Spatial mapping and temporal dynamics of mangrove: a case study in 'pro-mangrove' villages,

. This study used remote sensing technology, specifically Landsat 5 TM and Sentinel-2 MSI images, to map mangrove distribution in 'Pro Mangrove' Villages, Indragiri Hilir District, Riau Province, Indonesia, from 1989 to 2021. The multi-resolution segmentation (MRS) and Random Forest (RF) algorithms were used to identify changes in land cover over this period. The study found a notable increase of 482.62 hectares in mangrove area over the 32-year period, with a recovery clear from 2014, possibly due to rising mangrove conservation awareness. The study also identified possible disturbances such as exploitation before 1989. The Random Forest algorithm proved effective in mapping mangroves and surrounding land cover. The study underscores the utility of remote sensing technology in tracking mangrove dynamics, which is fundamental for informed conservation and sustainable land management strategies. The findings are expected to guide local authorities, conservation entities, and other stakeholders in devising strong mangrove conservation and management strategies to curb deforestation and promote sustainable land use practices in the Riau Province.


Introduction
Mangroves are an important ecosystem playing a pivotal role in the global carbon cycle, coastal protection, and the provision of habitats for a myriad of species.They also serve as significant indicators of environmental changes and the impact of human activities, such as pollution or deforestation, which can be gauged by tracking the health and extent of mangroves [1].However, the investigation and mapping of mangroves using traditional methods often encounter challenges due to the remote locations of mangroves and their unique environmental conditions [2].
Globally, mangroves span an area of 137,760 km², distributed across 118 countries [3].Among these, Indonesia stands out as it boasts the world's largest mangrove coverage.Between 1996 and 2020 [4], Indonesia experienced a deforestation of mangroves totalling 1,739.04km², with the Riau Province seeing a significant decline due to factors such as land conversion for agriculture, palm oil plantations, oil and gas exploitation, and urban development [5].
Given the challenges associated with traditional methods, remote sensing emerges as an effective solution for mangrove mapping.Data from this technology, especially from Sentinel-2, has been widely used for mangrove mapping, providing clearer insights into the dynamics of mangrove changes.This research aims to map the distribution of mangroves in 'Pro Mangrove' Villages, in Indragiri Hilir District, Riau Province, Indonesia, identify changes in land cover from 1989 to 2021, and analyse the factors contributing to these changes using object-oriented classification of satellite imagery, as previously described.
The procedure involves acquiring Landsat 5 TM and Sentinel-2 MSI images, segmenting them into homogeneous entities using the multi-resolution segmentation (MRS) algorithm [6], and categorizing the data using the Random Forest (RF) algorithm [7].This method enables a detailed examination of mangrove distribution and the factors contributing to land cover changes.
The results from this study are expected to recommend local authorities, conservation organizations, and other stakeholders on the conservation and management of mangroves in the region.Through a better understanding of the dynamics of mangrove changes, informed strategies can be developed to mitigate mangrove deforestation and support sustainable land use practices in Indonesia, particularly in the Riau Province and 'Pro Mangrove' Villages.

Material and methods
This research is in two sub-districts in Indragiri Hilir Regency, encompassing six promangrove villages: Kelurahan Sapat, Desa Sungai Piyai, Perigi Raja, Tanjungmelayu (Subdistric Kuala Indragiri), Desa Igal, and Desa Pulau Cawan (Sub-district Mandah) (Figure 1).The term 'pro mangrove' refers to villages that have made mangroves a key pillar in their community development.Most residents work as coconut farmers, while fishing is the second largest profession, with dominant fishing gear such as nets and longlines.The community has long used mangroves for daily needs such as firewood and construction, and as an economic resource through mangrove biota like mud crabs, snails, and shellfish.Additionally, pesut (Orcaella brevirostris) or aquatic mammals around Pulau Cawan has been an interesting phenomenon reported since 2016, with the distribution of pesut in coastal areas influenced by seasonal factors, water levels, and food availability.Nine mangrove families have been identified in the research location, including the Acanthaceae family (4 species), Arecaceae (1 species), Combretaceae (2 species), Euphorbiaceae (1 species), Lythraceae (3 species), Meliaceae (2 species), Pteridaceae (4 species), Rhizophoraceae (8 species), and Rubiaceae (1 species).
The research methods used in this study involves several key steps: (a) Data Collection: The study sourced data from two primary channels, namely the Landsat 5 TM Level 2A satellite imagery captured on May 24, 1989, and September 13, 1989, and the Sentinel-2 MSI Level-2A satellite imagery captured on December 1, 2021.(b) Data Processing: To mitigate the challenges posed by cloud cover, the study combined images from different dates to unveil areas previously obscured by clouds or shadows.The Level 2A data underwent atmospheric correction to enhance its analytical precision.(c) Synthetic Band Creation [8]: In addition to the original bands from the satellite imagery, the study introduced various synthetic bands as auxiliary input layers.These include vegetation indices such as NDVI and SAVI, the built-up area index NDBI, the water index MNDWI, and diverse spectral indices pertinent to mangroves such as MVI, CMRI, MI, NDMI, and MMRI.(d) Object-Based Classification [9]: This approach aggregates pixels through an image segmentation procedure instead of operating on singular pixels.It considers spectral and textural data, along with shape attributes and environmental correlations.(i) Multi-Resolution Segmentation (MRS): The MRS algorithm amalgamates pixels into entities until they satisfy user-specified homogeneity prerequisites.All image bands participate in this segmentation procedure.(ii) Random Forest (RF) Classification: The RF algorithm uses multiple decision trees to predict the final class of entities.The ultimate predictions are determined through majority voting from all available decision trees [7,8,10].(e) Classification Evaluation: Classification outcomes are tested using an error matrix to obtain user, producer, and overall accuracy.(f) Change Detection: Change detection aids in understanding the dynamics of mangrove coverage and distribution.Each data series in this study is assigned an identifier, and mangrove changes are assessed by aggregating all the identifiers obtained [5].The annual rate of mangrove loss is calculated using Puyravaud's [11] formula, using mangrove coverage data at two time points.

Mangrove distribution
The study successfully mapped the distribution of mangroves, categorizing them into two classes: nipa (Nypa fruticans Wurmb) and other types of mangroves.Nipa thrives in mangrove environments, favoring brackish waters such as estuaries, tidal rivers, muddy estuaries (semi-liquid), and coastlines [12].Mangroves grow along the coastlines and rivers in the study area [13], with some extending up to 21 kilometers upstream from the coastline.
The thickness of the mangrove vegetation varies significantly, ranging from a few meters to over 3 km in some areas, such as on both sides of Batang Tuaka (Village Sapat and Tanjungmelayu).In Beting Island (Village Tanjungmelayu), almost the entire island is covered by mangroves, including Nypa fruticans Wurmb and other types of mangroves.These habitats are found in Villages Sapat, Tanjungmelayu, and Perigi Raja.However, in Villages Pulau Cawan and Igal, the spread and area of nipa are less, to the extent that Landsat 5 TM cannot detect nipa mixed with other mangrove vegetation (Figure 2).The mangrove classification yielded satisfactory accuracy levels, with producer and user accuracy each exceeding 80%, except for the nipa class with user accuracy below 80%, while omission and commission errors were estimated between 8 -25%.

Fig 2. Mangrove land cover classification result
The Producer's Accuracy measures the map accuracy from the map maker's point of view, showing how often real features on the ground are correctly classified on the map or the probability that a certain land cover of an area on the ground is classified as such.The Producer's Accuracy is the complement of the Omission Error, which is calculated as Producer's Accuracy = 100% -Omission Error.But the User's Accuracy measures the accuracy from the map user's point of view, showing how often the class on the map will be present on the ground.This is called reliability.The User's Accuracy is the complement of the Commission Error, which is calculated as User's Accuracy = 100% -Commission Error.
Factors such as imprecise designation of pixel training areas and spectral similarity between nipa and other mangroves could influence errors [14].Spectral similarity when selecting object samples in creating specific class rules posed another challenge.Another factor could be high surface water level fluctuations potentially causing errors, and land cover around mangroves in the same morphological area, namely tidal areas [15].Still, producer accuracy values show that the use of the Random Forest algorithm in object-based classification can be an effective alternative technique in mapping mangroves and surrounding land cover [8] , although there are still errors in separating mangrove classes from other classes [7].In 1989, the total area of mangroves in the study area was 20,681.14hectares, with the nipa community comprising 57.1% and other mangroves making up the rest.By 2021, the total mangrove area had gone up to 21,163.75 hectares, with the nipa community dominating 62.8% of the area and other types of mangrove's accounting for the rest.Table 2 explains the area and changes in the distribution and expansion of mangroves over time in each Pro Mangrove village/sub-district.In 2021, the Ministry of Environment and Forestry (KLHK) published an update of the national mangrove map (PMN), which assessed the total national mangrove area at 3,364,080 hectares.The research site covers 0.63% of the total national mangrove area, equivalent to 21,163.75 hectares.This figure indicates a difference of only 448.08 hectares (<2%) from the mangrove area calculation in this research, as presented in Table 3.Despite differences in resolution and data collection techniques, this comparison shows consistency between the results and national data.
The PMN data derives from images with a higher spatial resolution, namely SPOT 6 and 7.These images, with multi-spectral and panchromatic band diffusion, can present surface objects with a level of detail of 1.5 m. this research used Sentinel-2 MSI satellite images with image pixels re-sampled to 10 meters.The delineation technique of mangroves in PMN was carried out manually through visual interpretation and digitization.The PMN shows smoother detail, especially for water body objects.This indicates that the PMN data has a higher resolution and may be more accurate in displaying small details, such as tributaries and boundaries of mangrove areas.

Mangrove changes
The analysis of mangrove changes dynamics from the two data sources used shows relatively static changes, although changes (gain and loss) occurred over 32 years (Figure 3).The increase in mangrove area over three decades amounted to 223.31 hectares, with percentage change values ranging from 0.6 to 12.8.Negative values show changes towards the reduction of mangrove area (Table 4).Meanwhile, according to Puyravaud [11], the rate of change in mangrove area in the study location is low, at 0.03% per year.Most mangrove gains are primarily seen in Desa Sapat and Igal, while the most significant losses in mangrove areas occur predominantly in Desa Sapat and Perigi Raja. Figure 7 presents the spatial distribution of these mangrove changes, along with detailed transformations of mangroves into other land cover classes.These detailed changes in land cover highlight the conversion of mangroves, including nipa, into different types of land cover such as oil palm and coconut plantations [5].
Extensive exploitation has occurred throughout the mangrove area in the study location.This research is significantly aided by very high-resolution satellite images provided through the Google Earth Pro (GEP) service, which bolster confidence regarding these exploitation events.Two excerpts from the GEP data are presented here.The first excerpt illustrates the evolution of changes on Sangkar Ayam Island.In 2002, Pulau Sangkar Ayam kept its original and pristine condition.However, over the next 12 years, a noticeable reduction in the mangrove stands on the island occurred, resulting in a smaller Pulau Sangkar Ayam.By 2020, two basins had formed in the northern part of the island and one basin in the southern part.

Conclusion
This research successfully mapped the distribution of mangroves in Indragiri Hilir Regency, Riau Province, Indonesia, between 1989 and 2021.The study revealed an increase in the mangrove area by 482.62 hectares over this 32-year period, with fluctuations over the years.The recovery of mangrove cover was seen in 2014, supported by observation data from 2016, suggesting that awareness of the importance of mangroves has contributed to the recovery and improvement of mangrove conditions in the study location.The remote sensing-based methodology and Random Forest algorithm were used to map mangroves with high accuracy.To support conservation policies and natural resource management in Indonesia, it is recommended that local authorities and conservation organizations use the findings of this research to develop future mangrove management strategies.

Fig 1 .
Fig 1.The research site is in six Pro Mangrove villages in Indragiri Hilir Regency

Fig 3 .
Fig 3. Mangrove changes status over 1989 -2021 In one of the estuary areas of Batang Pelandung (the second excerpt), the 2002 image recording reveals many exploited mangrove areas.Some of these areas had been encroached upon by the community for direct use of mangrove wood.Fast forward 12 years, and many areas were now covered by mangrove vegetation, although some exploited areas remained.The image from 2016 showed an improvement in mangrove coverage compared to 2014.The most recent recording from 2020 indicates that most study areas have better mangrove coverage compared to previous years.

Table 1 .
Mangrove mapping accuracy assessment result

Table 2 .
Mangrove area (hectares) per village in 1989 and 2021 over Pro-Mangrove Village/Subdistrict Comparison of mangrove area and mangrove density condition based on PMN in 2021 with this research

Table 4 .
Status and area (hectares) of mangrove changes from 1989 -2021