 × (soil auger radius² (cm)× soil thickness (cm))

(3)

Soil organic carbon content was analyzed using a Vario EL III elemental analyzer (Elementar Analysensysteme GmbH, Hanau, Germany; C% accuracy: ±0.5%). The soil carbon stock was computed by multiplying BD by the soil carbon content and then by the corresponding corrected soil depth, as per the formula proposed by Kauffman and Donato [45]:

SOC = BD (g cm⁻³) × Carbon content (%) ×corrected soil column (cm)

(4)

The temporal dynamics of soil carbon stocks were found to be significant within the 0–60 cm soil layer [8]. In this study, we first calculated SOC for eight individual depth intervals and then aggregated them into five standardized layers (0–20, 20–40, 40–60, 60–80, and 80–100 cm) to analyze temporal variations. The raw data are presented in

Table A1. Summary of soil carbon storage in different layers of mangrove forests at different forest ages.

Age Stage	0–5 cm (Mg C/ha)	5–10 cm (Mg C/ha)	10–20 cm (Mg C/ha)	20–30 cm (Mg C/ha)	30–45 cm (Mg C/ha)	45–60 cm (Mg C/ha)	60–80 cm (Mg C/ha)	80–100 cm (Mg C/ha)
$\mathrm{Stage}\mathrm{I}$	4.88 ± 2.63	5.53 ± 3.23	10.78 ± 5.29	11.19 ± 5.22	17.35 ± 11.18	16.93 ± 11.88	25.38 ± 17.21	20.72 ± 14.75
$\mathrm{Stage}\mathrm{II}$	6.48 ± 3.14	6.78 ± 4.02	13.25 ± 4.98	12.86 ± 5.33	18.86 ± 13.63	16.37 ± 8.54	17.81 ± 9.44	15.18 ± 9.64
$\mathrm{Stage}\mathrm{III}$	8.92 ± 4.61	8.72 ± 3.41	17.6 ± 8.36	18.81 ± 8.02	25.89 ± 10.99	25.1 ± 8.33	30.08 ± 15.08	24.85 ± 14.52

and

Table A2. Summary of soil carbon storage in different mangrove species at different layers.

Species	0–5 cm (Mg C/ha)	5–10 cm (Mg C/ha)	10–20 cm (Mg C/ha)	20–30 cm (Mg C/ha)	30–45 cm (Mg C/ha)	45–60 cm (Mg C/ha)	60–80 cm (Mg C/ha)	80–100 cm (Mg C/ha)
Sonneratia apetala	4.02 ± 0.56	4.2 ± 1.65	10.92 ± 1.46	7.67 ± 1.32	13.76 ± 3.54	11.08 ± 3.23	11.02 ± 4.67	11.06 ± 3.42
Avicennia marina	4.77 ± 2.82	5.41 ± 3.06	10.26 ± 4.92	10.73 ± 4.11	12.7 ± 4.96	13.64 ± 5.15	16.31 ± 8.83	11.76 ± 7.43
Aegiceras corniculatum	6.95 ± 2.4	7.56 ± 2.64	14.91 ± 4.24	15.22 ± 4.24	24.13 ± 7.59	23.87 ± 8.07	29.72 ± 12.73	24.62 ± 11.58
Rhizophora stylosa	13.35 ± 2.46	9.63 ± 2.38	24.21 ± 10.53	29.42 ± 3.05	33.68 ± 8.32	30.05 ± 7.06	39.55 ± 12.91	36.21 ± 16.05
Bruguiera gymnorrhiza	13.84 ± 3.28	13.31 ± 1	26.12 ± 4.84	25.03 ± 5.51	39.84 ± 8.1	36.76 ± 5.04	47.79 ± 13.51	38.83 ± 6.26
Kandelia obovata	11.57 ± 1.79	12.84 ± 0.25	20.19 ± 1.79	22.56 ± 1.96	40.9 ± 10.48	33.07 ± 3.01	41.09 ± 2.53	37.41 ± 2.83

.

2.5. Extracting the Age Stage of Mangroves

Combined with the long-term NDVI time series and the HSL_MangroveChina_LASAC_share, we extracted mangrove age information. Understanding the influence of the introduction of exotic species—Sonneratia apetala—on the temporal dynamic of ACG and SOC is the main purpose of our study. Previous studies reported that Sonneratia apetala was first introduced into China in the 1990s and was later incorporated into mangrove replantation programs in the 2000s [46,47]. By the 2010s, several studies had reported that Sonneratia apetala possesses a more tolerant and invasive niche than native mangrove species [48,49,50]. To define an appropriate temporal window for assessing changes in ACG and SOC, we drew on published estimates of carbon stocks for dominant mangrove species across age stages (Table 3). For Sonneratia apetala, aboveground biomass can exceed ~180 Mg C·ha⁻¹ within approximately 15 years and remains at a comparable level by ~40 years, indicating an early growth plateau.

Briefly, we chose the spatial distribution of mangroves in 1990, 2000, 2013, and 2018 to filter three stages of mangrove vegetation. To identify three distinct mangrove age stages, we used historical proxy mangrove classification datasets (1990, 2000, 2013, and 2018) with high overall accuracy and determined their intersections across time periods. Long-term NDVI imagery was then applied to further validate and refine the classification, ensuring that the three mangrove stages corresponded accurately to their respective temporal periods. Previous studies have demonstrated that an NDVI threshold of 0.63 can effectively differentiate dwarf shrub mangroves from mature mangrove stands, making it a sensitive indicator for species or structural classification [51]. This conservative cutoff minimizes the risk of overestimating stand age, particularly in areas dominated by Sonneratia apetala, a fast-growing species [6,50]. Specifically, Stage III, Stage II, and Stage I were signals of the pixels with mangrove ages over 34, 24, and 11 years, respectively. Those different stages of mangrove vegetation were filtered based on three equations:

$$\mathrm{Stage}\mathrm{III}\in A_{1990}\cap A_{2000}\cap A_{2018}\mathrm{And}{NDVI}_{\mathrm{Stage}\mathrm{III}}^t>0.63(1990\le t\le 2024)$$

(5)

$$\mathrm{Stage}\mathrm{II}\in (A_{2018}-\mathrm{Stage}\mathrm{III})\cap A_{2000}\cap A_{2018}\mathrm{And}{NDVI}_{\mathrm{Stage}\mathrm{II}}^t>0.63(2000\le t\le 2024)$$

(6)

$$\mathrm{Stage}\mathrm{I}\in {(A}_{2018}-\mathrm{Stage}\mathrm{III}-\mathrm{Stage}\mathrm{II})\cap A_{2013}\mathrm{And}{NDVI}_{\mathrm{Stage}\mathrm{I}}^t>0.63(2013\le t\le 2024)$$

(7)

$A_{1990},A_{2000},A_{2013,}\mathrm{and}$ $A_{2018}$ is the mangrove area in 1990, 2000, 2013, and 2018, respectively, and ${NDVI}_{\mathrm{Stage}\mathrm{III}}^t$, ${NDVI}_{\mathrm{Stage}\mathrm{II}}^t$ and ${NDVI}_{\mathrm{Stage}\mathrm{I}}^t$ are signals of the pixel’s NDVI at different stages.

Table 3. The aboveground vegetation biomass and belowground vegetation biomass and soil carbon stock at different ages.

Location	Latitude	Types	Age	Aboveground Vegetation Biomass (Mg C/ha)	Belowground Vegetation Biomass (Mg C/ha)	Soil Carbon Stock (Mg C/ha)	Reference
Qiao Island	22° N	S. apetela (SA)	1	1.04	0.13	77.09
			4	97.66	22.46	81.62
			9	146.87	39.42	85.2	[6]
			15	184.88	47.94	101.22
			40	194.66	58.39
Leizhou Bay	20–21° N	SA	5	12.81 (AGB + BGB)		68.81
			13	65.04 (AGB + BGB)		75.12	[52]
			20	65.48 (AGB + BGB)		89.86
Futian Bay	22.5°	SA	6	144	86	44
			20	258	58	48	[53]
Malaysia	2°06′ N	natural	3	37.8	42.05
			12	66.01	32.81
			25	75.61	32.64		[54]
		A. marina (AM)	6	69.74	38.25
			12	66.01	32.81
Qiao Island	22° N	SA	4	99.42	24	92.06
		SA	9	147.42	37.71	92.66	[55]
		SA	15	212.57	53.71	101.08
		SA	40	188.57	87.99	170.88
Yingluo Bay	21° N	AM	15	29.64	21.34	150.41	[56]
		AM	45	84.17	34.3	234.45
Leizhou, Guangdong	109°03′ E, 20°30′ N	SA	4	16.3	3.6	27.7 (0–20 cm)	[57]
			10	37.2	11. 8	67.5 (0–20 cm)
Gaoqiao, Guangdong	21°30′ N	AM	5	16.4 (AGB + BGB)
		Bruguiera gymnorhiza (BG)	10	41.4 (AGB + BGB)			[58]
		BG	30	96.1 (AGB + BGB)

Table 3. The aboveground vegetation biomass and belowground vegetation biomass and soil carbon stock at different ages.

Location	Latitude	Types	Age	Aboveground Vegetation Biomass (Mg C/ha)	Belowground Vegetation Biomass (Mg C/ha)	Soil Carbon Stock (Mg C/ha)	Reference
Qiao Island	22° N	S. apetela (SA)	1	1.04	0.13	77.09
			4	97.66	22.46	81.62
			9	146.87	39.42	85.2	[6]
			15	184.88	47.94	101.22
			40	194.66	58.39
Leizhou Bay	20–21° N	SA	5	12.81 (AGB + BGB)		68.81
			13	65.04 (AGB + BGB)		75.12	[52]
			20	65.48 (AGB + BGB)		89.86
Futian Bay	22.5°	SA	6	144	86	44
			20	258	58	48	[53]
Malaysia	2°06′ N	natural	3	37.8	42.05
			12	66.01	32.81
			25	75.61	32.64		[54]
		A. marina (AM)	6	69.74	38.25
			12	66.01	32.81
Qiao Island	22° N	SA	4	99.42	24	92.06
		SA	9	147.42	37.71	92.66	[55]
		SA	15	212.57	53.71	101.08
		SA	40	188.57	87.99	170.88
Yingluo Bay	21° N	AM	15	29.64	21.34	150.41	[56]
		AM	45	84.17	34.3	234.45
Leizhou, Guangdong	109°03′ E, 20°30′ N	SA	4	16.3	3.6	27.7 (0–20 cm)	[57]
			10	37.2	11. 8	67.5 (0–20 cm)
Gaoqiao, Guangdong	21°30′ N	AM	5	16.4 (AGB + BGB)
		Bruguiera gymnorhiza (BG)	10	41.4 (AGB + BGB)			[58]
		BG	30	96.1 (AGB + BGB)

2.6. Data Analysis

ACG and SOC were quantified using UAV-LiDAR data and in situ measurements. By integrating long-term NDVI imagery with the HSL_MangroveChina_LA-SAC_share dataset, we further estimated mangrove age stages and revealed the temporal dynamics of both ACG and SOC. Finally, structural equation modeling (SEM) was applied to identify the key drivers of SOC (Figure 2).

The Standard Deviational Ellipse was used to detect the temporal–spatial change trend of the mangrove area on ArcMap 10.3. The mean center was applied to demonstrate the temporal change of different mangrove stages [59]. To compare mangrove canopy height and soil carbon stocks, we performed a one-way ANOVA, and statistically significant results (p < 0.05) were further examined using LSD post-hoc tests. Results are presented as mean values with error bars indicating standard deviation (SD).

Structural equation modeling (SEM) was used to quantify the direct and indirect effects of eight primary drivers (TN, TP, SA, Stage, Root, VPD, Temp, and Pre) on SOC and ACG. In the SEM framework, Root and Stage represented species functional type and stand age, respectively. Models were implemented using the psem function in the ‘piecewiseSEM’ package (R version 4.3.3 (R Core Team, Vienna, Austria), with generalized least squares (GLS) employed to account for spatial autocorrelation among age stages and to improve model robustness [60,61]. The best-fitting model was identified as the one with the lowest AIC and a statistically significant chi-square test (p < 0.05) [62,63].

3. Results

3.1. ACG and SOC Based on Field Survey

Our findings revealed distinct variations in the spatial distribution of mangrove canopy height (CH) in Zhanjiang (Figure 2). Specifically, mangroves along the western coast of Zhanjiang City exhibited a mean CH of 3.67 m, whereas those in the eastern coastal zone had a lower mean CH of 2.48 m. The spatial distribution of aboveground carbon stock (ACG) showed a similar pattern. Mazhang had the highest ACG (41.38 Mg C/ha), while Luishawan recorded the lowest value (21.83 Mg C/ha). In contrast, the spatial pattern of soil carbon stock (SOC) differed from that of ACG. Gaoqiao exhibited the highest SOC (193.42 Mg C/ha), whereas Leizhou showed the lowest (45.25 Mg C/ha). Overall, SOC was several times greater than ACG across all regions. At the species level, Sonneratia apetala had the highest ACG (52.45 Mg C/ha) but the lowest SOC (79.87 Mg C/ha). Other species displayed no significant differences in ACG, while Bruguiera gymnorhiza and Kandelia obovata exhibited relatively high SOC levels (Table 1 and Figure 3).

3.2. Remote-Sensing-Based Extraction of Forest Age Stages and Species Composition

Based on the classification of three mangrove age stages, we analyzed the spatiotemporal changes in mangrove distribution across Zhanjiang (Figure 4). The areas of Stage III, Stage II, and Stage I mangroves were 1259.45 ha, 3763.05 ha, and 630.27 ha, respectively. Stage III mangroves were primarily distributed in the northwest of Zhanjiang (around GQ) and the central region (around TCD). Stage II mangroves were mainly located in the western (around SX) and southeastern parts (around LZ), whereas Stage I mangroves were concentrated in the central and southeastern regions (around WLW). The centroid of mangrove distribution gradually shifted from northern Zhanjiang toward the middle-eastern region (around LZ). Overall, older mangroves were concentrated in the northwest (GQ) and central (TCD) areas, while younger mangroves were mainly found in the southeast (WLW and LZ). These results are consistent with previous studies: Wang et al. (2024) [16] also reported that mangroves in GQ were in an older developmental stage; Huang et al. (2024) [7] found that the mangroves in TCD were stable and mature; and Zhao et al. [39] identified Sonneratia apetala—an exotic species introduced around 2000—in the LZ and WLW regions, confirming that these areas represent younger mangrove stands (<34 years).

By combining the spatial distribution of mangrove species with forest age stages, we analyzed the area occupied by five species—Kandelia obovata, Aegiceras corniculatum, Sonneratia apetala, Rhizophora stylosa, and Avicennia marina—across different successional stages. According to Figure 5, species composition varied significantly across stages. In the old-growth stage (Stage III), Sonneratia apetala accounted for only 9% of the total area, while native species such as Rhizophora stylosa, Aegiceras corniculatum, and Avicennia marina dominated. In contrast, Sonneratia apetala increased sharply in the younger stages (Stage I and Stage II), emerging as the second most dominant species. The area occupied by Avicennia marina in the younger stages was also greater than that in the old-growth stage. Overall, Sonneratia apetala and Avicennia marina were the dominant species in younger mangroves, whereas the native species Avicennia marina, Aegiceras corniculatum, and Rhizophora stylosa were more dominant in older stands. As shown in Figure 6, notable differences in root-type composition were observed between younger (Stages I and II) and older mangroves (Stage III). Younger stands exhibited a higher proportion of pneumatophores compared with older mangroves.

3.3. Temporal Dynamics in the Aboveground and Soil Carbon Stocks of Mangroves

Based on the classification of mangrove age stages, significant differences were observed among the three stages (Figure 7). Canopy height (CH) showed a clear negative relationship with forest age, with Stage I exhibiting the highest CH, followed by Stage II and Stage III. Because aboveground carbon stock (ACG) was derived from CH, its temporal pattern across age stages was consistent with that of CH.

Mangrove age exerted a significant effect on SOC in the upper soil profile (0–60 cm), whereas deeper layers (60–100 cm) showed no age-dependent differences (Figure 8). Stage III mangroves stored consistently higher SOC than Stage I and Stage II, whereas SOC in younger stands (<30 years) showed no significant differences between stages.

3.4. Driving Factors of Temporal Variations in the Aboveground and Soil Carbon Stocks

Based on the results of the in situ survey results, Sonneratia apetala had the highest ACG density (Table 1 and Figure 3). Figure 5 and Figure 6 indicate that Sonneratia apetala was the predominant species in younger mangroves (Stage I and Stage II), while it accounted for only a minor proportion in older stands (Stage III). Zan et al. [47] found that S. apetala adapts well to coastal environments in southern China, including the Shenzhen coastline near Zhanjiang, where it can reach heights of up to 10 m. As a non-native species, S. apetala was widely introduced in the 1990s for mangrove restoration due to its rapid growth and high environmental tolerance [44,47], and it has since become increasingly prevalent along China’s coastline [46]. Taken together, these findings suggest that the widespread introduction of S. apetala has enabled younger mangroves to accumulate greater ACG than older stands, resulting in the observed negative relationship between ACG and mangrove age stage. We further analyzed the influence of mangrove species and age on the ACG using SEM (Figure 9) and found that the forest age had a significant impact on ACG, whereas species identity did not play a major direct role. This finding suggests that most native mangrove species exhibit similar ACG values, consistent with the patterns observed in Figure 3. Overall, the gradual expansion of Sonneratia apetala over the past two decades has reshaped the temporal trajectory of ACG due to its high tolerance and rapid growth.

We applied structural equation modeling (SEM) to identify the primary drivers of SOC dynamics. Root structure was the most influential factor affecting surface SOC (0–5 cm), with a strong positive path coefficient (0.68, p < 0.01). Forest age (Stage), TN, TP, SA, and species identity also exhibited significant positive correlations with surface SOC. At the 0–100 cm profile, root structure and TN remained significantly positively correlated with SOC, whereas TP showed a significant negative correlation (Figure 9). Across all soil depths, mangrove species dominated by pneumatophores had the lowest SOC. At the 20–60 cm depth interval, SOC differed significantly among the three other root types (Figure 10). These results indicate that differences in root structure restrict carbon accumulation in species with pneumatophores, thereby amplifying temporal variation in SOC among mangrove communities.

4. Discussion

By integrating in situ surveys with UAV-LiDAR and long-term NDVI series, we identified significant asymmetry and asynchrony in the temporal dynamics of aboveground carbon (ACG) and soil organic carbon (SOC), reflecting distinct sequestration mechanisms in the above- and belowground components. The introduction of Sonneratia apetala, a fast-growing exotic species, accelerated and reshaped the temporal trajectory of ACG, whereas SOC accumulation was mainly regulated by root structure and soil nutrient availability. Notably, SOC in younger mangroves (10–24 and 24–34 years) did not increase substantially, suggesting that soil carbon restoration requires longer time scales to reach levels comparable to those in older stands (>34 years). RS further revealed that the introduction of S. apetala altered species composition in younger mangroves and reshaped the temporal dynamics of ACG. This shift was accompanied by a transition in root structure toward pneumatophores, which contributed to slower SOC accumulation. As a result, younger stands will require more time to develop soil carbon stocks equivalent to those of mature forests. These findings highlight the fundamental differences between aboveground and belowground carbon sequestration processes and underscore the critical roles of species composition, root architecture, and environmental factors in regulating mangrove carbon dynamics.

Our study shows clear asymmetry and asynchrony in the temporal trajectories of ACG and SOC, revealing distinct restoration time lags between these pools. The rapid growth of Sonneratia apetala accelerates ACG turnover and can produce ACG levels comparable to older native stands within <20 years. In contrast, SOC recovery proceeds much more slowly: the soil carbon pool requires substantially longer to approach that of older native mangroves and may remain lower within the several frames of typical restoration projects. Yu et al. [6] also found that the ACG of younger Sonneratia apetala was higher than older native mangroves, while the SOC of Sonneratia apetala was lower than older native mangroves. The younger mangrove (<30 years) had a similar SOC, while the older mangrove (>40 years) had a higher SOC than the younger mangrove. These findings highlight the need for restoration strategies to explicitly prioritize SOC recovery and to establish long-term baselines and monitoring frameworks that can effectively track belowground carbon trajectories.

The differing sensitivities of ACG and SOC indicate distinct mechanisms of above- and belowground carbon sequestration in mangroves. ACG can accumulate rapidly when highly competitive species are introduced, as they capture resources more efficiently to support fast growth. The sequestration of soil carbon was a complex mechanism, since the soil carbon pool is supplied by autochthonous inputs (aboveground litter and root production) and by allochthonous carbon delivered via tidal exchange and lateral water movement from adjacent ecosystems [54,55]. Using structural equation modeling (SEM), we further identified root structure as the most significant factor influencing SOC, with environmental variables such as total nitrogen (TN) and total phosphorus (TP) also playing key roles, which was consistent with previous studies [7,21,64,65]. These results suggest that effective SOC restoration should explicitly consider belowground processes (e.g., root production) and site-specific soil conditions.

Sonneratia apetala could rapidly accumulate ACG due to its special growth strategy, compared to other native species. Sonneratia apetala exhibits rapid early structural development, high specific leaf area (SLA), and elevated leaf nitrogen content, coupled with relatively low to moderate wood density [66,67]. Such traits promote fast canopy closure, high photosynthetic rates, and rapid ACG accumulation in the first two decades after planting, as shown in Table 3. In contrast, native species generally possess higher wood density and greater leaf construction costs [68,69], resulting in a more conservative growth strategy and slower biomass accumulation.

The RS data revealed clear alterations in mangrove species composition, indicating a shift toward the dominance of Sonneratia apetala. Zhao et al. [40] similarly reported that S. apetala is now the predominant mangrove species in Zhanjiang. However, the introduction and widespread colonization of this species may reduce the soil carbon sequestration capacity of mangrove ecosystems. Its gradual dominance should therefore be viewed as a potential early warning signal of ecological degradation. Moreover, the transition toward pneumatophore-dominated root systems appears to further weaken belowground carbon storage potential. Although species-specific differences in SOC were not directly detected in Zhanjiang, previous studies have shown that stands dominated by S. apetala can exhibit lower SOC than native communities [8]. Nevertheless, it is premature to conclude that S. apetala inherently suppresses soil carbon sequestration without direct evidence on SOC turnover (e.g., repeated depth-resolved coring, root production and decomposition rates, or δ¹³C/¹⁴C and biomarker constraints), particularly given that SOC dynamics are strongly age-dependent. In contrast, mixed native communities with greater functional trait diversity generally exhibit higher SOC levels [45]. Because Sonneratia apetala introductions can simplify community composition and homogenize root architectures, they may reduce belowground functional heterogeneity, thereby dampening SOC sequestration and increasing ecological vulnerability. Taken together, these considerations argue for diversity-oriented restoration strategies and long-term, process-based monitoring of soil carbon.

5. Conclusions

Our study highlights the asymmetry and asynchrony in the temporal dynamics of ACG and SOC in mangrove ecosystems. ACG accumulates rapidly in younger mangroves dominated by Sonneratia apetala but plateaus with forest maturity, while SOC increases more gradually over time due to root structure and environmental factors. Our findings emphasize the need for stage-specific restoration strategies that enhance SOC through native species while managing the rapid ACG growth in younger mangroves. Future research should refine temporal analyses to identify critical transition points and optimize strategies for sustaining long-term carbon storage. In particular, quantitative measurements of SOC and ACG turnover rates, as well as stand age, are essential for pinpointing key time windows and informing rational planning for mangrove restoration.