Temporal and Spatial Patterns of Blue Carbon Storage in Mangrove and Salt Marsh Ecosystems in Guangdong, China

Abstract

Coastal blue carbon ecosystems serve as vital carbon sinks in global climate regulation, yet their long-term carbon storage dynamics remain poorly quantified at regional scales. This study quantified the spatiotemporal evolution of mangrove and salt marsh carbon storage in Guangdong Province, China, over three decades (1986–2020), by integrating a new mangrove and salt marsh detection framework based on Landsat image time series and the Integrated Valuation of Ecosystem Services and Tradeoffs (InVEST) model. The proposed detection framework provided two coastal vegetation detection methods, exploring the potential of utilizing phenological features to improve the mangrove and salt marsh discrimination accuracy with Landsat data. The overall accuracies of both mangrove and salt marsh detection results exceeded 90%, suggesting good consistency with the validation data. The mangrove extent showed a trend of decreasing from 1986 to 1995, then fluctuated from 1995 to 2005, and presented an upward trend from 2005 to 2020. The overall trend of the salt marsh area was upward, with small fluctuations. The mangrove carbon storage in Guangdong increased from 414.66 × 10⁴ Mg C to 490.49 × 10⁴ Mg C during 1986–2020, with Zhanjiang having the largest mangrove carbon storage increase. The salt marsh carbon storage in Guangdong grew from 8.73 × 10⁴ Mg C in 1986 to 14.39 × 10⁴ Mg C in 2020, with Zhuhai as the salt marsh carbon sequestration hotspot. The temporal dynamics of carbon storage in mangroves and salt marshes could be divided into three stages, namely a decreasing period, a fluctuating period, and a rapid increase period, during which ecological and economic policies played a crucial role. The multi-decadal blue carbon datasets and their temporal-spatial change analysis results here can provide a scientific basis for nature-based climate solutions and decision-support tools for carbon offset potential realization and sustainable coastal zone management.

1. Introduction

Coastal blue carbon ecosystems (mangroves, tidal salt marshes, and seagrass beds) play a pivotal role in climate change mitigation [1,2,3]. They excel at marine carbon sequestration mechanisms: anaerobic sediment conditions that slow down decomposition, tidal sediment deposition that buries organic matter, high plant productivity through photosynthesis [4,5], etc. As efficient blue carbon sinks, these coastal ecosystems provide multiple ecological services [6] for climate change adaptation [2,7,8], as well as for the well-being and health of coastal communities [9]. However, rapid urbanization [10,11], land reclamation [12,13], and climate change [14,15] have caused extensive degradation of these ecosystems, thereby giving rise to substantial carbon emissions in coastal regions [16]. In 2020, China announced its plan to strengthen its Nationally Determined Contributions (NDCs) to achieve carbon neutrality by 2060 [17]. And the “Dual Carbon” strategy (carbon peaking and neutrality) [18] has highlighted the urgency of protecting and restoring blue carbon ecosystems as a nature-based solution to achieve climate goals [2,19]. Investigating the spatiotemporal characteristics of carbon storage is crucial for understanding the long-term evolution of blue carbon ecosystems, contributing to NDCs, and fulfilling international blue carbon commitments.

Traditional methods mainly utilize in situ field data to estimate carbon storage with high precision [20,21,22], but they require substantial labor and material resources, limited to site-scale research [23,24,25,26]. With the development of remote sensing technology, researchers explored the potential of remote sensing and model simulation to monitor carbon storage dynamics [27,28,29,30,31]. Fu et al. (2019) investigated the spatiotemporal dynamics of terrestrial ecosystem carbon storage in the Su–Xi–Chang region, China, with the Integrated Valuation of Ecosystem Services and Tradeoffs (InVEST) model [32]. Kafy et al. (2023) examined the spatiotemporal patterns of carbon storage in forests using the InVEST model and the generated land use/land cover maps from Landsat satellite imagery [31]. Li et al. (2023) applied the InVEST-PLUS model to analyze the pattern of terrestrial ecosystem carbon storage in Liaoning Province, China [27]. Deng et al. (2022) assessed carbon storage in wetlands of the Guangdong–Hong Kong–Macau Greater Bay Area, China, using available land use maps and the InVEST model [21]. Li et al. (2022) investigated the dynamics of carbon storage in salt marshes across the eastern coastal wetlands in China by combining remote sensing and the InVEST model [33]. Most studies focused on quantifying terrestrial carbon storage, while overlooking the analysis of temporal dynamics in blue carbon ecosystem carbon stocks in Guangdong Province, China.

Phenological information is regarded as an important feature for the wetland vegetation mapping [34,35,36,37]. Researchers [34,35] employed satellite imagery in two phenological periods (green-up season and senescence season) to map coastal salt marshes. The phenological characteristics of mangroves and salt marshes are different [38]. Mangroves are evergreen forests, while salt marshes are deciduous. Zhang et al. (2022) proposed a method that integrated tide-level and phenological features, and simultaneously generated maps of salt marshes, mangroves, and tidal flats in East Asia in 2020 using Sentinel-2 satellite image time series [37]. However, the Sentinel-2 mission, which has been operational since 2015, has a higher revisit frequency, compared to the Landsat missions (launched in 1972). Moreover, the periodic tidal inundations and frequent cloudy weather along coastlines make it challenging to monitor blue carbon ecosystems with Landsat imagery. In this study, we explored the potential of the green-up season and senescence season image composite method [34] for accurate mangrove and salt marsh mapping with Landsat image time series.

Blue carbon storage monitoring is important for advancing the Sustainable Development Goals (SDGs), particularly SDG 13 (Climate Action), SDG 14 (Life Below Water), and SDG 15 (Life on Land). Based on carbon stock estimation results, coastal managers can quantify emission reductions from conservation efforts, and align the carbon stock gains with national climate targets under the Paris Agreement. Environmental policymakers can rely on robust carbon stock data to design blue carbon frameworks, such as the Kunming–Montreal Global Biodiversity Framework [9]. And scientists working on nature-based solutions can leverage monitoring insights to refine climate change predictive models, and address methodological gaps.

Guangdong Province, with its extensive coastline and abundant blue carbon ecosystem resources, is one of the most economically vibrant coastal provinces in China, and has seen its mangrove and salt marsh ecosystems under significant pressures due to rapid urbanization and overdevelopment [39,40]. Mangroves and salt marshes coexist in the intertidal wetlands in Guangdong, forming many mangrove–salt marsh ecotones [41,42]. The fragmented, dynamic, and heterogeneous intertidal wetlands present significant challenges for the detailed and precise monitoring of mangroves and salt marshes within the ecotones. Researchers have conducted extensive research on the application of remote sensing for mapping and monitoring mangroves [43,44,45,46,47,48,49,50] and salt marshes [35,37,51,52,53,54,55,56]. However, most studies produced mangrove or salt marsh distribution datasets separately, there are few studies on long-term and large-scale monitoring of mangrove-salt marsh ecotones [57]. In recent years, Guangdong has placed high priority on the protection and restoration of coastal ecosystems, advancing these efforts through policy innovation, ecological restoration projects, and multi-party collaboration [58,59,60,61]. However, systematic assessments of spatiotemporal carbon storage changes in mangroves and salt marshes in Guangdong Province remain limited. This study focuses on quantifying the carbon stock evolutions of mangroves and salt marshes in Guangdong over the past three decades with Landsat image time series, aiming to provide actionable insights for optimizing coastal ecosystem management and enhancing blue carbon contributions to global climate mitigation.

2. Materials

2.1. Study Area

Guangdong Province (20°09′~25°31′ N, 109°45′~117°20′ E), with a total area of approximately 179,700 km², is located in the southeastern coastal region of China, adjacent to the South China Sea, bordering Jiangxi, Fujian, Guangxi, Hainan, Hong Kong, and Macau. It has 14 coastal cities, including Chaozhou, Shantou, Jieyang, Shanwei, Huizhou, Shenzhen, Dongguan, Guangzhou, Zhuhai, Zhongshan, Jiangmen, Yangjiang, Maoming, and Zhanjiang (Figure 1). Guangdong has vast maritime areas with excellent harbors and numerous islands, diverse coastal ecosystem resources with mangroves, salt marshes, and mudflats. It lies in the East Asian monsoon region, with abundant light, heat, and water resources. The precipitation mainly concentrated from April to September. The average annual temperature is 21.9 °C. Guangdong has the longest mainland coastline of 4084.48 km and the largest area of mangroves in China [39]. It has 14 true mangrove and 10 semi-mangrove species, with Acrostichum aureum, Aegiceras corniculatum, Acanthus ilicifolius, Kandelia obovata, Excoecaria agallocha, as well as Avicennia marina as the most prevalent mangrove species here [39]. The coastal substrates in Guangdong also create a suitable environment for the growth of blue carbon ecosystems here.

2.2. Datasets

2.2.1. Landsat Data and Preprocessing

All the United States Geological Survey (USGS) surface reflectance (SR) images from the Landsat-5 TM and Landsat-8 OLI along Guangdong coastlines on the Google Earth Engine platform in 1986, 1990, 1995, 2000, 2005, 2010, 2015, and 2020 were utilized to monitor mangroves and salt marshes. We removed bad-quality observations, such as clouds, cloud shadows, cirrus, snow/ice, with the “QA_PIXEL” band [62,63]. For every Landsat image, we used six spectral bands (resolution: 30 m), which were blue, green, red, near infrared, short-wave infrared 1, and short-wave infrared 2 [34].

2.2.2. Reference Data

To collect the reference data for the three mapping classes (mangroves, salt marshes, and others), we carried out field surveys and gathered UAV images along the coastlines in Guangdong using DJI Phantom 4 RTK (DJI, Shenzhen, China), with flight altitude of 100 m and 70% sidelap (the spatial resolution is about 3 cm) in 2020 [64]. We also collected very high spatial resolution (VHR) satellite images in Guangdong during the study period from 1986 to 2020. We visually interpreted the VHR images from UAV and satellite to supplement the reference data. As the number of the VHR images was limited before 2005, we also consulted 5 local experts (1 professor specializing in mangrove ecology and 4 technical specialists from mangrove nature reserves in Guangdong), and created reference data with ArcGIS software (v9.3) for algorithm training and validation.

3. Methods

3.1. Detection of Mangroves and Salt Marshes

Figure 2 presents a flowchart of the detection method in mapping mangroves and salt marshes with Landsat image time series. It has two major steps: (1) creating an annual coastal vegetation mask from multi-temporal Landsat imagery, (2) generating annual mangrove and salt marsh maps.

We first gathered Landsat images (cloud percentage < 30%) during each year, and then generated annual cloud-free image composites by calculating the median values of each band. With the image composite for each year, we computed two vegetation indices and one water-related spectral index, including normalized difference vegetation index (NDVI) [65], enhanced vegetation index (EVI) [66], and land surface water index (LSWI) [67]. These indices were combined to generate the green vegetation mask, refer to Equation (4) for details [68].

$$NDVI={\displaystyle \frac{{\rho }_{nir}-{\rho }_{red}}{{\rho }_{nir}+{\rho }_{red}}}$$

(1)

$$EVI=2.5\times {\displaystyle \frac{{\rho }_{nir}-{\rho }_{red}}{{\rho }_{nir}+6\times {\rho }_{red}-7.5\times {\rho }_{blue}+1}}$$

(2)

$$LSWI={\displaystyle \frac{{\rho }_{nir}-{\rho }_{swir}}{{\rho }_{nir}+{\rho }_{swir}}}$$

(3)

$$Green Vegetation=\left(EVI\ge 0.1\right) and \left(NDVI\ge 0.2\right) and (LSWI>0)$$

(4)

where ${\rho }_{red}$, ${\rho }_{blue}$, ${\rho }_{nir}$, and ${\rho }_{swir}$ are the SR values of red, blue, near-infrared, and shortwave infrared (1.55–1.75 micrometers for Landsat-5 TM imagery, and 1.57–1.65 micrometers for Landsat-8 OLI imagery) bands in Landsat images.

We digitized coastlines through the visual interpretation of the Landsat imagery each year. Then, we used the coastline data to mask out the terrestrial vegetation, and generated a coastal vegetation mask for subsequent identification of mangroves and salt marshes.

Owing to the prevailing cloudy and rainy weather conditions along coastlines, it is difficult to collect sufficient high-quality Landsat images for wetland classification [69]. We proposed a mangrove and salt marsh detection framework (Figure 2) to generate annual mangrove and salt marsh distribution maps in scenarios with limited high-quality satellite acquisitions. We first judge whether there exist enough cloud-free observations for producing both a green-up season image composite and a senescence season image composite for the study region in each year. Here, the green-up season and the senescence season are defined as the periods of salt marsh growth initiation and decline, respectively [37]. If there are not enough good-quality images, we used Method One; otherwise, we resorted to Method Two (Figure 2). As to Method One, we first combined the annual cloud-free image composite and the coastal vegetation mask to generate the image composite of coastal vegetation; then used the six spectral bands and NDVI, EVI, and LSWI indices as the input features, and applied a random forest classifier to identify mangroves and salt marshes. For the random forest classifier, the number of variables per split was set at the square of the number of variables, and the number of trees was set at 100 [38]. As to Method Two, we generated both a green-up season image composite and a senescence season image composite based on the satellite images acquired in one year:

(1) To produce the green-up season image composite, we collected Landsat images (cloud percentage < 70%) acquired from April 1 to November 1, which characterized the green-up season [35], and then used EVI to select the green-up season images [34]. Higher EVI suggests closer to the green-up season [34] and image pixels less affected by clouds or tidal inundations. Similar to Zhang et al. (2022) [37], we flagged each Landsat scene with the spatially average EVI value of salt marsh and mangrove samples, and then sorted them in descending order to obtain the highly ranked images as the green-up season images. Then we generated the cloud-free green-up season image composite by a median synthesis of the selected green-up season satellite images.

(2) To obtain the senescence season image composite, we gathered Landsat images (cloud percentage < 70%) acquired from 1 January to 1 March, and 1 December to 31 December, which denoted the senescence season [35]; then used plant senescence reflectance index (PSRI) to choose the senescence images for salt marshes [34,37]. Higher PSRI denotes increasing senescence of vegetation [34,37]. We constructed the cloud-free senescence season image composite by a median synthesis of the highly ranked Landsat images with a high average PSRI of the salt marsh samples.

$$PSRI={\displaystyle \frac{{\rho }_{red}-{\rho }_{blue}}{{\rho }_{nir}}}$$

(5)

Given that the vegetation phenology may vary in different estuaries, the average EVI/PSRI values in Guangdong may fail to adequately represent the actual phenological features in some estuaries. We divided the whole study region into several subregions based on estuarine systems in Guangdong, and performed the above steps for each subregion to obtain the green-up season/senescence season image composites. The coastal vegetation mask was used to clip the two image composites to generate composites of coastal vegetation. For each subregion, we combined the clipped green-up season/senescence season image composite to generate an image of twelve spectral bands (six bands each in the two composites), and employed the twelve spectral bands, as well as the indices including NDVI, EVI, LSWI, and PSRI of the two composites as the classification features. The random forest classifier was applied to obtain the mangrove and salt marsh detection results.

For validation, we adopted the stratified equalized random sampling to avoid bias and arbitrariness in the accuracy assessment [37,70]. For each year, 300 independent points for mangroves, salt marshes, and others were randomly generated (Figure S1). We constructed a confusion matrix for each year, and used overall accuracy and Kappa coefficient to evaluate the accuracy of the detection results.

3.2. Dynamic Degree of Area

We utilized a dynamic degree of area [71,72] to study the dynamics of mangroves and salt marshes in Guangdong Province. A greater absolute value of this index indicates a larger ecosystem area change [71]. And the positive/negative value of this index suggests the area is increasing/decreasing. The calculation equation is presented as follows:

$$D_i={\displaystyle \frac{{A_{i2}-\mathrm{A}}_{i1}}{A_{i1}}}\times {\displaystyle \frac{1}{T}}\times 100\%$$

(6)

where $i$ is a certain type of ecosystem; $D_i$ is the dynamic degree of a certain ecosystem; $A_{i1}$ is the area of a certain ecosystem at the beginning of the study period; $A_{i2}$ is the area of a certain ecosystem at the end of the study period; $T$ is the length of the study period.

3.3. Carbon Storage Assessment for Mangroves and Salt Marshes with the InVEST Model

We employed the InVEST Carbon Storage and Sequestration model to estimate the carbon storage of mangroves and salt marshes, and analyzed the carbon storage spatiotemporal dynamics (Figure 3). The InVEST model divides the carbon storage of ecosystems into four basic carbon pools: aboveground biomass carbon, belowground biomass carbon, soil organic carbon, and dead organic carbon. We estimated the carbon storage based on the ecosystem distribution maps and carbon density data of each ecosystem [21,32]. With the InVEST model, the carbon density of each ecosystem is considered as a constant value. For each ecosystem, the quantity of carbon storage in a specific region is determined by multiplying the carbon stock per unit area for each ecosystem type by its area. The equations are listed as follows:

$$C_i=C_{i\_above}+C_{i\_below}+C_{i\_soil}+C_{i\_dead}$$

(7)

$$C_{i\_total}=C_i\times A_i$$

(8)

where $i$ is a certain type of ecosystem; $C_i$ is the carbon density of a certain type of ecosystem; $C_{i\_above}$, $C_{i\_below}$, $C_{i\_soil}$, and $C_{i\_dead}$ represent the aboveground biomass carbon density, belowground biomass carbon density, soil organic carbon density, and dead organic carbon density, respectively. $C_{i\_total}$ is the total amount of carbon stock for a certain ecosystem type; and $A_i$ is the area of a certain ecosystem type $i$.

Considering that the study areas from Wang et al. (2013) [22], Gao et al. (2017) [73], and Qin et al. (2023) [74] were in Guangdong Province, we collected the mangrove carbon density data from these studies for mangrove carbon storage estimation (

Table 1. Carbon densities for each ecosystem type in Guangdong Province (Mg C/hm²).

Ecosystem Type	${\mathit{C}}_{\mathit{i}\_\mathit{a}\mathit{b}\mathit{o}\mathit{v}\mathit{e}}$	${\mathit{C}}_{\mathit{i}\_\mathit{b}\mathit{e}\mathit{l}\mathit{o}\mathit{w}}$	${\mathit{C}}_{\mathit{i}\_\mathit{s}\mathit{o}\mathit{i}\mathit{l}}$	${\mathit{C}}_{\mathit{i}\_\mathit{d}\mathit{e}\mathit{a}\mathit{d}}$	Sources
mangroves	153.23	28.28	230	0.66	[22,73,74]
salt marshes	3.22	3.99	106.5	1.02	-

). Given the scarcity of literature on salt marsh carbon density in Guangdong, we adopted field investigation and laboratory analysis methods [75] to gather the carbon density data of salt marshes in Zhuhai, Guangdong Province for salt marsh carbon storage estimation (Table 1).

4. Results

4.1. Accuracy Assessment

Table 2. Detection accuracies in Guangdong Province.

Year	1986	1990	1995	2000	2005	2010	2015	2020
Overall Accuracy	96%	96%	96%	95%	94%	96%	95%	98%
Kappa	0.94	0.95	0.94	0.93	0.92	0.94	0.92	0.97

shows the overall accuracies and Kappa coefficients of the mangrove and salt marsh detection results. The variable tidal conditions and cloudy weather along the coastlines might contribute to the classification errors. The overall accuracies of the mangrove and salt marsh detection results all exceeded 90%, and the Kappa coefficients were greater than 0.90, demonstrating the good consistency between our results and those from the validation data.

4.2. Spatiotemporal Patterns of Mangroves

Mangroves in Guangdong Province are mainly distributed in 14 coastal cities (Figure 4). The distribution maps of mangroves showed similar patterns from 1986 to 2020. Mangroves were mainly distributed in the western coastal regions in Guangdong, while the mangrove area in eastern Guangdong was relatively smaller [76]. The mangrove areas in Guangdong in 1986, 1990, 1995, 2000, 2005, 2010, 2015, and 2020 were 10,060.48 ha, 7818.43 ha, 7769.19 ha, 8293.3 ha, 7740.93 ha, 8129.91 ha, 10,627.61 ha, and 11,900.3 ha, respectively. And the mangrove area in Guangdong showed a trend of decreasing from 1986 to 1995, fluctuated from 1995 to 2005, and then exhibited a trend of increasing from 2005 to 2020. The total mangrove area increased by 1839.82 ha from 1986 to 2020, with a dynamic degree of 0.54% (

Table 3. Temporal area changes of mangroves and salt marshes in Guangdong Province.

Year	Mangroves		Salt Marshes
	Area Change (ha)	Dynamic Degree	Area Change (ha)	Dynamic Degree
1986–1990	−2242.05	−5.57%	−4.86	−0.16%
1990–1995	−49.24	−0.13%	−91.66	−2.42%
1995–2000	524.11	1.35%	130.94	3.94%
2000–2005	−552.37	−1.33%	−135.29	−3.40%
2005–2010	388.98	1.00%	134.28	4.07%
2010–2015	2497.7	6.14%	274.78	6.92%
2015–2020	1272.69	2.40%	185.1	3.46%
1986–2020	1839.82	0.54%	493.29	1.91%

). From 1986 to 2020, the mangrove area in Zhanjiang grew the fastest, with an area increase of 1011.57 ha. The mangrove areas in Jiangmen, Guangzhou, Maoming, Shantou, Zhuhai, Huizhou, Zhongshan, Chaozhou, and Yangjiang increased by 601.39 ha, 269.21 ha, 268.34 ha, 184.62 ha, 178.31 ha, 89.21 ha,41.88 ha, 15.23 ha, and 11.61 ha, respectively. Conversely, declines in mangrove areas were observed in Shenzhen, Dongguan, Shanwei, and Jieyang during the study period.

From 1986 to 1990, the mangrove area in Guangdong decreased by 2242.05 ha, with a dynamic degree of −5.57% (Table 3). Among the 14 coastal cities, the mangrove area in Zhanjiang decreased the fastest, with the reduction mainly concentrated in Lianjiang and Potou. From 1990 to 1995, the mangrove area in Guangdong continued to decrease by 49.24 ha, with a dynamic degree of −0.13%. During this period, the mangrove area in Shenzhen had the highest degree of decline, with a dynamic degree of −9.23%. From 1995 to 2000, the mangrove area in Guangdong increased by 524.11 ha, with a dynamic degree of 1.35%. And the mangrove dynamic degree in Jiangmen was 13.69%, which was the largest among the 14 coastal cities in Guangdong. From 2000 to 2005, the mangrove area in Guangdong decreased by 552.37 ha, with a dynamic degree of −1.33%. From 2005 to 2010, the mangrove area increased by 388.98 ha, with a dynamic degree of 1.00%. From 2010 to 2015, the mangrove area continued to increase by 2497.7 ha, with a dynamic degree of 6.14%. During this period, Zhanjiang presented a total mangrove area increase of 2327.91 ha, which was the largest mangrove area increase in Guangdong. From 2015 to 2020, the mangrove area continued to grow by 1272.69 ha, with a dynamic degree of 2.4%. And the period from 2010 to 2020 saw the largest mangrove area increase in Guangdong.

4.3. Spatiotemporal Patterns of Salt Marshes

Salt marshes in Guangdong Province are also mainly distributed in the 14 coastal cities (Figure 5). The salt marsh distribution maps exhibited consistent patterns over the period from 1986 to 2020. Salt marshes were mainly distributed in Zhuhai and Jiangmen. The areas of salt marshes in 1986, 1990, 1995, 2000, 2005, 2010, 2015, and 2020 were 760.95 ha, 756.09 ha, 664.43 ha, 795.37 ha, 660.08 ha, 794.36 ha, 1069.14 ha, and 1254.24 ha, respectively. From 1986 to 2020, the overall trend of the salt marsh area was upward, with small fluctuations. The total salt marsh area in Guangdong increased by 493.29 ha, with a dynamic degree of 1.53%. The salt marsh area in Zhuhai increased the fastest, with a total area increase of 439.86 ha. The salt marsh areas in Jiangmen, Zhanjiang, Yangjiang, Shantou, Guangzhou, Shenzhen, Dongguan, Chaozhou, and Maoming increased by 176.35 ha, 162.33 ha, 116.37 ha, 49.84 ha, 33.36 ha, 8.95 ha, 6.37 ha, 5.75 ha, and 0.55 ha, respectively. Salt marsh areas in Jieyang, Shanwei, Huizhou, and Zhongshan experienced declines from 1986 to 2020.

From 1986 to 1990, the area of salt marshes decreased by 4.86 ha, with a dynamic degree of −0.16% (Table 3). From 1990 to 1995, the area of salt marshes continued to decrease by 91.66 ha, with a dynamic degree of −2.42%. From 1995 to 2000, the area of salt marshes increased by 130.94 ha, with a dynamic degree of 3.94%. From 2000 to 2005, the area of salt marshes decreased by 135.29 ha, with a dynamic degree of −3.4%. During this period, Zhuhai had the largest area reduction rate in salt marshes, with an area decrease of 119.25 ha. From 2005 to 2010, the area of salt marshes increased by 134.28 ha, with a dynamic degree of 4.07%. From 2010 to 2015, the area of salt marshes continued to increase by 274.78 ha with a dynamic degree of 6.92%. During this period, Jiangmen exhibited the highest growth rate in the salt marsh area in the province, with its dynamic degree reaching 28.35%. From 2015 to 2020, the area of salt marshes continued to increase by 185.10 ha, with a dynamic degree of 3.46%. Similar to mangroves, Guangdong witnessed the largest increase in salt marsh area during the period from 2010 to 2020.

4.4. Spatiotemporal Analysis of Carbon Storage in Mangroves and Salt Marshes

The mangrove carbon storage in Guangdong Province in 1986, 1990, 1995, 2000, 2005, 2010, 2015, and 2020 were 414.66 $\times$ 10⁴ Mg C, 322.25 $\times$ 10⁴ Mg C, 320.22 $\times$ 10⁴ Mg C, 341.82 $\times$ 10⁴ Mg C, 319.06 $\times$ 10⁴ Mg C, 335.09 $\times$ 10⁴ Mg C, 438.04 $\times$ 10⁴ Mg C, and 490.49 $\times$ 10⁴ Mg C, respectively (Figure 6). From 1986 to 2020, the total mangrove carbon storage increased by 75.83 $\times$ 10⁴ Mg C in Guangdong. From 1986 to 1995, the mangrove carbon storage decreased by 94.44 $\times$ 10⁴ Mg C, with mangrove carbon storage losses mainly concentrated in areas such as Zhanjiang and Shenzhen. From 1995 to 2000, the mangrove carbon storage increased by 21.6 $\times$ 10⁴ Mg C, with carbon storage growth in Yangjiang, Jiangmen, Zhongshan, etc. From 2000 to 2005, the mangrove carbon storage decreased by 22.77 $\times$ 10⁴ Mg C. During this period, Yangjiang presented the largest reduction in mangrove carbon storage, while the mangrove carbon storage values in Zhanjiang, Zhuhai, Shenzhen, Dongguan, and Chaozhou showed an upward trend. From 2005 to 2020, the mangrove carbon storage increased by 171.44 $\times$ 10⁴ Mg C. Mangrove carbon storage values increased in most coastal cities, with Zhanjiang having the largest mangrove carbon storage increase.

The distribution patterns of mangrove carbon storage in Guangdong Province exhibit significant spatial heterogeneity, characterized by higher values in western Guangdong and lower values in eastern Guangdong (Figure 7). Areas with higher carbon storage values are mainly concentrated in Zhanjiang, Yangjiang, and Jiangmen, with large mangrove distribution areas. In contrast, areas with lower carbon storage are primarily found in Jieyang, Chaozhou, and Shanwei, where the spatial distribution of mangroves is uneven and fragmented, with a predominance of small-sized mangrove patches.

To investigate the long-term carbon storage changes of mangroves in Guangdong, we subtracted the mangrove carbon storage data acquired in 1986 from the data in 2020, analyzed the spatial patterns (Figure 8). From 1986 to 2020 in Guangdong, the mangrove carbon storage in 1912.48 ha remained stable, whereas the mangrove carbon storage over an area of 9987.82 ha (8148.01 ha) showed an increasing (decreasing) trend.

The carbon storage of salt marshes in Guangdong Province in 1986, 1990, 1995, 2000, 2005, 2010, 2015, and 2020 were 8.73 $\times$ 10⁴ Mg C, 8.67 $\times$ 10⁴ Mg C, 7.62 $\times$ 10⁴ Mg C, 9.13 $\times$ 10⁴ Mg C, 7.57 $\times$ 10⁴ Mg C, 9.11 $\times$ 10⁴ Mg C, 12.27 $\times$ 10⁴ Mg C, and 14.39 $\times$ 10⁴ Mg C, respectively (Figure 6). From 1986 to 2020, the carbon storage of salt marshes increased by 5.66 $\times$ 10⁴ Mg C, with an overall trend of first decreasing, then fluctuating, and finally increasing. From 1986 to 1995, the carbon storage of salt marshes increased by 1.11 $\times$ 10⁴ Mg C, mainly concentrated in cities such as Shantou, Jiangmen, and Huizhou. From 1995 to 2000, the carbon storage of salt marshes increased by 1.51 $\times$ 10⁴ Mg C. During this period, the carbon storage of salt marshes in Zhanjiang, Shenzhen, Huizhou, and Shantou decreased, while other cities witnessed increases in salt marsh carbon storage. From 2000 to 2005, the carbon storage of salt marshes decreased by 1.55 $\times$ 10⁴ Mg C; cities including Zhuhai, Jiangmen, and Shenzhen experienced the most significant declines of salt marsh carbon storage. From 2005 to 2020, the carbon storage of salt marshes increased by 6.82 $\times$ 10⁴ Mg C, with Zhuhai having the largest increase in salt marsh carbon storage in Guangdong.

The distribution pattern of carbon storage in salt marshes in Guangdong Province is displayed in Figure 9. The high carbon storage areas are mainly concentrated in Zhuhai, Jiangmen, Zhanjiang, Shantou, and Jieyang, while the low carbon storage areas are in Yangjiang, Shenzhen, Zhongshan, Guangzhou, Shanwei, and Huizhou. From 1986 to 2020, the salt marsh carbon storage in 15.27 ha remained stable, whereas the salt marsh carbon storage over an area of 1238.97 ha (745.68 ha) showed an increasing (decreasing) trend in Guangdong (Figure 8).

5. Discussion

5.1. Mangroves and Salt Marshes in Guangdong

We proposed a new mangrove and salt marsh detection framework, considering the high-quality satellite image acquisition and phenological characteristics for accurate mangrove and salt marsh mapping. When sufficient good-quality satellite images were available, we generated a green-up season image composite and a senescence season image composite in one year to obtain phenological features; otherwise, we generated one annual coastal vegetation composite for tidal wetland vegetation classification. While researchers used Sentinel-2 satellite imagery with higher temporal resolution to obtain the phenological information for wetland classification [35,37], this study explored the potential of obtaining phenological characteristics from the long-term Landsat image time series to analyze the distribution dynamics of mangroves and salt marshes. The overall accuracies from 1986 to 2020 were greater than 90%, indicating the reliable classification results. The long-term distribution data of mangroves and salt marshes here provided a good dataset for in-depth analysis of blue carbon dynamics in Guangdong.

The mangrove areas in Guangdong measured by Wang et al. (2017) [77] with Landsat satellite time series and Object-based Image Analysis method, were 10,200 ha in 1987 and 8700 ha in 2000, respectively, which closely matched the 10,060.48 ha in 1986 and 8293.30 ha in 2000, respectively, in this study. The mangrove identification results from Wu et al. (2013) [78] (7733.20 ha in 1990 and 8722 ha in 2000) were consistent with our detection results (7818.43 ha in 1990 and 8293.30 ha). The mangrove detection results in 2015 from Jia et al. (2018) [50] (9855 ha) and Ma et al. (2019) [76] (9700 ha) aligned with our results (10,627.85 ha). The mangrove area in 2020 was 11,928.87 ha in our study, similar to the 10,651.30 ha based on the Third National Land Survey (2017–2019) which was derived from high spatial resolution satellite imagery; and similar to the 10,745 ha from Zhao et al. (2022) [79] which was based on 10-m-resolution Sentinel-1 and Sentinel-2 images in 2019 in Guangdong. Furthermore, we found our method generated more accurate mangrove detection results by comparing with other published literature [50] (Figure A1).

Compared to mangroves, fewer studies have focused on salt marsh distribution monitoring in Guangdong [80]. One important reason may be that mangroves dominate as the primary blue carbon ecosystems in Guangdong, while salt marshes are distributed in small, fragmented patches along the coast, contributing to greater uncertainties in salt marsh mapping. Hu et al. (2021) [80] mapped salt marshes along China’s coast with time series Sentinel-1 SAR data in 2019 and the decision tree classifier, but only identified Spartina alterniflora (S. alterniflora) with an area of 371.54 ha in Guangdong. Hu et al. (2021) [80] ignored the native coastal salt marsh species (e.g., Phragmites australis, Cyperus malaccensis) (Figure 10) in Guangdong, leading to a significant underestimation of the total salt marsh area. Based on Landsat and Sentinel-2 imagery, Chen et al. (2022) [81] and Gu et al. (2021) [82] applied a support vector machine classifier to identify coastal salt marshes in China, and stated that the detected salt marsh extent was conservative. Our salt marsh detection results from 1995 to 2020 in Guangdong were about 20% higher than those from Chen et al. (2022) [81] and Gu et al. (2021) [82]. Liu et al. (2018) [83] mapped the S. alterniflora invasion in mainland China with Landsat images, and presented that the S. alterniflora area in Guangdong was 780 ha in 2015. In this study, we mapped 1069.14 ha of salt marshes in 2015, which was reasonable considering the extensive distribution of native salt marsh species across the coastlines in Guangdong. Notably, the trend of the salt marsh areas from 2000 to 2015 in our study was similar to that of the S. alterniflora areas in Guangdong from Mao et al. (2019) [53], suggesting the S. alterniflora invasion profoundly altered the spatial distribution of salt marshes in Guangdong. Different from previous studies, which mostly employed the traditional classifiers for single-step extraction of all coastal land cover types [81], we first generated the coastal vegetation mask to delineate the regions where mangroves and salt marshes were further classified. Moreover, when enough observations existed, we added the phenological features for accurate mangrove and salt marsh identification.

As to the carbon storage of mangroves and salt marshes in Guangdong, we found that the carbon storage values of mangroves were much larger than those of salt marshes, which was mainly due to the larger distribution areas and carbon density values of mangroves. Interestingly, the mangrove carbon stock from 2017 to 2020 in Hainan Province, China was 70.32 $\times$ 10⁴ Mg C [84], much smaller than the mangrove carbon stock in 2020 in Guangdong (490.49 $\times$ 10⁴ Mg C). Nevertheless, the salt marsh carbon stock in Liaoning Province, China was 5.41 $\times$ 10⁶ Mg C [85], much larger than the salt marsh stock in 2020 in Guangdong (14.39 $\times$ 10⁴ Mg C). Furthermore, the mangrove and salt marsh carbon stocks in North America were 386.39 $\times$ 10⁶ Mg C and 343.49 $\times$ 10⁶ Mg C, respectively [86], much larger than those in Guangdong. Compared with terrestrial forests, the carbon stocks of mangroves in 2015 accounted for only about 10% of the coniferous forest carbon stocks in 2012 in Guangdong, while the salt marsh carbon stocks were comparable to the shrub carbon stocks [87]. The carbon stocks of terrestrial forests significantly exceeded those of the two blue carbon ecosystems in Guangdong [87], highlighting the critical importance of green carbon in achieving carbon neutrality.

5.2. Factors Affecting Mangroves and Salt Marshes

Mangroves and salt marshes, functioning as unique yet fragile ecosystems, face multifaceted threats, such as climate change, sea level rise, anthropogenic disturbances, and biological invasions [56,88,89]. In the context of global warming, the rising frequency of extreme weather events, such as severe cold spells and tropical cyclones, poses a growing threat to the survival of mangroves and salt marshes [90,91,92]. The sea level rise has exacerbated the threats posed by marine disasters, including storm surges, coastal erosion, and saltwater intrusion to blue carbon ecosystems [93]. According to long-term sea level data from 1980 to 2021, the rate of sea level rise along China’s coasts has reached 3.4 mm/a, exceeding the global average during the same period [94]. However, the average sedimentation rate of mangroves in Guangdong is 13.5 mm/year [74], significantly higher than the rate of sea level rise. Currently, studies on salt marshes in Guangdong remain scarce, but research results in Texas, America, presented mineral trapping rates 4.1 times higher in mangroves compared with adjacent salt marshes [95]. Hence, mangroves in Guangdong are less affected by the sea level rise.

Non-native species are a significant factor influencing the dynamics of salt marshes and mangroves in Guangdong. Researchers reported that the invasion of S. alterniflora exerts adverse effects on mangroves, which may lead to the replacement of mangroves by S. alterniflora [57,96]. On the one hand, S. alterniflora has high primary productivity and biomass with a slower degradation rate, contributing to increased carbon storage [97]. On the other hand, its strong adaptability makes it spread quickly and widely, leading to biodiversity losses in coastal wetlands (plant, microbial, and animal diversity) [98]. And the invasion of S. alterniflora has severely affected the nutrient structure, benthic community biodiversity, and other biological indicators of the coastal habitat [97,99]. Figure 11 illustrates the invasion of S. alterniflora along the coastlines in Guangdong. With the implementation of the “Special Action Plan for the Prevention and Control of Spartina alterniflora (2022–2025)” [70], the adverse impacts of this invasive species on mangroves will be substantially mitigated.

Sonneratia apetala (S. apetala), a mangrove species native to Bangladesh and Sri Lanka, was introduced to Guangdong for mangrove afforestation programs and S. alterniflora control in the early 1990s. According to the latest research, S. apetala was primarily distributed in Guangdong Province, accounting for 14.8% of the total mangrove area in China in 2022 [100]. S. apetala can provide food and habitats for wildlife, and protection against storms. Researchers stated that S. apetala can provide the highest protection against cyclones, compared with other mangrove species [101]. But in some mangrove habitats in Guangdong, this species demonstrates stronger competitiveness compared to native mangrove plants, exhibiting noticeable invasive characteristics [102] and exerting negative effects on the structural integrity and functional dynamics of local mangrove ecosystems [103,104]. As to carbon sequestration capacity, some researchers suggested that S. apetala presented greater potential for carbon sequestration than most native species [105], while other researchers found that the native K. obovate had larger content and pools of soil organic carbon and nitrogen than those of exotic S. apetala. Therefore, planting native mangroves in existing S. apetala plantations is suggested to enhance the carbon storage capacity in Guangdong [106,107].

Moreover, anthropogenic activities constitute one of the most critical factors influencing the spatiotemporal distribution of mangroves and salt marshes. The seaward land reclamation and intensive aquaculture have caused the rapid loss of coastal ecosystems [108,109], while the restoration and protection projects have reversed the trend of habitat decline [42].

5.3. Implications for Blue Carbon Ecosystem Management in Guangdong

The timeline of key blue carbon ecosystem-related policies and economic plans in China is presented in Figure 12. Besides carbon sequestration and storage, mangroves and salt marshes provide other valuable ecosystem services, including elevation maintenance during sea-level rise, protection from storms and flooding, habitat and food provision, cultural values, nutrient cycling [95], etc. With the growing recognition of the invaluable ecosystem services, the conservation and restoration of the blue carbon ecosystems in China have received more attention over the past three decades. According to the Regulations of the People’s Republic of China on Nature Reserves (released in 1994), priority should be given to in-situ conservation through nature reserve designation for ecosystems with high representativeness and unique ecological functions. This serves as a cornerstone approach for protecting blue carbon ecosystems in China. The National Plan for Wetland Conservation and Restoration, approved by the State Council in 2003, proposed that “by 2010, 50% of the natural wetlands will be conserved, and 18,000 ha of mangrove ecosystems will be cultivated or restored”. The increasing trend of the mangrove area from 2010 to 2020 demonstrated the effectiveness of the above conservation policies in Guangdong. However, the impacts of urbanization and socioeconomic activities on blue carbon ecosystems should not be overlooked [110]. Researchers identified multiple anthropogenic activities, like coastal reclamation, pollutant discharge, aquaculture practices, and port construction, as the primary causes of coastal ecosystem degradation in China [110,111]. The rapid reduction in Guangdong’s mangrove area has been driven by agricultural coastal land reclamation during the 1960s–1970s, enclosed mudflat aquaculture in the 1980s–1990s, and urbanization-driven construction since the 1990s [112]. Li et al. (2011) [110] reported that from 1986 to 2005, the intensive coastal land reclamation and mudflat development activities resulted in a loss of 8.39 km² in mangroves, while the built-up land in coastal zones expanded by 678 km². Figure A2 illustrates the examples of mangrove losses in Guangdong.

By analyzing the correlations among the timeline of blue carbon ecosystem-related policies, economic development plans, and changes in blue carbon storage between 1980 and 2020, we identified several notable findings. Following the initiation of the National Ecological Environment Construction Plan of China in 1998, carbon storage in both mangroves and salt marshes showed an upward trend (Figure 6). After China’s accession to the WTO in 2001, a notable decline in carbon storage of mangroves and salt marshes was observed in Guangdong. Conversely, with the implementation of the National Plan for Wetland Conservation and Restoration, and the Wetland Protection Regulations of Guangdong Province, carbon stocks in mangroves and salt marshes rebounded. In 2012, following the integration of ecological civilization construction into the national policy framework, carbon storage in mangroves and salt marshes experienced a dramatic increase in subsequent years. By contrast, the new urbanization strategy enacted in 2014 may have inhibited the growth of carbon storage in blue carbon ecosystems.

Guangdong Province places significant emphasis on mangrove protection. By 2024, Guangdong has established three internationally important wetlands and two provincially important wetlands, primarily designated for mangrove conservation. Additionally, there are 47 nature reserves with mangrove ecosystems in Guangdong. Wetland Protection Regulations of Guangdong Province, issued in 2006, also underscores the protection and restoration of mangroves. With these measures and policies, the mangrove area kept rising from 2005 to 2020 in Guangdong (Table 3, Figure A3). Although there are currently no nature reserves specifically dedicated to the restoration of coastal salt marshes, mangrove and salt marsh ecotones are widely distributed, and salt marshes adjacent to mangroves are afforded protection within these reserves. Furthermore, the local government has restored salt marshes through ecological restoration projects (Figure A4).

In this study, we found that 9.54% of the original mangrove area in Guangdong remained intact from 1986 to 2020, while 40.64% of mangrove ecosystems were lost, with far fewer salt marshes surviving intact. The limited expanse and fragmented distribution of salt marshes in Guangdong, coupled with the province’s extensive coastlines, present substantial challenges for the conservation of salt marsh ecosystems. Moreover, the S. alterniflora invasion poses a critical threat to the survival of native salt marsh species in Guangdong. Strengthened conservation measures are therefore necessary to safeguard Guangdong’s salt marshes.

Interestingly, researchers found that mangrove reforestation (replanting mangroves where they previously existed) provides much greater carbon storage than afforestation (planting mangroves where no mangroves previously existed) [113], the areas where mangroves previously colonized could be given priority for mangrove restoration. Moreover, for areas showing large carbon losses, like the conversion of mangroves into aquaculture, immediate conservation actions should be taken to restrict further blue carbon losses. And in carbon-gain regions, such as restored mangroves in Guangdong, tax incentives can be proposed for eco-tourism development aligned with the SDG 14 targets. Therefore, the long-term time-series blue carbon stock datasets and change analysis results in this study (Figure 8) provide valuable data support for blue carbon ecosystem restoration planning and management.

5.4. Limitations and Future Improvements

This study is subject to three key limitations: (1) limited spatial resolution of the satellite imagery, (2) reliance on a single carbon density dataset, and (3) a lack of sufficient in-situ data for driving factor analysis of the carbon storage changes. The mangrove and salt marsh distribution maps generated in this study were derived from Landsat time series with a spatial resolution of 30 m, ignoring coastal vegetation patches smaller than 900 m². Over time, both anthropogenic and environmental factors have continuously driven the escalating fragmentation of ecosystems. In addition, the ongoing ecological restoration programs might not be fully captured by remote sensing, due to the weak spectral signals from the newly planted seedlings. These factors may collectively lead to the underestimation of carbon storage in mangrove and salt marsh ecosystems. With the rapid progress of AI-powered remote sensing technology and the growing archives of high-spatial-resolution satellite and aerial imagery, it is possible to generate fine-scale distribution maps of blue carbon ecosystems with accurate species detection and high temporal resolution in the future.

Another limitation was that we estimated carbon storage in mangroves and salt marshes with one carbon density dataset (Table 1) for the entire coastal region of Guangdong Province over time, whereas carbon densities of these ecosystems can vary in space, time, and species. For instance, carbon densities of the same species may differ across growth stages or seasons [114,115,116]. Therefore, it is essential to conduct long-term in-situ monitoring of ecosystem carbon densities across different local regions and for different tidal vegetation species of various growth stages, for further carbon stock estimation improvement [116]. Moreover, improved carbon stock models with the integration of deep learning techniques and multi-source remote sensing datasets provide another important research direction that can enhance the accuracy and reliability of carbon stock estimations [117].

In this study, we attributed the blue carbon ecosystem carbon storage changes to the national and provincial policy events through timeline alignment. To systematically unravel the complex driving mechanisms behind blue carbon stock dynamics in Guangdong Province, future studies can adopt an integrated analytical framework combining advanced statistical approaches, such as correlation methods [118], geographical detector model [119], and structural equation model [84], with multi-source geospatial datasets.

To combat climate change, China has integrated blue carbon into the NDCs, which include ocean carbon storage and conservation, restoration, or management of marine ecosystems. Thus, blue carbon ecosystems should be integrated into a carbon market [120], leveraging their superior sequestration efficiency and long-term carbon storage capabilities to achieve climate goals. The China certified emission reduction (CCER) scheme operates as a strategic complementary mechanism to the national carbon emissions trading system [120]. But blue carbon is now excluded from the CCER scheme due to unresolved methodological constraints, and regulatory gaps in coastal carbon accounting. To integrate blue carbon into the CCER scheme, many legal issues, such as marine spatial planning, defined legal rights to implement blue carbon programs, should be solved in the future [120]. Moreover, by integrating international innovations, such as carbon finance and community engagement, Guangdong can become a leader in blue carbon governance. Methods in this study can provide critical technical underpinning for China’s carbon neutrality targets, while the carbon stock datasets here enable prioritized restoration site selection aligning with SDG targets.

To maximize the climate benefits of blue carbon ecosystems, a triple-track planning framework is proposed to integrate: (1) conservation of intact blue carbon ecosystems through legally enforced coastal ecosystem protection zones, prioritizing carbon-rich hotspots; (2) restoration of degraded habitats using nature-based engineering (e.g., planting K. obovata seedlings in existing S. apetala plantations); (3) continuous monitoring with hybrid remote sensing technologies, coupled with AI-driven methods to detect carbon loss.

6. Conclusions

In order to promote the understanding of carbon storage dynamics in mangroves and salt marshes in Guangdong Province, we estimated mangrove and salt marsh carbon storage from 1986 to 2020 by combining the InVEST model and a new mangrove and salt marsh detection framework with Landsat image time series. The main conclusions are as follows:

(1)

The proposed method provided two coastal vegetation detection methods, exploring the potential of utilizing phenological features to improve the identification accuracy. The overall accuracies of the mangrove and salt marsh detection results exceeded 90%, suggesting good consistency with the validation data.

(2)

Over the study period, the mangrove extent showed a trend of decreasing from 1986 to 1995, then fluctuated from 1995 to 2005, and presented an upward trend from 2005 to 2020. The overall trend of the salt marsh area was upward, with small fluctuations from 1986 to 2020.

(3)

The mangrove carbon storage in Guangdong increased from 414.66 $\times$ 10⁴ Mg C to 490.49 $\times$ 10⁴ Mg C during 1986–2020, with Zhanjiang having the largest mangrove carbon storage increase. The distribution pattern of mangrove carbon storage in Guangdong exhibits significant spatial heterogeneity, characterized by higher values in western Guangdong and lower values in eastern Guangdong.

(4)

The salt marsh carbon storage in Guangdong grew from 8.73 $\times$ 10⁴ Mg C in 1986 to 14.39 $\times$ 10⁴ Mg C in 2020, with Zhuhai having the largest increase in salt marsh carbon storage. The high salt marsh carbon storage areas in Guangdong are mainly concentrated in Zhuhai, Jiangmen, Zhanjiang, Shantou, and Jieyang.

(5)

The temporal changes in mangrove and salt marsh carbon storage could be separated into three stages: a decreasing period, a fluctuating period, and a rapidly increasing period, during which the ecological and economic policies played a crucial role. The turning points of carbon storage dynamics for mangroves and salt marshes were consistent with the timing of the policy implementation.

(6)

The multi-decadal blue carbon datasets and their temporal-spatial change analysis results here can offer a scientific basis for coastal ecosystem restoration, nature-based climate solutions, and a decision-support tool for sustainable coastal zone management. In addition, based on Landsat imagery, the proposed method could be employed in other coastal ecotones in China, enabling regional and national blue carbon monitoring programs.