Mapping mangrove above ground carbon stock of benoa bay bali using sentinel-1 satellite imagery

Mangrove forest ecosystems distributed in tropical countries play an important role in carbon sequestration. The presence of mangrove forests is estimated to be capable of reducing CO2 levels caused by air pollution. The process of photosynthesis absorbs CO2 gas from the atmosphere and stores it as organic matter in the form of plant biomass. The amount of carbon stock stored in mangrove forests can be estimated using a regression model based on satellite imagery pixel values and above ground biomass (AGB) measurements. This study used the backscattering value of SAR Sentinel 1 images and field AGB measurements to map carbon stocks in the mangrove forest ecosystem of Benoa Bay, Bali. The processed backscattering value is the result of VV and VH polarization in the SAR Sentinel 1 image acquired in 2022. Meanwhile, AGB is calculated using the allometric equation based on the mangrove’s species found at the study site. The biomass at the study location was estimated to be in the range of 64.18 Mg/Ha to 175.24 Mg/Ha, with an average of 115.34 ± 25.33 Mg/Ha. Therefore, the carbon stock of Benoa Bay can be estimated to have values in the range of 30.16 MgC/Ha to 82.36 MgC/Ha, with an average of 54.21 ± 11.09 MgC/Ha. In terms of estimating AGB biomass based on SAR pixel values, the VH polarization produces a better model (R2 = 0.635) than the VV polarization.


Introduction
Mangrove ecosystems, like other coastal ecosystems, absorb and store carbon.These functions have been considered to play an important role in reducing CO2 levels in the Earth's atmosphere.Carbondioxide (CO2) absorbed during photosynthesis is used for the growth and development of mangrove plants [1,2].Photosynthesis absorbs CO2 gas from the atmosphere and stores it as organic matter in the form of plant biomass.The amount of organic matter stored in mangrove biomass per unit area and per period of time is the basis of mangrove forests' carbon-storage productivity [3].Despite accounting for less than 2% of marine environments, mangroves are responsible for 10 to 15% of global carbon burial [4].Mangroves are carbon-absorbing machines that absorb CO2 while growing and sink it to the seafloor when they die.Remote sensing sensors acquire electromagnetic waves reflected on the Earth's surface by vegetation and non-vegetation objects.Using a vegetation index, a subsequent examination of the images produced by the sensors is able to differentiate vegetation from other objects.Furthermore, a mathematical equation that combines several remote sensing bands to produce a single index value yields the vegetation index value.Other functions of the vegetation index include calculating percentages of vegetation cover and plant biomass, as well as estimating carbon sequestration and photosynthetic capacity.
From a remote sensing standpoint, above ground biomass (AGB) is recognized as an important ecological indicator for determining the environmental condition of mangrove forests.AGB is defined as the dry weight of organic matter produced and stored in above-ground living vegetation, including stems, branches, bark, seeds, and leaves [5].Estimation of AGB has received attention in recent decades as a result of growing awareness of global warming caused by climate change and the critical role of mangrove forest biomass in the absorption and release of greenhouse gas carbon due to deforestation [6].In addition, AGB can be estimated using destructive sampling methods, in which trees are felled and weighed, or non-destructive methods such as estimation based on allometric equations developed using dimensional trees such as diameter at breast height (dbh) and tree height.Recent research has focused on developing mathematical models that can be used to estimate AGB based on the satellite image pixel values.
The most widely used method is to extract pixel values of Visible-Infra Red (V-IR) bands (EM wave lenght 600 nm-950 nm) from medium resolution satellite images such as Landsat ETM 7, Landsat 8 OLI, ALOS, or the most recent Sentinel 2A/2B [7][8][9].However, detection via the above mentioned passive satellites remote sensing is frequently limited by weather and thus cannot penetrate cloud cover.Other satellite sensors for example as SAR (Synthetic Aperture Radar) images offer an alternative for mangrove mapping.Synthetic Aperture Radar (SAR) is an active remote sensing sensor that generates its own energy to transmit microwave (radio) signals in order to image a specific scene.It is not dependent on sunlight and can thus operate at any time of day or night.In addition, SAR is regarded as an all-weather imaging sensor because it is not affected by cloud cover and can penetrate dense canopy cover.Radar sensors work in the microwave portion of the electromagnetic spectrum, with frequencies ranging from 0.3 GHz to 300 GHz, or 1 m to 1 mm in wavelength [10,11].
Sentinel-1 provides SAR images with a medium to high geometric resolution (5-20 m on the ground) in the C-band.It can send a signal in either horizontal (VH) or vertical (VV) polarisation and receive in both.Complex values can be found in Dual Polarisation Level-1 Single Look Complex (SLC) products.The inter-channel phase information, in addition to the backscatter intensity that can be measured from each single polarisation, allows for a more in-depth analysis of backscattering properties [12].Several studies have demonstrated the use of Sentinel-1 SAR images to model the estimation of mangrove biomass, with varying results.A previous study in the Philippines found that VH polarisation has a better correlation with biomass than VV polarisation (Pearson correlation = 0.48-0.51)[13].Meanwhile, another study in the West African dryland discovered that Sentinel-1 had insufficient accuracy in mapping AGB biomass, with RMSE = 78.6 and MAE = 25.6 [14].Furthermore, several studies in China and Indonesia have attempted to estimate mangrove biomass using a combination of Sentinel 1 and Sentinel 2 data with several classification algorithms.The results obtained were representative with a coefficient of determination (R 2 ) = 0.787 and (R 2 ) = 0.821 respectively [15,16].
Questions have been raised regarding the accuracy and validity of using SAR images to estimate vegetation biomass.However, very few studies have investigated the association between SAR data of Sentinel-1 and specific vegetation biomass in particular mangroves.Previously published studies are limited to non-mangroves vegetation.Therefore this study aims to measure the accuracy of Sentinel-1 SAR data in mapping AGB biomass in a mangrove ecosystem.Furthermore, the results will be used to estimate the amount of carbon stock in the study area.

Study Location
This study was conducted in August to December 2021.The location of the study is the mangrove forest of Tahura Ngurah Rai Benoa Bay.Ngurah Rai Tahura is administratively located in Kuta District Badung Regency South Denpasar Bali Province, Indonesia.The geographic coordinate of the location is 115°9'31'' to 115°14'07'' East and 08°42'15'' to 08°47'27''South with an estimated area of 1,373.50Ha.Thirty sampling plots measuring 10x10 m for field measurement of biomass and carbon stock were established.The plots were distributed using a purposive sampling method based on representation and access to the area (Figure 1).

Image Processing
The Sentinel-1A SAR image used in this research was acquired on October 27, 2021 with Ground Area Detected (GRD) processing level.The image was downloaded from https://search.asf.alaska.edu.The specification of Sentinel-1 imagery data is explained in Table 1.The image was captured in Interferometric Wide Swath mode with VV and VH polarization and a pixel size of 10 m.The Sentinel application (SNAP) was used for image pre-processing, which included image sub-setting, orbit file application, radiometric calibration, speckle reduction, and Range-Doppler terrain correction.Radiometric image calibration converts the SAR image's pixel values to those that represent radar backscatter from reflecting surfaces, as well as incident angle effects.This process converts the SAR image's pixel values into radar intensity backscattering coefficients (σ).Moreover, speckle reduction using the Refined Lee Filter is performed to reduce the effect of speckles on the image to enable better backscatter analysis and interpretation.Finally, Range-Doppler terrain correction is used to re-project the SAR image into a map projection.Phase error 5°

Field Survey
The non-destructive sampling method was applied to measure biomass of mangrove vegetations.A total of 30 sampling points were established for biomass measurement.At each sampling point, identification of mangrove species and measurement of the circumference of the stem at breast height (GBH) were carried out, which were then converted into stem diameters (DBH) (Figure 2).Biomass estimation was carried out using equations according to the type of mangrove found (Equation 1-5).Furthermore, biomass values were converted into carbon stock values (Equation 6).

Above Ground Biomass and Carbon Stock
Benoa Bay is a tidal waterway on Bali's southern coast.Due to the reclamation of Serangan Island, the bay's mouth has narrowed by up to 75%, resulting in a semi-enclosed bay typology.Because it is a shallow estuary ecosystem with a number of rivers flowing into it (Tukad Punggawa, Tukad Balian, Tukad Badung, Tukad Mati, Tukad Soma, Tukad Mumbul, and Tukad Bualu), Benoa Bay is a unique ecosystem area.This esthetic condition distinguishes Benoa Bay from other shallow coastal waters, where a number of ecologically important communities, particularly mangrove communities, seagrass beds, macrozoobenthos, and infauna components with high abundance and diversity exist.This location is strategically situated at the epicenter of Bali's golden triangle, which includes the Sanur -Kuta -Nusa Dua area as a tourism area that has developed and advanced as a centre for economic growth based on tourism, trade, and services.The beaches in Benoa Bay's coastal area have nearly identical physical characteristics, with the exception of the mangrove beach, which is distinguished by mangrove vegetation and its associations in the form of alluvial and muddy deposits.On the coasts of Benoa and Serangan, there are also sandy sloping beaches.In general, the coastal area of Benoa Bay is lowland with a slope of 0 -2%.Mangrove forests grow on the sides of Benoa Bay from Tukad Loloan to Tanjung Benoa.The mangrove forest area in Benoa Bay accounts for 62.9% of Bali's total mangrove forest area.According to Minister of Forestry Decree Number: 544/Kpts-II/1993 dated September 25, 1993, the mangrove forest area totals was 1373.5 Ha [17].

Figure 3. Results of AGB Biomass and Carbon Stock Estimation of Mangrove Forest of Benoa Bay
Bali The findings of this study's field survey identified several dominant mangrove species, including Rhizophora mucronata, Rhizophora apiculata, Sonneratia alba and Bruguiera gymnorhiza.The mangrove species discovered during the survey are dominant species that play an important role in the mangrove ecosystem.To adapt to the high salinity tidal environment, mangrove vegetations develop morphological specializations such as aerial roots and other special physiological mechanisms, particularly for salt excretion [18].Taxonomically, true mangrove species are distinct from land plants because they are only found in mangrove forests and form pure stands, never mixing with land plants [19].
The above-ground biomass (AGB) of mangrove vegetation was measured in order to estimate carbon stocks and calculate CO2 sequestration potential.The substrate in the mangrove growing area influences the amount of biomass produced by mangroves.The biomass content of tree stands increases with age.The biomass at the study location was estimated to be in the range of 64.18 Mg/Ha to 175.24 Mg/Ha, with an average of 115.34 ± 25.33 Mg/Ha based on the measurement results of 864 mangrove stands from 30 sampling plots (Figure 1).As a result, the carbon stock of Benoa Bay can be estimated to have values in the range of 30.16MgC/Ha to 82.36 MgC/Ha, with an average of 54.21 ± 11.09 MgC/Ha.
The estimated AGB carbon stock of mangrove forest in Benoa Bay was relatively higher compare to several rehabilitated mangrove forest in Indonesia, for example in Tanakeke South Sulawesi (37.6 MgC/Ha), Deli Serdang North Sumatra (40-50 MgC/Ha) and Bregasmalang Central Java .The carbon stock in the mangrove forests of Benoa Bay Bali, however, is significantly lower when compared to natural mangrove forests in relatively undisturbed environments.Figure 4 compares carbon stocks from other locations in Indonesia and other Asian countries [3,20,21,[23][24][25][26][27][28]

Image Processing and AGB Carbon Stock Model
The processing of Sentinel-1 satellite imagery was the next analysis performed in this study.Backscattering values on the vertical (VV) and horizontal (VH) polarizations can be obtained after the pre-processing stage.The VV and VH pixel values were then extracted from 30 points according to the coordinates of the carbon stock measurement plot.The average backscatter value at VV polarization was -7.86 dB which for the highest value was -5.10 dB and the lowest value is -9.99 dB.Whereas the average value of backscatter in VH polarization was -13.43 dB which for the highest value was -11.21 dB and the lowest value is -16.44 dB. Figure 5 represents the pixel values obtained from the sampling plots.The Sentinel-1 SAR instruments operate in single (HH or VV) and dual polarisation (HH+HV or VV+VH), with one transmit chain (switchable to H or V) and two parallel receive chains for H and V polarisation.VV is a mode that transmits and receives vertical waves to generate the SAR image, whereas VH is a mode that transmits and receives horizontal waves to generate the SAR image.Sentinel-1 VV and VH backscatter, in particular, are known to be sensitive to changes in vegetation conditions as a result of natural or anthropogenic disturbance such as storm surge and land conversion [29,30].In 2021, a research at the same location using ALOS PALSAR-2 (2020) was conducted.The results for calculating the average AGB obtained were 133.74 Mg/Ha, which was lower than the results of this study, which were 160.53 Mg/Ha.This difference may be due to differences in spatial resolution between ALOS PALSAR imagery (7 meters) and Sentinel-1 SAR (10 meters).However, compare to its other predecessor satellites such as ERS-1, ERS-2, JERS, SIR-C/X-SAR, RADARSAT, SRTM, EnviSAT-ASAR, RADARSAR-II, LIGHTSAR [10], Sentinel-1 SAR offers better resolution thus become more reliable to be used for medium scale vegetation mapping.
The short revisit time (temporal resolution) of Sentinel-1 images is advantageous in forest change detection applications.Furthermore, the aforementioned images are free to use and easily accessible.Longer wavelength SAR (e.g., ALOS PALSAR products) is more suitable for forest applications because it penetrates deeper than shorter wavelength SAR (e.g., Sentinel-1 SAR products).Belowcanopy information such as tree trunks/stems and underlying undergrowth vegetation, deadwood (including standing deadwood), and soil organic matter all contribute significantly to the amount of aboveground biomass and carbon pools in tropical forests [32,33].
Mangrove forests play a crucial role in the global carbon cycle by sequestering and storing large amounts of carbon in their biomass and soils.These coastal forests are highly efficient at capturing and storing carbon dioxide from the atmosphere, making them important carbon sinks.Studies have shown that mangrove carbon stocks vary across different regions and countries [3,34].Global patterns indicate that mangrove carbon stocks are higher in the tropics compared to temperate regions.This is due to the favorable conditions for mangrove growth, such as high temperatures and nutrient availability.However, there are also variations within tropical regions, with differences in carbon stocks observed between hemispheres and latitudes.The loss of mangrove forests has significant implications for carbon stocks and the global carbon cycle.Deforestation and degradation of mangroves result in the release of stored carbon into the atmosphere, contributing to greenhouse gas emissions.A study estimated that between 2000 and 2012, global mangrove deforestation led to the loss of approximately 122 million metric tons of carbon and potential emissions of 316 million metric tons of CO2 [35,36].The consequences of mangrove carbon stock loss extend beyond the carbon cycle.Mangroves provide numerous ecosystem services, including coastal protection, habitat for biodiversity, and support for local communities.The destruction of mangroves not only disrupts these services but also leads to the loss of valuable carbon sinks.
The use of multi-temporal satellite images for example Landsat TM-8, Sentinel-1 and Sentinel-2 are important for mangrove monitoring.Image processing can detect various spatial environmental change parameters over multiple time periods.Other parameters, such as carbon content, require field data.This research demonstrates how a combination of field data and image processing results can be used to create a spatial database with high accuracy.To produce a more representative model of estimated biomass and carbon stock, sample selection must be varied to account for differences in density or canopy cover.

Conclusions
In conclusion, mangrove carbon stocks are an important component of the global carbon cycle.Their loss contributes to greenhouse gas emissions and has negative implications for coastal ecosystems and communities.This study has demonstrated the application of SAR data to identified the extent of mangrove forest, and with the combination of field survey to estimate AGB and carbon stock of Benoa Bay Bali.The biomass at the study location was estimated to be in the range of 64.18 Mg/Ha to 175.24 Mg/Ha, with an average of 115.34 ± 25.33 Mg/Ha.Therefore, the carbon stock of Benoa Bay can be estimated to have values in the range of 30.16MgC/Ha to 82.36 MgC/Ha, with an average of 54.21 ± 11.09 MgC/Ha.In terms of estimating AGB biomass based on SAR pixel values, the VH polarization produces a better model (R 2 = 0.635) than the VV polarization.Sample selection must be varied to account for differences in density or canopy cover in order to produce a more representative model of estimated biomass and carbon stock.Additional research is required to better understand mangrove carbon stocks and their role in climate change mitigation.Mangrove forest protection and restoration are critical for preserving their carbon sequestration potential and the valuable ecosystem services they provide.

Figure 1 .
Figure 1.Study Location the Mangrove Forest of Ngurah Rai Tahura Benoa Bay Bali

Figure 5 .
Figure 5. Pixel Value of VV and VH Polarization of Sentinel-1 SAR Data Benoa Bay Bali

Figure 6 .
Figure 6.Regression Results of SAR Pixel Value and Mangrove's AGB Biomass of Benoa Bay Bali

Table 1 .
Specification of Sentinel-1 imagery data