Spatiotemporal Distribution of Mangrove at Kuala Sepetang Forest Reserve, Malaysia, Using Remotely Sensed Data

Abstract

Mangrove plays an important role in mitigating the impact of climate change by sequestering CO₂ as the main greenhouse gas emission. They maintain water quality and clarity, filtering pollutants and trapping sediments originating from land. However, mangroves have been reported to be threatened by land conversion for other activities. The objectives of this chapter were: (1) to derive Normalized Difference Vegetation Index (NDVI) in the study area and (2) to map the status of the mangrove species distribution from year 2004 to 2019. Matang Mangrove Forest Reserves (MMFR) at Kuala Sepetang, Perak, Malaysia were determined using multi-temporal satellite imageries by Landsat TM and Sentinel-2 from year 2004 to 2019. The classification of land use land cover (LULC) was performed using the ISODATA Clustering method and Maximum Likelihood Classifier (MCL) method along with vegetation index differencing technique. The overall classification accuracy was 90% with Kappa statistic accuracy of 0.89. The results indicated changes in area of the mangrove forest to a cleared area occurred at 1340.01–1666.30 ha. (2004–2019). Combinations of these approaches were useful for change detection and an indication of the nature of these changes. The result also revealed that Rhizophora apiculata and Avicennia sonneratia are still being preserved and estimated to be 70% of the total species. Temporal changes of mangrove species for the 15–period showed that the mangrove species of R. apiculata was 58%, A. sonneratia was 6%, Bruguiera parviflora 8% and other mangrove species 28% in year 2019 as compared to 2004 in which R. apiculata was 71%, A. Sonneratia 6%, B. parviflora 10% and other mangrove species 14%. The findings indicated that the status and loss of mangroves due to direct impacts from surrounding land development activities.

1 Introduction

Mangroves are trees that grow near the coast and protect the coastal land against the destruction of tsunamis and storms in the tropics, subtropics and temperate coast (Sheridan and Hays 2003). Mangroves are not only play important roles in ensuring stability and sustainability of coastal ecosystems, but also in fulfilling important socio-economic benefits to coastal communities (Suratman 2008). Mangroves are also important to coastal ecosystems for the shoreline stabilization, ecosystem biology coastal, water quality maintenance, habitat for various aquatic, bird immigrants, recreation, tourism, fishing activities, a supply of forest products and provides the carbon (C) storage in biomass and soil. According to the National Oceanic and Atmospheric Administration (NOAA), mangroves can sequester carbon at the rate of two to four times greater than mature tropical forests. In addition, the soil of mangrove forests stores a significant amount of carbon compared with other types of forest given its high sediment concentration (Zhu et al. 2015). Thus, there is a need to quantify the carbon (C) storage in different components of this ecosystem. According to Azahar et al. (2003), the total area of mangrove forests was approximately 2% (645,852 ha) of the total land area in Malaysia in 1994. The distribution of mangroves area in Malaysia from year 1990 to 2017 are mainly at the west coast of Peninsular Malaysia (17%), east coast of Sabah (61%) and Sarawak coastlines (22%) (Hamdan et al. 2018). The factors that mainly contributed to the changes the biomass was direct conversion to other land uses in the form of aquaculture, developments of infrastructure, industry, and settlements. Besides, factors such as coastal erosion and pollution also affect the mangroves. To control mangrove, the proper and timely diagnosis will facilitate appropriate treatments to maintain a healthy forest of replanted mangrove trees (Ong et al. 2018). They are about over 30 species of mangroves in Malaysia, but the major species are Rhizophora apiculata (Bakau minyak), R. mucronata (Bakau kurap), and Bruguiera parviflora (Lenggadai) (Hamdan et al. 2013).

Several remote sensing methods have been developed for tree species identification such as using multispectral satellite images, hyperspectral images and airborne LiDAR (Abd Latif et al. 2012; Mohd Zaki and Abd Latif 2017). In this chapter, the Landsat TM and Sentinel-2 data were used in mapping the spatial patterns of mangroves with different wavelengths to distinguish biomass estimation and carbon stocks. Normalized Differential Vegetation Index (NDVI) is the most commonly used proxy for biomass estimation. According to (Aziz 2014), NDVI is a valuable tool to quantify vegetation biomass (Mohd Zaki et al. 2020). However, the use of NDVI sometimes underestimate the biomass of some woody mangroves because it represents only canopy properties (Hamdan et al. 2014). The use of remotely sensed images, such as a satellite image, can be cost and time-efficient. Remote sensing techniques and computer science made it possible to analyze mangrove biomass and productivity at present to compare them with global maps obtained with the classical technique (Ibrahim et al. 2015). The assessment of monitoring with satellite sensing based on the monthly, weekly, or daily recording the NDVI in several wavelengths which differ from day to day with the seasonal progress and development urban area. For example, the land use/cover of Pulau Indah, in Selangor, Malaysia was found to have changed dramatically over the year intervals of 1995, 1999, 2005 as detected by Landsat TM satellite driven by an expansion of urban areas and built-up land for West Port and cruise terminal development in the island (Suratman and Ahmad 2012).

2 Materials and Method

2.1 Study Area

The study area was located in the coast of Perak shoreline at Larut Matang, Kuala Sepetang which covers an area of 40,711 ha and along a 52 km stretch of the northern coast of Perak (Fig. 1). Kuala Sepetang consists of two zones namely Kuala Sepetang (North) and Kuala Sepetang (South). The study was carried out in North and South Kuala Sepetang. Kuala Sepetang consist of forest reserves namely Pulau Gula, Cabai Malai, Pulau Kelumpang, Sungai Baharu, Sungai Sepetang, Pulau Kecil, Pulau Selinsing, Jebong, Jebong (Tambahan), Pulau Sangga Besar, Pulau Sangga Kecil, and Telok Kertang a part of Sungai Limau.

Fig. 1

Kuala Sepetang Mangrove Forest in study area (OpenStreetMap 2016)

Firstly, the validation of data from Google Earth and Forest Type in MMFR, then the second step from Sentinel-2 in 2019, Sentinel-2 in 2015, Landsat TM in 2009 and Landsat TM in 2004 satellite imagery data which was known as satellite measured data. There are several steps involved for processing the satellite imagery. First, the pre-processing was carried to the satellite image data. This is followed by an extraction of the image to obtain reflectance data. Then, images were classified using an Iterative Self-Organizing Data Analysis Techniques (ISODATA) algorithm. An ISODATA algorithm requires the user to choose the initial estimates of class means, and then each pixel was assigned to classes with a similar mean. The unsupervised and supervised classification methods were used to differentiate the land cover classes and reference map was used to determine the classification accuracy. Lastly, the maps of the mangroves area were developed. The summary of methodology work flow is shown in Fig. 2.

Fig. 2

Flowchart of the general methodology

2.2 Data Processing

Satellite images of Sentinel-2 (2015 and 2019), Landsat TM (2004 and 2009) were downloaded from the USGS website. Forest type maps in MMFR at Kuala Sepetang, Perak were obtained for validation of the classification data. False color composite image of Kuala Sepetang Mangrove Forest was generated with the band combination of 1, 4, 7 and 2, 3, 8 for Landsat TM and Sentinel-2 images, respectively.

2.3 Calibration and Validation Data

In order to analyze the spectral characteristics of typical land cover types in the study area and assess the accuracy for mangrove forest classification, calibration (or training) and validation (or reference) pixels for all land cover categories were collected from the Landsat TM and Sentinel-2 images by validating data with visual interpretation of Google Earth. Calibration pixels were selected using a visual interpretation, prior knowledge of the study area and from the MMFR map.

2.4 Land Use Land Cover Classification

ISODATA clustering and Maximum Likelihood were used to evaluate the performance of the decision-tree (DT) algorithm for mapping mangrove forests. Unsupervised (ISODATA) and supervised classifications (Maximum Likelihood) were made after radiometric corrections were performed. Unsupervised classification is the process which is computer determines the spectrally separable class and then defines their information value. For supervised classification, analysts identify the information classes of interest in the image. The result of this classification was five classes with different pixel values. Then, these five classes were grouped as R. apiculata, Bruguiera and Avicennia, Dryland and Unproductive area. The classification was made with the reference of the Google Earth and map of MMFR.

2.5 Normalized Difference Vegetation Index (NDVI)

The NDVI was derived from the pre-processed satellites data:

$$ \mathrm{NDVI}=\frac{\rho_{\mathrm{NIR}}-{\rho}_{\mathrm{R}}}{\rho_{\mathrm{R}}+{\rho}_{\mathrm{R}}} $$

where ρ_R and ρ_NIR are the reflectance values for the red and NIR channels, respectively. Mangrove is a part of vegetation that absorbs solar radiation in different bands.

3 Results and Discussion

Figure 3 shows the NDVI images of the Kuala Sepetang mangrove forest in 2004, 2009, 2015, and 2019. The mangrove high vegetation is indicated in dark green with a high value of NDVI meanwhile low vegetation is indicated in white with low value of NDVI.

Fig. 3

NDVI map of Kuala Sepetang Mangrove Forest for years (a) 2004, (b) 2009, (c) 2015, and (d) 2019

Figure 4 shows a graph of the crown density of mangroves area based on the NDVI values in the year 2004 and 2019. The non-vegetation, lower dense, moderate dense, dense and higher dense are increased in 15 years. The area of moderate dense show an increasing area change of 72,341.15 ha compared to dense vegetation of 48,866.69 ha. Hence, this situation can conclude that the NDVI of mangrove areas is increasing at a very high rate in Kuala Sepetang.

Fig. 4

Crown density based on NDVI value for years 2004 and 2019

The mangrove area in the year 2004–2019 tend to be reduced, however the cleared area tends to increase from 1340.01 to 1666.30 ha. These findings are in line with the findings of Giri et al. (2011). Changes also occurred along the intertidal zone due to erosion and occurred in between the mangrove forest due to changes in species and tree harvesting rotation system. The overall change of area during the 15-year period indicated the loss mangrove area at 344.07 ha, while the cleared area is increase at 326.29 ha. Therefore, mangrove forest has lost 34.41 per year ha during 2004–2019 interval. The cleared area increased 32.63 ha per year ha during 2004–2019.

Distribution of mangrove species from the supervised classification from 2004, 2009, 2015 to 2019 are shown in Fig. 5. Results indicated that five major classes are classified comprising of R. apiculata, Bruguiera-Parviflora, Avicennia sonneratia, Dryland Forest, and Unproductive area.

Fig. 5

Mangrove types distribution for years 2004, 2009, 2015, and 2019

Table 1 shows the area of mangroves species from 2004 to 2019. R. apiculata is the most dominant species for the years 2004, 2009, 2015 and 2019. B. parviflora is the second-highest species which consist of 10% in 2004, 11% in 2009, 7% in 2015, and 8% in 2019, respectively. Both R. apiculata and B. parviflora area decreased by 58% and 8% within 15 periods of year. It indicates that the total area of these the dominant species has decreased to 23,465.09 ha in 2019 from 23,482.87 ha in 2004. The Dryland forest had an increase of 21% in the year 2019 than 8% in the year 2004. Analysis of the 2004 Landsat TM image estimated the total mangrove area is 23,482.87 ha, in which the areas of. R. apiculata and B. parviflora were 16,767.7 ha and 2252.79 ha respectively. Unproductive area covered area of 1340.01 ha, Dryland Forest 1771.74 ha, and Avicennia-Sonneratia 1350.63 ha. For the analysis of the 2009 Landsat TM image, it is estimated that the total mangrove area is 24,102.26 ha, in which the areas of R. apiculata was 15,727.4 ha, B. parviflora was 2568.06 ha, Unproductive area was 1215.72 ha, Dryland Forest was 2337.12 ha and Avicennia-Sonneratia was 2253.96 ha. Meanwhile, analysis of the 2015 Sentinel-2 image indicated the mangrove area at 23,465.25 ha, with the highest proportion recorded by R. apiculata at 15,730.5 ha, followed by B. parviflora, Unproductive area, Dryland Forest, and Avicennia-Sonneratia at 1582.85 ha, 2002.79 ha, 2327.55 ha and 1821.56 ha respectively. As for 2019s Sentinel-2, R. apiculata showed the value at 13,706.7 ha, followed by B. parviflora, Unproductive area, Dryland Forest, and Avicennia-Sonneratia at 1800.13 ha, 1666.3 ha, 4933.04 ha, and 1358.92 ha respectively. Changes in the mangrove species forest types from 2004 to 2019 are indicated in Table 1. Based on the table, 13% of loss in mangrove area is from Rhizophora species during the 2004–2019 study period. Rhizophora forest area decreased by 4% from 2004 to 2009, and was constantly decreasing in 2019. Generally, from 2004 to 2019, Rhizophora, B. parviflora had a loss of mangrove species at 3061 and 452.66 ha, while the Unproductive area, Dryland forest, and Avicennia-Sonneratia area increased at approximately 326.29, 316.13, and 8.29 ha.

Table 1 Area of mangrove land use and percentage changes

Figure 6 shows the percentage area changes of different land use category. The spatial distribution of mangrove species classification at Kuala Sepetang indicated that Rhizophora species dominated the study area image. In Kuala Sepetang, Rhizophora forest is a major forest type that comprises about 71% of the total forested area in the year 2019. Avicennia-Sonneratia recorded the lowest value at 6% in the year 2004, 9% in the year 2009, 8% in the year 2015, and 6% in the year 2019.

Fig. 6

Percentage of species distribution

Referring to Table 1, the area of mangrove changes that have occurred in Rhizophora forest loss at 13% over the study period from 2004 to 2019. Rhizophora forest area decreased by 4% from 2004 to 2009 and constantly decreasing to 2019. Generally, from 2004 to 2019, Rhizophora, Bruguiera-Parviflora had a loss of mangrove species at 3061 and 452.66 ha, while the Unproductive area, Dryland forest, and Avicennia-Sonneratia area increased approximately 326.29, 316.13, and 8.29 ha.

For the accuracy assessment in this study, simple random sampling was adopted. A total of 37, 45, 38, and 41 random pixels were extracted from the classified images of 2004, 2009, 2015, and 2019. The overall accuracies for 2004, 2009, 2015, and 2019 were 95%, 92%, 94%, and 94% respectively. Kappa statistics were 90%, 86%, 89%, and 90%. Kappa statistic was implemented to evaluate the accuracy of change detection and classification maps by measuring the agreement between the two images. User’s and producer’s accuracies of the individual class were relatively high, ranging from 34% to 100% which indicated a good agreement between the reference data and thematic maps generated from the images.

4 Conclusion

Understanding of spatial distribution of Kuala Sepetang Mangrove Forest helps to determine the location of deforestation, transitional forest, and erosion for the next future mangrove forest management. Hence, remote sensing application is a reliable method to monitor deforestation by comparing the temporal data by daily, monthly, yearly for large mangrove forest areas. In this study, mapping of mangrove area can contribute to the effectiveness of the area management as a whole since the conventional method of ground inventory for the mapping area is tremendously difficult, time-consuming and costly. Continuous monitoring of mangrove status using moderate resolution remote sensing imageries is essential for the large mangrove area for protection and restrictive productive forest. Hence, remote sensing is the best technique to estimate the changes of mangrove distribution. In conclusion, the present information on the status and changes of mangrove forest area will be useful for further studies and to monitor the mangrove ecosystem. It will help to formulate the strategic plans and afforestation of mangroves in the study area.