remote sensing models used for mapping and

REMOTE SENSING MODELS USED FOR MAPPING AND ESTIMATION OF BLUE CARBON BIOMASS IN SEAGRASS-MANGROVE HABITATS: A REVIEW Dalhatu Aliyu Sani1, 2, 3, Mazlan Hashim1, 2, *, Mohamad Shawkat Hossain 4 1

Geoscience & Digital Earth Centre (INSTEG), Research Institute for Sustainable Environment (RISE), Universiti Teknologi Malaysia, Johor Bahru, Malaysia 2 Faculty of Built Environment & Surveying, Universiti Teknologi Malaysia, Johor Bahru, Malaysia 3 Department of Geography, Yusuf Maitama Sule University, Kano, Nigeria 4 Universiti Malaysia Terengganu, Kuala Terengganu, Malaysia *corresponding author: E-Mail [email protected]

Abstract: Blue carbon ecosystems (seagrasses, mangroves and salt marches) play important role in the global carbon cycle. This makes them crucial and as such, more attention should be paid to these ecosystems. Recent studies revealed that remote sensing system has been utilized to map and estimate seagrass-mangrove biomass over large areas to circumvent time consumption and excessive cost of conventional techniques. Several remote sensing models have been successful applied for detecting seagrass-mangrove biomass and carbon. In this article, acoustic remote sensing and models with their strength, limitations and location applied were reviewed to enable users have the holistic view of their competences. Thus, achievements and gaps in the research domain were likewise highlighted in the article with a view to presenting the future prospects in suitable remote sensing models for mapping and estimating seagrass-mangrove biomass, and other biophysical parameters towards realizing the United Nations Sustainable Development Goals (SDG 14). KEYWORDS: acoustic methods, water column correction, bottom reflectance index 1. INTRODUCTION Blue carbon pools as reservoirs store and emit carbon. Global carbon pools comprise vegetation, atmosphere, ocean and the soil (Macreadie et al., 2017). The major blue carbon components comprise seagrass, saltmarsh and mangrove costal habitats (Pendleton et al., 2012). These components store and sequester tangible quantity of carbon in their above-ground biomass (AGB), below-ground biomass (BGB) and soil. More so, they are regarded as a cost-effective means of mitigating the effects of climate change such as global warming. Thus, mapping and modelling of seagrass-mangrove habitats is crucial in determining the total amount of carbon stock in a study area (Mishra et al., 2017; Sousa et al., 2017). Long-living mangrove forests hold an important quantity of autotrophic carbon when compared with seagrass meadows (Whitfield, 2017). Organic carbon may be exported to the adjacent habitats and influence carbon cycle (Thompson et al., 2017), while recent studies focused mainly on partial estimate and modelling of only single component/variable of blue carbon include AGB, BGB or sediment (Congdon et al., 2017; Misbari and Hashim, 2016a), without taking the three variables into account. This perhaps have resulted carbon estimation to be unrepresentative. Hence, it is indispensable to examine achievements and gaps via this review. When a blue carbon habitat represents more than single variable, traditional remote sensing (RS) is similar to a one-eyed man, i.e., majority of researches assess only one variable (Benson et al., 2017; Misbari and Hashim, 2016a) and consequently comprehending the worldwide blue carbon cycle will remain far away from practical. The RS ability can only observe the biophysical parameters, due to an inherent restraint of RS instruments (Duveiller et al., 2018). Ultimately, precise estimation of under-ground organic carbon remains challenging. Furthermore, due to lack of researches in long-term bases, that entails huge project cost as well as international collaborations, regional blue carbon dynamics assessment and regional threats continued to remained unaddressed. However, AGB, BGB, and carbon can be measured (indirect measurements) through RS models in order to overcome the inherent restraint of RS instruments. Similarly, previous research have reviewed and established the competence of satellite-based RS in estimating and mapping either seagrass-mangrove habitats biophysical parameters, mainly tree canopy/AGB (Adams et al., 2017; Gao et al., 2016; Karan et al., 2016). Therefore, this review article complimented other studies, through assessing the recent advancements on mapping, estimating and modeling seagrass-mangrove habitats using satellite-based approaches. The specific sets objectives addressed comprise: 1) to review relevant publications on acoustic RS techniques applicable for seagrass habitat, and 2) to review RS models employed for mapping and estimating biophysical components of seagrass-mangrove habitats.

2. SCOPE AND METHODOLOGY OF THE REVIEW 2.1.

Literature Search

The bibliographic search was carried out through the Scopus search engine, ScienceDirect, Web of Science, IEEE Xplore Digital Library, and Elsevier databases. Criteria used for eligibility was through incorporated only articles that were published from 2008 to 2018 (a decade) with the following terms either in their keywords abstract or title: (‘‘seagrass habitat’’ or ‘‘blue carbon biomass” or “mangrove” or “acoustic RS’’) and (‘‘bottom reflectance index (BRI)’’ or ‘‘depth-invariant index (DII)’’ or ‘‘water column correction’’ or “ Leaf Area Index (LAI)”). The total results were 84 publications and non-English articles were not included in the search. Thus, only publications that carried out satellite-based mapping, estimation and modelling on seagrass and mangrove biomass, and other biophysical parameters were reserved in the last segment of the evaluations. Fifty (50) articles on acoustic techniques and RS models used for mapping and estimating seagrass habitat were critically reviewed. Whereas the remaining 34 articles were utilized for citations in the residual segments of the review. 2.2.

Review Design

The review started through going over the acoustic RS techniques employed for seagrass habitat mapping, RS models (BRI, DII, LAI and cubical model) used for mapping and estimating seagrass biomass/carbon, an overview of acoustic techniques and RS modelling initiatives . Finally, conclusion and future directions were hilighted. Figure 1 presents the conceptual descriptions of the review article.

Figure 1. Conceptual diagram of the review article, which contained: a) acoustic RS techniques in seagrass habitat; b) RS models for detecting seagrass habitat (BRI, DII, LAI and cubical model). accuracy improvement for seagrass model and water column correction techniques; c) recent technology in for mangrove mapping; d) achievements and gaps in seagrass-mangrove habitats, e) an overview of the acoustic and RS modelling initiatives in seagrass habitat; and f) conclusion and future directions of the study. 3. APPLICATION OF ACOUSTIC REMOTE SENSING TECHNIQUES ON SEAGRASS HABITAT In active RS scope (for instance, acoustic), benthic habitat categories can be characterized through the investigation of backscatter data (McIntyre et al., 2018). The idea of acoustic scattering from seagrass is yet a less comprehended domain compared to sediment and rock (Bjørnø, 2017). The intensity of the backscatter in the seagrass presence is normally observed to be higher than muddy and sandy sea floor (Ierodiaconou et al., 2018). Therefore, acoustic response statistical analysis allow us to comprehend the fundamental substrate types via the mechanism of physical scattering (Barrell et al., 2015). The hypothesis and use of acoustic method for aquatic RS are very much outlined by (Jawak et al., 2015). Studies have demonstrated the likelihood of detecting seagrasses from differences of sound speed, based on laboratory-based experiments (Roelfsema et al., 2014). The advancement of acoustic RS approaches, which commenced around nineteenth century, has enhanced mapping competency where airplanes, satellite-based RS and in-situ sampling (for instance, grab, corer, photographs, video and trawls) were incapable to offer detailed information (Hossain et al., 2015a). Advancement of acoustic survey methods such as: 1) sonar side scan systems (example, star fish); 2) acoustic ground discrimination systems of single-beam; 3) echo-sounders multi-beam, and 4) one-dimensional (1D) single-beam equipment (echo-sounders and existing profiling sonars), are offering comprehensive information about seagrass blue carbon component.

These acoustic approaches have numerous strength and limitations on seagrass applications when compared with the optical RS techniques (Hossain, et al., 2015a). Active RS could be used in turbid and deep coastal waters to map seagrass (Poursanidis et al., 2018; Roelfsema, et al., 2014). Researchers have used acoustic devices to produce accurate seafloor maps, characterize the geological and biological seafloor characteristics, and develop algorithms to compute water depth, seagrass canopy height, and coverage at preferred depth ranges, mostly in single-species and occasionally in mixed-species seagrass meadows (Roelfsema et al., 2015b). Acoustic methods can produce satisfactory maps in a reasonable time when scanning in turbid waters and covering large seagrass meadows (Islam et al., 2017), although the relatively low canopy height makes seagrass species harder to distinguish, and it is difficult to map the distribution of seagrass meadows with low biomass (Klemas, 2013; Knudby and Nordlund, 2011). The comparative ability of mapping seafloor habitats using a variety of acoustic techniques and their strengths and limitations is published elsewhere (Brown et al., 2011; Lurton et al., 2015). Field-based observation and measurements are generally considered necessary for validating of the acoustic techniques results. For instance, efforts has been invested on employing satellite RS, using imagery for classification of bottom and acoustical survey techniques to validate methods (Lyons et al., 2015; Roelfsema et al., 2015a). However, a precise protocol has yet to be suggested by specialists. Therefore, determining the most effective acoustic approaches for seagrass meadows mapping is apprehensive as result of extensive field measurements due to the high cost of local and regional projects and time. Thus, there is scope for the integration of optical and acoustic techniques. Table 1 presents 18 publications (36% of the total reviewed articles) that applied acoustic RS approaches with their sensor applied, methods, strengths and limitations as well as the regions where the studies were conducted. This were detailed discussed in section 7 of the review article. 4. REMOTE SENSING MODELS USED FOR MAPPING SEAGRASS HABITAT Over the year, various RS models have been developed to map and estimate seagrass biomass and other biophysical components within the habitat (Hashim et al., 2014b; Hill et al., 2014; Hossain et al., 2015b; Sagawa et al., 2010), some of these models were applied in clear water (less than 5m) while some in less clear or turbid water (more than 5m) across the globe. Among the frequent utilized models for detecting seagrass meadow, include depth-invariant index (DII) to remove light scattering and absorption effects within both atmosphere and water body, and bottom reflectance index (BRI) was developed to improve seagrass mapping accuracy. 4.1

Accuracy Improvement for Seagrass Model

4.1.1 Techniques of Water Column Correction: It is essential to measure attenuation of the light and correct the water column effects on benthic reflection used for applications that involve the production of SAV as well as seagrass ecosystem maps (Klemas, 2013; Pu and Bell, 2013). The most generally utilized water column correction method is the Lyzenga’s (Lyzenga, 1981; Maritorena, 1996). Lyzenga (1978) expressed the relationship between radiance and bottom reflectance through the following equation: 𝐿𝑖 = 𝐿𝑑,𝑖 + 𝑎𝑖 𝑟𝑖 exp(−𝐾𝑖 gZ),

(1)

where 𝐿𝑖 stand for the radiance in bandi, 𝐿𝑑,𝑖 represent the average radiance that was recorded over deep-water in bandi (external reflection from the water surface and scattering in the atmosphere). 𝑎𝑖 is constant that comprise; solar irradiance, the transmittance of the atmosphere and the water surface, and the reduction of the radiance due to refraction at the water surface (mWcm–2 sr–1). 𝑟𝑖 is the bottom surface reflectance, 𝐾𝑖 is the effective attenuation coefficient of the water (m–1) of bandi, g is a geometric factor to account for the path length through the water, Z is the water depth (m), and exp stands for exponential. Lyzenga (1978), further suggested the calculation of a depth-invariant index (DII) to remove light scattering and absorption effects within both atmosphere and water body as follows: DII𝑖𝑗 =

𝐾𝑗 𝐼𝑛(𝐿𝑖 −𝐿𝑑,𝑖 )−𝐾𝑗 In(𝐿𝑖 −𝐿𝑑,𝑖 ) √𝐾𝑖 2 +𝐾𝑖 2

,

(2)

where the i and j subscripts correspond to two different bands of the satellite image and 𝐼𝑛 stands for natural logarithm. The DII is found to be effective for correcting clear water (types I and II) (Bukata et al., 2018) but inefficient when water clarity deceases from type II to type III (Sagawa, et al., 2010). To improve mapping accuracy Sagawa et al. (2010) proposed an alternative bottom reflectance index (BRI) expressed by the following equation:

BRI 

 Li  Lsi  exp   KigZ  

(3)

By replacing the numerator in Equation 3) by 𝑎𝑖 𝑟𝑖 exp KigZ of Equation 1), the BRI can be rearranged as: BRI𝑖𝑗 = 𝑎𝑖 𝑟𝑖 ,

(4)

where a and r represents same as in Equation 1). BRI can competently be utilized for II and III type of coastal waters, and allows the comparison of not only the distinction in reflectance proportions. linearizing the RS esteem concerning water profundity by subtracting an optically profound water an incentive from the entire waveband, trailed by change of the outcomes into regular logarithm lastly relapse examination for each band against relating profundity esteems between bands but as well distinction in absolute reflectance for individual band. Tassan (1996), modified Lyzenga’s technique and proposed water column correction used for greater turbid water in (1996) but its mathematically complex and required field validation. Other methods employs for water column corrections include: 1) linearizing the value of RS with water depth via subtracting an optically profound water value of the whole waveband, trailed by transformation of the outputs into natural logarithm as well as regression analysis for individually band against the corresponding depth values in the final stage (Abd Rahman et al., 2017); 2) By applying simple subtraction of a profound water RS reflectance from individually pixel (Rößler, 2014) and 3) employment of adjacent profound water optical properties (Yamashita et al., 2008). Wicaksono and Hafizt (2013), revealed the establishment of an optical model used for water column correction to map seagrass blue carbon component and developed strong correlations amongst the LAI and reflectance value. The inversion approaches/artificial neural network (ANN) inversions (Hedley et al., 2016; Hossain, et al., 2015a) match remotely estimated spectral values with query tables. Very few researchers have tested relative competence of water column correction for improving mapping accuracy. For instance, the RT model was used to produce a seagrass habitat map through comparing the Lyzenga’s optical model/Stumpf’s ratio model (Sagawa, et al., 2010). Thus, all these research examples pointed out that water column correction is crucial for creating quantitative empirical relationships amongst benthic features and spectral values, and for improving accuracy of the map. Table 2 presents 32 publications, which employed RS models to map and quantify seagrass biomass/carbon, these include cubical model 4%; depth-invariant index (DII) 41%; bottom reflectance index (BRI) 14%; and leaf area index (LAI) 41% publications respectively. 6. AN OVERVIEW OF ACOUSTIC TECHNIQUES AND REMOTE SENSING MODELLING INITIATIVES Acoustic techniques and RS models has become an essential tool for mapping and estimating blue carbon biomass in seagrass-mangrove habitats, from the perspectives of spatial and temporal. These techniques provide novel prospects for costal management from geographic information. This section analyzed some articles that applied acoustic techniques and RS modelling initiatives from 2008 to 2018, their limitations and strengths were reviewed as well as summarized in Table 1 and 2 of the article. The set of publications reviewed in this article exposed the significance of acoustic techniques and RS models in addressing issues of mapping and estimation accuracy in seagrass-mangrove habitats. Eighteen (36% the total reviewed articles) case studies that employs acoustic RS techniques were presented in Table 1. Moreover, thirty-three (64% the total reviewed articles) publication that applied RS models (DII; BRI; LAI and Cubical Model) used in mapping and estimating biomass in seagrass habitat were also presented in Table 2. Finally, the 51 articles reviewed were presented as a chart connoting percentages of publications based on seven global continents include: Africa has 2% of the acoustic RS publications while in the other publications it has 0%; Antarctica 0%; Asia 2%, 26%, 11%, 11% and 8% respectively; Australia 2% in LAI, 9% acoustic RS and the other publications 0%; Europe 6% for acoustic RS, whereas 0% for the remaining once; North America for BRI 2%, LAI 6% and acoustic RS 15% and others 0%, and South America has 0% as revealed by the review article. This review will be significant to stockholders such as blue carbon researchers, coastal management, industries and peoples who have interest in coastal resources to comprehend the dynamics of seagrass-mangrove habitats. Most importantly, it will support the realization of United Nation’s (UN) Sustainable Development Goals (SDG 14)

Table 1. Some Publications That Employed Acoustic Remote Sensing Techniques in Seagrass Habitat

Citation Barrell, et al. (2015); Hamana and Komatsu (2016); Kovacs et al. (2018); Saunders et al. (2015); Rahnemoonfar et al. (2017); Duffy et al. (2018); Poursanidis, et al. (2018); Rahnemoonfar et al. (2018); Sagawa et al. (2008); Greene et al. (2018); Ballard et al. (2018); Roelfsema et al. (2009); Komatsu et al. (2002); Ierodiaconou et al. (2007); Komatsu et al. (2012); Sánchez-Carnero et al. (2012); Barrell and Grant (2013); Stevens et al. (2008)

Model Type Cubical Model For Seagrass Biomass Depth-Invariant Index (DII)

Bottom Reflectance (BRI) Leaf Area Index (LAI)

Index

Sensor Side scan sonar

Method  Using of acoustic signals to estimate carbon stores in seagrass habitat

Strength  Comparatively high accuracy when compared with independent groundbased reference data,  Precise sensors for mapping the seagrass disturbance. Mapping techniques that yield high definition, twodimensional (2D) sonar imagery of seagrass habitats.

Limitation  Agreement between acoustic and satellite data sets is limited due environmental impact  The work lacks real-time seagrass identification patterns of disturbance.

Table 2. Remote Sensing Models Used for Mapping and Estimation of Seagrass Biomass Citation Strength Hashim et al. (2014a)  Ability to map seagrass distribution and standing crop to a depth of about 10 m.  Efficient for clear waters Manuputty et al. (2017); Thalib et al. (2018); Wahab et al. (2017); Knudby and Nordlund (2011); Manuputty et al. (2016); Thalib, et al. (2018); Geevarghese et al. (2018); Chen et al. (2016); Hafizt et al. (2017); Siregar et al. (2018); Fauzan et al. (2017); Noiraksar et al. (2014); Hanjaniputri (2017) Misbari and Hashim (2016a); Sagawa, et al. (2010); Hashim, et al.  Combination of bathymetry data with attenuation coefficients. (2014a); Misbari and Hashim (2014); Sagawa et al. (2012) Samper-Villarreal et al. (2018); Hedley, et al. (2016); Hedley et al. (2017); (Misbari and Hashim (2016b); Pu and Bell (2013)); Wicaksono and Hafizt (2013); Yang et al. (2010); Borfecchia et al. (2013); JeanBaptiste and Jensen (2006); Mabrouk et al. (2012); Santana and Encinas (2011); Adi et al. (2013);(Yang and Yang (2009));



Seagrass LAI represents the abundance of seagrass properties in the area

Region Canada; Japan; Australia; Australia; Texas; Wales UK; Greece; Florida; Japan; Texas; Laguna Madre ;Australia; Japan; Australia; Japan; Spain; Canada; Washington

Limitation  Not widely used and tested in another part of tropical coastal region.  Largely unsuitable when transparency decreases



Can be less effective with increase in depth



Sensor, method and environmental limitation contribute to the low accuracy of seagrass LAI mapping



Figure 2: present percentage of the reviewed articles based on global continents. 7. CONCLUSION AND FUTURE DIRECTION In summary, RS and modelling techniques has yielded in a tremendously powerful tool for investigating substantial seagrass-mangrove biophysical components in the regional and global scale. It provides a novel view of the seagrassmangrove biomass mapping and estimation, which could be difficult to achieve through sparse in-situ observations. The utilization of acoustic-based RS techniques for mapping and estimating seagrass blue carbon habitat were documented, as significant advances have been attained recently via using RS techniques as revealed by the reviewed article. RS models (DII; BRI; LAI and Cubical Model) used in detecting seagrass biomass and other biophysical components was reviewed. Similarly, recent technology employs for mangrove biomass/carbon mapping and estimation were also highlighted, and finally an overview of acoustic RS techniques and modelling initiatives were likewise presented in this reviewed article. Future research should focus on: 1) mapping and estimation of the 3 variables (AGC, BGC, and sediment) in order to come up with the total estimation of carbon pools in seagrassmangrove habitats, for fulfilling the UNFCCC convention; 2) development of advanced tools for Landsat time-series images processing to achieve better results for observing changes in smaller species like in seagrass cover for purposive biomass monitoring as well as other biophysical characteristics; 3) LiDAR and Landsat can be utilized together for estimation AGB and changes within mangroves locations, similarly to terrestrial forested areas; 5) embracing diverse mapping methods can be appropriate for different applications, and 6) incorporating financial values in biomass and organic carbon mapping, estimation and modelling with related blue carbon ecosystem services will encourage restoration and conservation of important blue carbon components. This is essential, as unravelling seagrass-mangrove habitats complexities will offer better mapping and estimation of biomass with acceptable accuracy for coastal ecosystems management. Furthermore, the important components of seagrass-mangrove carbon stock’s assessment will be actualized if scientists can develop new methods to offer fresh insights on the issue of acoustic-based-RS and modelling dynamics to support the realization of the SDG 14. Acknowledgments

The authors desire to acknowledge the federal government of Nigerian for providing monetary intervention to the first author through Tertiary Education Trust Fund (TETFUND). Facilities of research utilized in Univerisiti Teknologi Malaysia (UTM) are also acknowledged. References Abd Rahman, M. Z., Abu Bakar, M. A., Razak, K. A., Rasib, A. W., Kanniah, K. D., Wan Kadir, W. H., Omar, H., Faidi, A., Kassim, A. R. and Abd Latif, Z. (2017). Non-destructive, laser-based individual tree aboveground biomass estimation in a tropical rainforest. Forests. 8(3), 86. Adams, M. P., Collier, C. J., Uthicke, S., Ow, Y. X., Langlois, L. and O’Brien, K. R. (2017). Model fit versus biological relevance: Evaluating photosynthesis-temperature models for three tropical seagrass species. Scientific reports. 7.

Adi, N. S., Phinn, S., Roelfsema, C. and Samper-Villarreal, J. (2013). Integrating field and remote sensing approach for mapping seagrass leaf area index. Proceedings of the 2013 ACRS 2013: 34th Asian Conference on Remote Sensing: Indonesian Remote Sensing Society (ISRS/MAPIN), 3266-3273. Ballard, M. S., Lee, K. M., Sagers, J. D., Venegas, G. R., McNeese, A. R., Rahman, A. F., Dubin, J. T. and Wilson, P. S. (2018). Measurements and modeling of acoustic propagation in a seagrass meadow. Proceedings of the 2018 Proceedings of Meetings on Acoustics 175ASA: ASA, 005002. Barrell, J. and Grant, J. (2013). Detecting hot and cold spots in a seagrass landscape using local indicators of spatial association. Landscape ecology. 28(10), 2005-2018. Barrell, J., Grant, J., Hanson, A. and Mahoney, M. (2015). Evaluating the complementarity of acoustic and satellite remote sensing for seagrass landscape mapping. International Journal of Remote Sensing. 36(16), 40694094. Benson, L., Glass, L., Jones, T. G., Ravaoarinorotsihoarana, L. and Rakotomahazo, C. (2017). Mangrove Carbon Stocks and Ecosystem Cover Dynamics in Southwest Madagascar and the Implications for Local Management. Forests. 8(6), 190. Bjørnø, L. (2017). Underwater Acoustic Measurements and Their Applications Applied Underwater Acoustics (pp. 889-947)Elsevier. Borfecchia, F., De Cecco, L., Martini, S., Ceriola, G., Bollanos, S., Vlachopoulos, G., Valiante, L. M., Belmonte, A. and Micheli, C. (2013). Posidonia oceanica genetic and biometry mapping through high-resolution satellite spectral vegetation indices and sea-truth calibration. International journal of remote sensing. 34(13), 46804701. Brown, C. J., Smith, S. J., Lawton, P. and Anderson, J. T. (2011). Benthic habitat mapping: A review of progress towards improved understanding of the spatial ecology of the seafloor using acoustic techniques. Estuarine, Coastal and Shelf Science. 92(3), 502-520. Bukata, R. P., Jerome, J. H., Kondratyev, A. S. and Pozdnyakov, D. V. (2018). Optical properties and remote sensing of inland and coastal waters. CRC press. Chen, C.-F., Lau, V.-K., Chang, N.-B., Son, N.-T. and Chiang, S.-H. (2016). Multi-temporal change detection of seagrass beds using integrated Landsat TM/ETM+/OLI imageries in Cam Ranh Bay, Vietnam. Ecological informatics. 35, 43-54. Congdon, V. M., Wilson, S. S. and Dunton, K. H. (2017). Evaluation of Relationships Between Cover Estimates and Biomass in Subtropical Seagrass Meadows and Application to Landscape Estimates of Carbon Storage. Southeastern Geographer. 57(3), 231-245. Duffy, J. P., Pratt, L., Anderson, K., Land, P. E. and Shutler, J. D. (2018). Spatial assessment of intertidal seagrass meadows using optical imaging systems and a lightweight drone. Estuarine, Coastal and Shelf Science. 200, 169-180. Duveiller, G., Hooker, J. and Cescatti, A. (2018). The mark of vegetation change on Earth’s surface energy balance. Nature communications. 9(1), 679. Fauzan, M. A., Kumara, I. S., Yogyantoro, R., Suwardana, S., Fadhilah, N., Nurmalasari, I., Apriyani, S. and Wicaksono, P. (2017). Assessing the Capability of Sentinel-2A Data for Mapping Seagrass Percent Cover in Jerowaru, East Lombok. The Indonesian Journal of Geography. 49(2), 195-203. Gao, Y., Yu, G., Yang, T., Jia, Y., He, N. and Zhuang, J. (2016). New insight into global blue carbon estimation under human activity in land-sea interaction area: a case study of China. Earth-Science Reviews. 159, 36-46. Geevarghese, G. A., Akhil, B., Magesh, G., Krishnan, P., Purvaja, R. and Ramesh, R. (2018). A comprehensive geospatial assessment of seagrass distribution in India. Ocean & Coastal Management. 159, 16-25. Greene, A., Rahman, A. F., Kline, R. and Rahman, M. S. (2018). Side scan sonar: A cost-efficient alternative method for measuring seagrass cover in shallow environments. Estuarine, Coastal and Shelf Science. 207, 250-258. Hafizt, M., Manessa, M., Adi, N. and Prayudha, B. (2017). Benthic Habitat Mapping by Combining Lyzenga’s Optical Model and Relative Water Depth Model in Lintea Island, Southeast Sulawesi. Proceedings of the 2017 IOP Conference Series: Earth and Environmental Science: IOP Publishing, 012037. Hamana, M. and Komatsu, T. (2016). Real-Time Classification of Seagrass Meadows on Flat Bottom with Bathymetric Data Measured by a Narrow Multibeam Sonar System. Remote Sensing. 8(2), 96. Hanjaniputri, B. U. (2017). Distribution of Blue Carbon Ecosystems In Small Islands Using Landsat-8 (Case Study In Pramuka Island). Jurnal Geografi Gea. 17(1), 82-91. Hashim, M., Misbari, S., Yahya, N. N., Ahmad, S., Reba, M. N. and Komatsu, T. (2014a). An approach for quantification of submerged seagrass biomass in shallow turbid coastal waters. Proceedings of the 2014a Geoscience and Remote Sensing Symposium (IGARSS), 2014 IEEE International: IEEE, 4439-4442.

Hashim, M., Yahya, N. N., Ahmad, S., Komatsu, T., Misbari, S. and Reba, M. (2014b). Determination of seagrass biomass at Merambong Shoal in Straits of Johor using satellite remote sensing technique. Malay. Nat. J. 66, 20-37. Hedley, J., Russell, B., Randolph, K. and Dierssen, H. (2016). A physics-based method for the remote sensing of seagrasses. Remote sensing of environment. 174, 134-147. Hedley, J. D., Russell, B. J., Randolph, K., Pérez-Castro, M. Á., Vásquez-Elizondo, R. M., Enríquez, S. and Dierssen, H. M. (2017). Remote Sensing of Seagrass Leaf Area Index and Species: The Capability of a Model Inversion Method Assessed by Sensitivity Analysis and Hyperspectral Data of Florida Bay. Frontiers in Marine Science. 4, 362. Hill, V. J., Zimmerman, R. C., Bissett, W. P., Dierssen, H. and Kohler, D. D. (2014). Evaluating light availability, seagrass biomass, and productivity using hyperspectral airborne remote sensing in Saint Joseph’s Bay, Florida. Estuaries and coasts. 37(6), 1467-1489. Hossain, M., Bujang, J., Zakaria, M. and Hashim, M. (2015a). The application of remote sensing to seagrass ecosystems: an overview and future research prospects. International Journal of Remote Sensing. 36(1), 61114. Hossain, M. S., Bujang, J. S., Zakaria, M. H. and Hashim, M. (2015b). Assessment of the impact of Landsat 7 Scan Line Corrector data gaps on Sungai Pulai Estuary seagrass mapping. Applied Geomatics. 7(3), 189-202. Ierodiaconou, D., Laurenson, L., Burq, S. and Reston, M. (2007). Marine benthic habitat mapping using Multibeam data, georeferencedvideo and image classification techniques in Victoria, Australia. Journal of Spatial Science. 52(1), 93-104. Ierodiaconou, D., Schimel, A. C., Kennedy, D., Monk, J., Gaylard, G., Young, M., Diesing, M. and Rattray, A. (2018). Combining pixel and object based image analysis of ultra-high resolution multibeam bathymetry and backscatter for habitat mapping in shallow marine waters. Marine Geophysical Research. 39(1-2), 271-288. Islam, S. M., Raza, S. K., Moniruzzamn, M., Janjua, N., Lavery, P. and Al-Jumaily, A. (2017). Automatic seagrass detection: A survey. Proceedings of the 2017 Electrical and Computing Technologies and Applications (ICECTA), 2017 International Conference on: IEEE, 1-5. Jawak, S. D., Vadlamani, S. S. and Luis, A. J. (2015). A synoptic review on deriving bathymetry information using remote sensing technologies: models, methods and comparisons. Advances in Remote Sensing. 4(02), 147. Jean-Baptiste, N. and Jensen, J. R. (2006). Measurement of mangrove biophysical characteristics in the Bocozelle ecosystem in Haiti using ASTER multispectral data. Geocarto International. 21(4), 3-8. Karan, S. K., Samadder, S. R. and Maiti, S. K. (2016). Assessment of the capability of remote sensing and GIS techniques for monitoring reclamation success in coal mine degraded lands. Journal of environmental management. 182, 272-283. Klemas, V. (2013). Remote sensing of emergent and submerged wetlands: an overview. International journal of remote sensing. 34(18), 6286-6320. Knudby, A. and Nordlund, L. (2011). Remote sensing of seagrasses in a patchy multi-species environment. International Journal of Remote Sensing. 32(8), 2227-2244. Komatsu, T., Igararashi, C., Tatsukawa, K.-i., Nakaoka, M., Hiraishi, T. and Taira, A. (2002). Mapping of seagrass and seaweed beds using hydro-acoustic methods. Fisheries science. 68(sup1), 580-583. Komatsu, T., Sagawa, T., Sawayama, S., Tanoue, H., Mohri, A. and Sakanishi, Y. (2012). Mapping is a key for sustainable development of coastal waters: examples of seagrass beds and aquaculture facilities in Japan with use of ALOS images Sustainable Development-Education, Business and Management-Architecture and Building Construction-Agriculture and Food Security InTech. Kovacs, E., Roelfsema, C., Lyons, M., Zhao, S. and Phinn, S. (2018). Seagrass habitat mapping: how do Landsat 8 OLI, Sentinel-2, ZY-3A, and Worldview-3 perform? Remote Sensing Letters. 9(7), 686-695. Lurton, X., Lamarche, G., Brown, C., Lucieer, V., RIce, G., Schimel, A. and Weber, T. (2015). Backscatter measurements by seafloor‐mapping sonars: guidelines and recommendations. A collective report by members of the GeoHab Backscatter Working Group. (May), 1-200. Lyons, M., Roelfsema, C. M. and Phinn, S. R. (2015). Seagrass time-series analysis products, in Moreton Bay, Australia, with links to GeoTIFFs. Lyzenga, D. R. (1978). Passive remote sensing techniques for mapping water depth and bottom features. Applied optics. 17(3), 379-383. Lyzenga, D. R. (1981). Remote sensing of bottom reflectance and water attenuation parameters in shallow water using aircraft and Landsat data. International journal of remote sensing. 2(1), 71-82.

Mabrouk, L., Hamza, A., Mahfoudi, M. and Bradai, M. N. (2012). Spatial and temporal variations of epiphytic Ostreopsis siamensis on Posidonia oceanica (L.) Delile leaves in Mahdia (Tunisia). CBM-Cahiers de Biologie Marine. 53(4), 419. Macreadie, P. I., Serrano, O., Maher, D. T., Duarte, C. M. and Beardall, J. (2017). Addressing calcium carbonate cycling in blue carbon accounting. Limnology and Oceanography Letters. 2(6), 195-201. Manuputty, A., Gaol, J. L., Agus, S. B. and Nurjaya, I. W. (2017). The utilization of Depth Invariant Index and Principle Component Analysis for mapping seagrass ecosystem of Kotok Island and Karang Bongkok, Indonesia. Proceedings of the 2017 IOP Conference Series: Earth and Environmental Science: IOP Publishing, 012083. Manuputty, A., Lumban-Gaol, J. and Agus, S. B. (2016). Seagrass Mapping Based on Satellite Image Worldview-2 by Using Depth Invariant Index Method. Indonesian Journal of Marine Sciences/Ilmu Kelautan. 21(1). Maritorena, S. (1996). Remote sensing of the water attenuation in coral reefs: a case study in French Polynesia. International Journal of Remote Sensing. 17(1), 155-166. McIntyre, K., McLaren, K. and Prospere, K. (2018). Mapping shallow nearshore benthic features in a Caribbean marine-protected area: assessing the efficacy of using different data types (hydroacoustic versus satellite images) and classification techniques. International Journal of Remote Sensing. 39(4), 1117-1150. Misbari, S. and Hashim, M. (2014). Evaluation of median filtering impact on satellite-based submerged seagrass mapping accuracy in tropical coastal water. Proceedings of the 2014 Proceedings of 35th Asian Conference on Remote Sensing, Myanmar, Misbari, S. and Hashim, M. (2016a). Change detection of submerged seagrass biomass in shallow coastal water. Remote Sensing. 8(3), 200. Misbari, S. and Hashim, M. (2016b). Light Penetration Ability Assessment of Satellite Band for Seagrass Detection Using Landsat 8 OLI Satellite Data. Proceedings of the 2016b International Conference on Computational Science and Its Applications: Springer, 265-274. Mishra, U., Jastrow, J., Matamala, R. and Fan, Z. (2017). Observational needs for estimating Alaskan soil carbon stocks under current and future climate. Journal of Geophysical Research (Biogeosciences). 122, 415-429. Noiraksar, T., Sawayama, S., Phauk, S. and Komatsu, T. (2014). Mapping Sargassum beds off the coast of Chon Buri Province, Thailand, using ALOS AVNIR-2 satellite imagery. Botanica Marina. 57(5), 367-377. Pendleton, L., Donato, D. C., Murray, B. C., Crooks, S., Jenkins, W. A., Sifleet, S., Craft, C., Fourqurean, J. W., Kauffman, J. B. and Marbà, N. (2012). Estimating global “blue carbon” emissions from conversion and degradation of vegetated coastal ecosystems. PloS one. 7(9), e43542. Poursanidis, D., Topouzelis, K. and Chrysoulakis, N. (2018). Mapping coastal marine habitats and delineating the deep limits of the Neptune’s seagrass meadows using very high resolution Earth observation data. International Journal of Remote Sensing. 1-18. Pu, R. and Bell, S. (2013). A protocol for improving mapping and assessing of seagrass abundance along the West Central Coast of Florida using Landsat TM and EO-1 ALI/Hyperion images. ISPRS journal of photogrammetry and remote sensing. 83, 116-129. Rahnemoonfar, M., Rahman, A. F., Kline, R. J. and Greene, A. (2018). Automatic Seagrass Disturbance Pattern Identification on Sonar Images. IEEE Journal of Oceanic Engineering. Rahnemoonfar, M., Yari, M., Rahman, A. and Kline, R. (2017). The First Automatic Method for Mapping the Pothole in Seagrass. Proceedings of the 2017 Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, 73-80. Roelfsema, C., Lyons, M., Dunbabin, M., Kovacs, E. M. and Phinn, S. (2015a). Integrating field survey data with satellite image data to improve shallow water seagrass maps: the role of AUV and snorkeller surveys? Remote Sensing Letters. 6(2), 135-144. Roelfsema, C., Phinn, S., Udy, N. and Maxwell, P. (2009). An integrated field and remote sensing approach for mapping seagrass cover, Moreton Bay, Australia. Journal of Spatial Science. 54(1), 45-62. Roelfsema, C. M., Lyon, M., Kovacs, E., Maxwell, P., Saunders, M., Samper-Villarreal, J. and Phinn, S. R. (2015b). Multi-temporal mapping of seagrass cover, species and biomass of the Eastern Banks, Moreton Bay, Australia, with links to shapefiles. Roelfsema, C. M., Lyons, M., Kovacs, E. M., Maxwell, P., Saunders, M. I., Samper-Villarreal, J. and Phinn, S. R. (2014). Multi-temporal mapping of seagrass cover, species and biomass: A semi-automated object based image analysis approach. Remote Sensing of Environment. 150, 172-187. Rößler, S. (2014). Methods for multitemporal mapping of submerged aquatic macrophytes using multi-and hperspectral remote sensing, Technische Universität München.

Sagawa, T., Boisnier, E., Komatsu, T., Mustapha, K. B., Hattour, A., Kosaka, N. and Miyazaki, S. (2010). Using bottom surface reflectance to map coastal marine areas: a new application method for Lyzenga's model. International Journal of Remote Sensing. 31(12), 3051-3064. Sagawa, T., Mikami, A., Komatsu, T., Kosaka, N., Kosako, A., Miyazaki, S. and Takahashi, M. (2008). Mapping seagrass beds using IKONOS satellite image and side scan sonar measurements: a Japanese case study. International Journal of Remote Sensing. 29(1), 281-291. Sagawa, T., Watanabe, T., Watanuki, A., Koike, T., Kamimura, H. and Komatsu, T. (2012). Can ALOS-3/HISUI detect seaweed beds more precisely than ALOS/AVNIR-2? Proceedings of the 2012 Remote Sensing of the Marine Environment II: International Society for Optics and Photonics, 85250Y. Samper-Villarreal, J., Mumby, P. J., Saunders, M. I., Roelfsema, C. and Lovelock, C. E. (2018). Seagrass Organic Carbon Stocks Show Minimal Variation Over Short Time Scales in a Heterogeneous Subtropical Seascape. Estuaries and Coasts. 1-12. Sánchez-Carnero, N., Rodríguez-Pérez, D., Couñago, E., Aceña, S. and Freire, J. (2012). Using vertical Sidescan Sonar as a tool for seagrass cartography. Estuarine, Coastal and Shelf Science. 115, 334-344. Santana, O. A. and Encinas, J. I. (2011). Leaf Area Index and Canopy Openness estimation using high spatial resolution image QuickBird. Revista Caatinga. 24(2), 59-66. Saunders, M. I., Bayraktarov, E., Roelfsema, C. M., Leon, J. X., Samper-Villarreal, J., Phinn, S. R., Lovelock, C. E. and Mumby, P. J. (2015). Spatial and temporal variability of seagrass at Lizard Island, Great Barrier Reef. Botanica Marina. 58(1), 35-49. Siregar, V., Agus, S., Subarno, T. and Prabowo, N. (2018). Mapping shallow waters habitats using OBIA by applying several approaches of depth invariant index in North Kepulauan Seribu. Proceedings of the 2018 IOP Conference Series: Earth and Environmental Science: IOP Publishing, 012052. Sousa, A. I., Santos, D. B., Da Silva, E. F., Sousa, L. P., Cleary, D. F., Soares, A. M. and Lillebø, A. I. (2017). ‘Blue Carbon’and Nutrient Stocks of Salt Marshes at a Temperate Coastal Lagoon (Ria de Aveiro, Portugal). Scientific reports. 7. Stevens, A. W., Lacy, J. R., Finlayson, D. P. and Gelfenbaum, G. (2008). Evaluation of a single-beam sonar system to map seagrass at two sites in northern Puget Sound, Washington. Geological Survey (US). Tassan, S. (1996). Modified Lyzenga's method for macroalgae detection in water with non-uniform composition. International Journal of Remote Sensing. 17(8), 1601-1607. Thalib, M., Nurdin, N. and Aris, A. (2018). The Ability of Lyzenga’s Algorithm for Seagrass Mapping using Sentinel2A Imagery on Small Island, Spermonde Archipelago, Indonesia. Proceedings of the 2018 IOP Conference Series: Earth and Environmental Science: IOP Publishing, 012028. Thompson, D. R., Hochberg, E. J., Asner, G. P., Green, R. O., Knapp, D. E., Gao, B.-C., Garcia, R., Gierach, M., Lee, Z. and Maritorena, S. (2017). Airborne mapping of benthic reflectance spectra with Bayesian linear mixtures. Remote Sensing of Environment. 200, 18-30. Wahab, I., Madduppa, H. and Kawaroe, M. (2017). Seagrass species distribution, density and coverage at Panggang Island, Jakarta. Proceedings of the 2017 IOP Conference Series: Earth and Environmental Science: IOP Publishing, 012084. Whitfield, A. K. (2017). The role of seagrass meadows, mangrove forests, salt marshes and reed beds as nursery areas and food sources for fishes in estuaries. Reviews in Fish Biology and Fisheries. 27(1), 75-110. Wicaksono, P. and Hafizt, M. (2013). Mapping seagrass from space: Addressing the complexity of seagrass LAI mapping. European Journal of Remote Sensing. 46(1), 18-39. Yamashita, Y., Jaffé, R., Maie, N. and Tanoue, E. (2008). Assessing the dynamics of dissolved organic matter (DOM) in coastal environments by excitation emission matrix fluorescence and parallel factor analysis (EEM‐ PARAFAC). Limnology and Oceanography. 53(5), 1900-1908. Yang, C., Yang, D., Cao, W., Zhao, J., Wang, G., Sun, Z., Xu, Z. and Ravi Kumar, M. (2010). Analysis of seagrass reflectivity by using a water column correction algorithm. International Journal of Remote Sensing. 31(1718), 4595-4608. Yang, D. and Yang, C. (2009). Detection of seagrass distribution changes from 1991 to 2006 in Xincun Bay, Hainan, with satellite remote sensing. Sensors. 9(2), 830-844.

PARETO OPTIMAL SOLUTION FOR DETECTION OF MH370 DEBRIS USING REMOTE SENSING SATELLITE DATA Maged Marghany School of Humanities, Geography Section, Universiti Sains Malaysia, 11800 USM Penang, Malaysia 2 Faculty Geospatial and Real Estate, Geomatika University College, Kuala Lumpur Email :[email protected]

1

KEYWORDS : Multi-objective algorithm. Pareto optimization, Indian Ocean circulation, MH370 flight, debris. ABSTRACT: Regardless of the superior area, marine, and communication technologies, the mystery of the Malaysia Airline flight MH370 cannot be explicated. Excluding twelve countries that allied for the search and rescue efforts of missing the flight MH370 on March 8th, 2014, it is very sophisticated to analyze the dramatic situation of the flight MH370 that non-existent from secondary microwave radar. The core objective is to develop a multi-objective optimisation via Pareto dominance to scale back the uncertainties for the debris automatic detection in satellite information like China satellite. Additionally, multi-objective optimisation, supported the genetic algorithmic rule is developed to forecast the debris flight movements from Perth, west of Australia i.e. the crashed claimed space. The Pareto optimization proved that within a water depth of 3000 m the remaining debris of 60% of total debris would sink down with highest cumulative percentage of 95%. As the debris would undergo the impacts of turbulent across the Southern Indian Ocean. Moreover, The detritus has been found in Réunion Island do not seem to belong to MH370. In fact, the detritus would sink below the ocean surface of 3000 water depths at intervals less than a few months as explained above. It can be said that the flight MH370 detritus can doubtless travel up to 50 km/day with massive eddies of a dimension of 100 km wide. 1. INTRODUCTION Notwithstanding that advanced technologies are not able to answer the critical question of where is the Flight MH370. Excluding twelve countries that allied for the search and rescue efforts of missing the flight MH370 on March 8th, 2014, it is very sophisticated to analyze the dramatic situation of the flight MH370 that non-existent from secondary microwave radar. MH370 routes of 5 nmi / 8–10 km wide are delineated conversely differed in breadth as 20 nmi / 35–40 km (Asia News 2014 and Excell , 2014; Zweck, 2014a; staff writer 2014). There have been many optical satellite sensor data that have been claimed to be objects of happiness to the flight MH370. One amongst these satellite information may additionally be a Thai satellite data that have detected 300 floating objects on the surface of the Indian ocean, concerning 200 kilometres from the international search space for the missing Malaysia Airlines MH 370 at a 10am Perth time on the 24th of March. During this context, the THEOS satellite optical sensors payload alternatives every high resolution in panchromatic mode and the wide subject of analyzing in multispectral bands and has been tailor-made to Thailand's particular needs with a global imaging capability. Additionally, it includes a 2-m resolution for black and white images and 15m resolution of the panchromatic image. Nevertheless, the images claimed to belong to flight MH370 are dominated by means of cloud covers (Marghany, 2015). Subsequently, the Malaysian navy microwave radar suggested that MH370 cosmopolitan over the Straits of Malacca whereas the white circle suggests disappearing of flight from the radar screen. China, as a result, properly deployed 10 high-resolution satellites to scurry the South China Sea, digital globe untied its crowdsourcing platform Tomnod and airliner defence and vicinity mobilized its 5 satellites to search out some leads (Marghany 2014; Grady 2014; Linlin, 2014; Zweck, 2014b). Under this fact, physical oceanography theories and models must impartially be instigated to analyse the mystery of flight MH370. The observation methods that tutor for the undergraduates of physical oceanography students do not add to this case. Indeed, common and modified models are required to confirm the understanding of Inmarsat satellite. In fact, there are a number of researchers who actually used the physical oceanography models and do not very discover on the

other hand the models are operated. The scientists enforced the drifter models to track the trajectory model of MH370 crashing. In this view, Martini (2015) commented that a prediction model and consequently the temporal order ought to be a very little off when you consider that it is entirely been 18 months considering that the crash. Martini (2015) similarly this pair possibly raised due to the fact that the model was potential loped with the historical surface current data and false numerical rubble. Martini (2015), withal, raised up the subsequent question why has not any MH370 debris been discovered in Australian and Tasmanian beaches as foreseen through the numerical particles mannequin? So, that model is no longer accurate to track the rubble of the crashed flight MH370 that scattered within the search space. Additionally, there are several choice dynamic ocean parameters extraordinarily struggling from the debris moves on the surface and through the water column (Marghany 2017). The main enquiry would be raised up what applicable sensors are frequently accustomed monitor and discover flight MH370 debris? The high-resolution sensors both on board of satellite or airborne can find out and determine the flight MH370 debris. Even HF ground can realize any MH370 objects occupancy the coastal zone. This is additionally required the really worth processes of object automatic detection through exploitation highresolution of the microwave, satellite data with 1 m high-resolution sensor of the spot mode of both RADARSAT-2 SAR, TerraSAR-X satellite information. The RADARSAT-2 SAR satellite contains an artificial aperture radar (SAR) with a couple of polarization modes, as properly as a definitely the polarimetric mode data are not inheritable. Its absolute best decision is 1 m within Spotlight mode (3 m within Ultra-Fine mode) with 100 m positional accuracy demand. Additionally, RADARDSAT-2 SAR Scan slim SCNB beam is its and an excessive return period of 7 days. Further, has nominal close to and great deal resolutions of 7 m. If the dimension of the flight is 24 m, potential that it can also definitely be detected in RADARDSAT-2 SAR Scan narrow. This implies that, as high cloud covers are dominated within the southern ocean, it is entreated to use airborne SAR sensors like unpopulated aerial vehicle synthetic aperture radar (UAVSAR, by using JPL, L-band) with a 22-km-wide ground swath at 22° to 65° (Marghany 2014 and Marghany et al., 2016). This investigation hypothesizes that the optimal solution of the Pareto algorithm can deliver accurate answer about the MH370 debris in the South Indian Ocean. The main objective of this work is to determine uncertainties associated with tracking MH370 debris. To this end, several satellite images are used to identify the physical characteristics of MH370 debris and its trajectory movements across the Indian Ocean. 2. SEARCH AREA The bathymetry of the suspected zone of crashing of NH370 is dominated by complicated underwater topography. The bathymetry of the search area is simulated from a survey which was once performed from May to December 2014, gathering data over 200,000 rectangular kilometres through the Joint Agency Coordination Centre (JACC) of Geoscience Australia. The seabed is dominated by two round Broken Ridges, a vast linear, mountainous sea ground structure that once fashioned the margin between two geological plates. These plates evolved and unfold apart between 20 and one hundred million years ago, under comparable methods found today at spreading plate margins (such as the Mid-Atlantic Ridge) (Geoscience Australia, 2015). Figure 1 suggests a located new seabed features which are: (i) seamounts (remnant submarine volcanoes), up to 1400 metres excessive and frequently forming a semi-linear chain; (ii) ridges (semi-parallel) up to 300 meters high, and (iii) depressions up to 1400 metres deep (compared to the surrounding seafloor depths) and regularly perpendicular to the smaller semi-parallel ridges (Smith and Marks 2014).

Figure 1. Bathymetry of the MH370 search area. The essential query is how the looking out operation failed to observe any wreckage with the topography under water? Side scan sonar provides a two-dimensional map of a region on either aspect of the sonar which could not discovers ways ample due to the complex water topography with outcrops, seamounts and quite a number other changes in remedy in many locations at some stage in the deep Indian Ocean. This is concluded that it is very difficult to become aware of wreckage with the problematic topography of the Indian Ocean. 3. PARETO OPTIMAL BASED ON GENETIC ALGORITHM In the genome, for each member of the population, the population is initialized by random assignment of a 0 or 1 to each of the 32 bits. Subsequently, the first 20 and 12 bits are transcribed into an integer representing the i,j coordinates, respectively to evaluate the fitness. The locations of trajectory movement of debris thenceforth are simulated (Anderson et al., 2013; 2013Serafino,2015; Marghany et al., 2016). Let X be a compact set of n feasible decisions in the Euclidean space with closed unit interval [0, 1], and Y is the feasible set of m criterion vectors in . Then Pareto front can be expressed as

P(Y )  { y1  Y : { y2 Y : y2

y1 , y2  y1 }  0 }

(1)

Let assume that large parameter space could be searched by the genetic algorithm (GA) to determine effective solutions. With regard to this, the predictive algorithm involves the nonlinear approximation function which is based on historical time series information on sea surface current, sea level variation, wave height variation and the Indian Ocean floor features to forecast the current location of MH370 debris to any feature state (Anderson et al., 2003; Anderson generic function 



m 1n 1

2013; Anderson

 which can state as follows:

2014). Let

 xi  be the observation made with a

(2)

 1 1   xn  n  m , xn  n  m   2 2 

The sequence observations that itemize the rational numbers are represented by generic function





 xnn1 .This

means that

is satisfying

    xi   xn  n 1

(3)

n 1

Let a hydrodynamic system of the southern Indian Ocean with m hydrodynamic parameters and n flight MH370 debris, and a utility function of each hydrodynamic parameters as

  f ( vi )

(4)

where v i is a vector of the flight MH370 debris and vi  ( v1 ,v2 , ......,vn ) . Then the feasibility constraint m equals

 v j  b j j=(1,2,3,…….,n). Finally, the Euler–Lagrange equations are maximized to find the Pareto

j 1 optimal allocation for the flight MH370 debris trajectory movements across the southern Indian Ocean.

m

Li (( x kj )k . j ,( k )k ,

  j  j )  f ( v )   k ( k  f k 2

k

k

( v )) 

n

m

j 1

k 1

  j ( b j   v j ) (5)

k here, L is Lagrangian with respect to each debris v for k=1,….,m and the vectors of multipliers are

  j  j , respectively and

k

and

k  j . The time series of archived data of significant wave heights, sea surface

current, sea level spatial variations and wind velocity March 2014 to March 2016 are collected from the Jason2/Ocean Surface Topography Mission (OSTM), and QuikSCAT respectively to simulate the contemporary and feasible debris trajectory movements throughout the Southern Indian Ocean. This information perhaps assists in finding precise information regarding the impacts of Southern Indian Circulation on the trajectory movements of MH370 debris. 4. Results and discussion Ocean circulation is the keystone of determining the MH370 debris drifting across the Indian Ocean. In this view, Figure 2 affords the simulated mechanical phenomenon movements of MH370 (white circles and blue rectangular) that supported multi-objectives of Indian Ocean circulation from Jason-2/Ocean Surface Topography Mission (OSTM), wind pace from QuikSCAT, debris, and ocean bottom topography, severally. Figure 2 suggests that MH370 debris must drift in an anti-clockwise route with the root mean square error of contemporary rate of 10 cm/sec that coincided with the Southern Indian circulation movement (Figure 3). It is attention-grabbing to are searching for out that the MH370 debris below the existing consequences had unsuccessful in the hassle of the Indian Ocean amongst the month of September and October 2014.

Figure 2. Multi-objective algorithm for suspected MH370 debris trajectory movements during (a)MarchApril 2014,(b)May-June, (c) July-August 2014, and (d) September-October 2014.

Figure 3. Southern Indian Ocean circulation. Furthermore, Figure 4 indicates that the Pareto optimization verified that within the water depth of 3000 m the endure MH370 debris of 60% must sink down with the highest cumulative percentage of 95%. As the debris would undergo the impacts of turbulent across the Southern Indian Ocean (Figure 3). In fact, these turbulent flows could be stirred down the MH370 till accumulate on the water depth more than 2000m. In addition, it is difficult to obtain any information about the MH370 debris or fuselage because of the complicated seafloor topography (Figure 1).

Figure 4. Pareto optimization for debris concentration in water depth. In fact, Marghany (2015)and (2017a) (2017b) and (2017c) noted that the dynamic instability, either detritus is additional buoyant than water, within which case they float, or they are less buoyant, within which case they sink. Hence, the turbulent actions with 50 km/ day of the big southern Indian curl with a dimension of one hundred km would cause the detritus to submerged thorough of 3,000 m to 8,000 m across the Southern Indian Ocean. The detritus has been discovered in Réunion Island do not appear to belong to MH370. In fact, the detritus would sink beneath the ocean surface of 3000 water depths at intervals much less than a few months as defined above. If there is no clue confirms the existence of particles both from far off sensing records or floor search throughout the Southern Indian Ocean, this implies the MH370 have landed vertically via the ocean surface and stony-broke proper down to many items through the water column as an end result of the immense hydrostatic pressure of 29,430,000 Pa. This confirms the notion of subgenus Chen et al., (2015) (Marghany et al., 2016). Nevertheless, the part of the detritus would no longer have floated for many months at the water’s surface alternatively would have drifted underwater a thousand meters deep. In fact, the Antarctic Circumpolar Current (ACC) can motive instabilities for the detritus flight movements. During this concern, the MH370 detritus may additionally transport westward and spin in a very large scale counter-clockwise eddy rotation and drifted westward to the African east i.e. Mozambique and Madagascar coastal waters (Marghany 2017c).

5. CONCLUSIONS This investigation has used optimization methods of the Genetic algorithm to look at the influence of ocean surface circulation on flight MH370 debris. The southern Indian Ocean throughout the months of March-April has dominated by using anticlockwise massive gyre transferring with the most velocity rate of 0.5 m/s and slowly drifts westward. It potentially that flight MH370 particles can potentially path up to 50 km/day with massive eddies of a width of one hundred km wide. The find out about indicates that flight MH370 particles could not pass to Africa within a period of the 24 months and with much less than 2 months it would sink earlier than washing up on Réunion Island. However, it can be stated that the turbulent drift due to massive Southern Indian gyre would make the particles submerged in deep water greater than 2000 m throughout the Southern Indian Ocean. In conclusion, the Pareto algorithm suggests that faux and uncertainty data had been delivered with the aid of satellite data. In conclusion, MH370 ought to be by no means fly and crash in the offshore of Perth, Australia. References Anderson S.J., Edwards P.J., Marrone P., and Abramovich Y.A. 2003. Investigations with SECAR - a bistatic HF surface wave radar, Proceedings of IEEE International Conference on Radar, RADAR 2003, Adelaide. Anderson S.J., Darces M., Helier M., and Payet N. 2013 . Accelerated convergence of Genetic algorithms for application to real-time inverse problems, Proceedings of the 4 th Inverse Problems, Design and Optimization Symposium, IPDO-2013, Albi, France, 149-152. Anderson S.J., 2013. Optimizing HF Radar Siting for Surveillance and Remote Sensing in the Strait of Malacca IEEE Tran. on Geosc. and Rem. Sens., 51, 1805-1816. Anderson S.J., 2014. HF radar network design for remote sensing of the South China Sea: In Marghany M.(ed.), Advanced Geoscience Remote Sensing. Intech, Retrieved August 10, 2014, from http://cdn.intechopen.com/pdfs-wm/46613.pdf. Asia News 2014. Missing Malaysian flight MH370: Is satellite data not enough? 2014 Geospatial World, 9:13. Chen, G., Gu, C., Morris, P. J., Paterson, E. G., Sergeev, A., Wang, Y. C., & Wierzbicki, T. (2015). Malaysia Airlines Flight MH370: Water Entry of an Airliner. Notices of the American Mathematical Society, 62(4), 330344. Excell J. 2014. Down deep. The Engineer,

296 3.

Grady B. 2014 NSR Analysis: OU Or Contribution The Business of Pre-Planning For Breaking News. Sat magazine, June , 2014 p.60. Geoscience Australia (2015). MH370: Bathymetric Survey. http://www.ga.gov.au/about/what-wedo/projects/marine/mh370-bathymetric-survey. [Access on August 29 2015]. Linlin G., (2014). Opinion can satellites help find flight MH370?https://newsroom.unsw.edu.au/news/sciencetechnology/can-satellites-help-find-flight-mh370. [Access on August 28 2015]. Marghany M. (2014). Developing genetic algorithm for surveying of MH370 flight in Indian Ocean using altimetry satellite data. 35th Asian conference of remote sensing, at Nay Pyi Taw, Mynamar, 27-31 October 2014. a-a-r-s.org/acrs/administrator/components/com.../OS-081%20.pdf. Marghany M. (2015). Intelligent Optimization system for uncertainty MH370 debris detection. 36 th Asian conference of remote sensing, at the Crowne Plaza Manila Galleria in Metro Manila, Philippines, 19-23 October 2015. acrs2015.ccgeo.info/proceedings/TH4-5-6.pdf.

Marghany, M., Mansor, S. and Shariff, A.R.B.M., 2016. Genetic algorithm for investigating flight MH370 in Indian Ocean using remotely sensed data. In IOP Conference Series: Earth and Environmental Science(Vol. 37, No. 1, p. 012001). IOP Publishing. Marghany,M. 2017a. Simulation of Indian ocean circulation impacts on MH370 debris using multi-objective evolutionary algorithm and Pareto optimal solution. 38th Asian Conference on Remote Sensing - Space Applications: Touching Human Lives, ACRS 2017; The Ashok Hotel New Delhi; India; 23 October 2017 through 27 October 2017. Marghany M 2017b. Multi-Objective Evolutionary Algorithm for MH370 Debris. Ann Mar Biol Res 4(1): 1020,pp:1-6. Marghany, M. 2017c. Multi-objective optimization evolutionary algorithm for investigation of fake MH370 debris.International Journal of Civil Engineering & Geo-Environmental (Special Publication for NCWE2017). pp.108-113. Martini K. (2015). How currents pushed debris from the missing Malaysian Air flight across the Indian Ocean to Réunion. Deep sea news [ http://www.deepseanews.com/2015/07/how-currents-pushed-debris-from-the-missingmalaysian-air-flight-across-the-indian-ocean-to-reunion/] Staff writer (2014). Missing flight satellite finds 122 floating objects http://whotv.com/2014/03/26/missingflight-satellite-finds-122-floating-objects/[Acess on August 28 2015]. Serafino, G. (2015). "Multi-objective Aircraft Trajectory Optimization for Weather Avoidance and Emissions Reduction." Modelling and Simulation for Autonomous Systems. Springer International Publishing, 226-239 Smith, W.H. and Marks, K.M., 2014. Seafloor in the Malaysia airlines flight MH370 search area. Eos, Transactions American Geophysical Union, 95(21), pp.173-174. Zweck J. 2014a. How Satellite Engineers are Using Math to Deduce the Flight Path of the Missing Malaysian Airliner. Retrieved August 10, 2014, from www.utdallas.edu/~zweck/MH370.pdf. Zweck J. 2014b. How Did Inmarsat Deduce Possible Flight Paths for MH370? SIAM News. Retrieved August 10, 2014, from http://www.siam.org/news/news.php?id=2151.

EVALUATING THE EFFICIENCY OF PLEAIDES AND SPOT 6 MULTISPECTRAL FUSION IMAGE FOR MAPPING CORAL REEF SUBSTRATES IN SELINGAN ISLAND, SABAH, MALAYSIA Roslinah Samad (1), Shimatun Jumani Ibrahim (1), Md. Nazri Safar (1), Hazil Sardi Soliano (1) 1

Malaysian Remote Sensing Agency Ministry of Energy, Science, Technology, Environment and Climate Change No. 13, Jalan Tun Ismail, 50480 Kuala Lumpur, Malaysia Email: [email protected], [email protected], [email protected], [email protected]

KEY WORDS: Coral reef substrates, multispectral, fusion image, supervised classification, accuracy assessment ABSTRACT: Mapping coral reef substrate using traditional ground base method is inefficient and very costly. However, satellite remote sensing is one of the technologies that have the potential to map accurately the distribution of coral reef substrates. The spatial component of the coral reef can be delineated using high resolution satellite imagery. Coral reef diversity has rapidly degradating over the last decade so the need for an accurate distribution coral reef substrates map is very crucial for better planning, management and monitoring of coral reef area. This paper evaluated the efficiency of using image fusion techniques in the mapping of coral reef substrates. The aim of the study was to determine and compare the accuracy of coral reef substrates classification between multispectral and fusion image of Pleaides and SPOT 6 data. Multispectral data of high resolution Pleaides and SPOT 6 satellite imageries were used in the study. Several radiometric correction techniques such as conversion from digital number to radiance and reflectance, atmospheric correction, water column correction and sun glint correction were applied to the individual Pleaides and SPOT 6 multispectral images and fusion images. The enhanced imageries were then classified using maximum likelihood classification in order to generate coral reef substrates map. The accuracy of each coral reef substrates classification map was subsequently derived from the accuracy assessment for each classification. The study had found that Pleaides data produced the most accurate results of coral reef substrates classification with 89.20% of overall accuracy. This was followed by fusion image of Pleaides with SPOT 6 which produced 87.20% of overall accuracy. SPOT 6 data produced the lowest accuracy of coral reef substrates classification with 86.80% of overall accuracy. The maximum likelihood classification clearly distinguished coral reef substrates within the study area into five classes, namely, live coral, dead coral, rubble, sand and unclassified. The study concludes that the fused multispectral image of Pleaides with SPOT 6 did not improved the accuracy level of coral reef classification compared with the original image of Pleaides which produced the highest overall accuracy.

1. INTRODUCTION Coral reef ecosystem is one of the most complex marine environmental systems which play an important role in maintaining marine biological diversity for future generation. It becomes an area for spawning, nursery, breeding and feeding ground of many marine species. Coral reef ecosystem also provides benefits to the country as it generates income for the local communities through fishery, marine-based industries and tourism activities. Most of the islands surrounded by coral reef ecosystem have beautiful beaches with clear blue waters and rich in the diversity of marine flora and fauna species. However, most of these islands are highly threatened as tourism activities leads to pressure the coral reef ecosystem. With the increase in tourist arrivals, there would be an impact such as damage of coral reef covers and marine pollution that caused declining in the coral reef areas distribution in the country. In order to maintain and protect the distribution of these coral reef areas and at the same time fulfilling the demand from tourism activities, relevant agencies should take the necessary action to map and monitor coral reef ecosystems for sustainable tourism industry. Mapping coral reef substrate using traditional ground base method is inefficient and very costly. However, satellite remote sensing is one of the technologies that have the potential to map accurately the distribution of coral reef substrates at landscape scale. On the basis that different coral reef substrates have distinctive optical properties, the spatial component of the coral reef can be delineated using high resolution satellite imagery. Coral reef diversity

has rapidly been degrading over the last decade, so the need for an accurate distribution of coral reef substrates map is very crucial for better planning, management and monitoring of coral reef areas. Remote sensing technologies have been recognized as a useful tool to monitor and map coral reef substrate. However, the accuracy of identifying coral reef substrates has been limited due to the spectral and spatial resolution of the satellite sensors. This accuracy also depends on the processing techniques being adopted to extract coral reef information from these satellite imageries. In order to know which satellite sensors are better for mapping coral reef substrates, the study evaluated the efficiency of different sensors from two satellites namely, Pleaides and SPOT 6. Both satellites produce high resolution multispectral data that are suitable for mapping coral reef substrates. The aim of the study was to determine and compare the accuracy of coral reef substrate classification between multispectral and fusion image of Pleaides and SPOT 6 data.

2. MATERIALS AND METHODS 2.1 Study Area The area selected in the study was Selingan Island. Selingan Island is one of the three islands that was gazetted as a Turtle Island Park under the Sabah State. Selingan Island lies 40 km north of Sandakan, Sabah in the Sulu Sea. The other two islands gazetted in this marine park are Gulisan Island and Bakungan Kecil Island. The park is famous for the endangered green and hawksbill turtle nesting in the Sabah State where the turtle eggs are placed in open-air hatcheries set up since 1966 (Maipol, 2001). However, the quality of coral reefs in the park had deteriorated due to the economic and human needs, such as tourism activities. The general survey carried out by the Marine Research Unit of Sabah Parks in 1998 showed that the average live coral cover for the three islands in the Turtle Island Park was about 27%, while the rest was dead coral, sand and rubble (Maipol, 2001). 2.2 Data High resolution data of Pleaides multispectral (2 m resolution) dated 11 March 2017 and SPOT 6 multispectral (6 m resolution) dated 3 September 2017 were used in the study. Both satellites used visible and near infrared wave lengths to detect information on the earth surface but both of them are using different range of wavelengths. The characteristics of multispectral data from Pleaides and SPOT 6 satellites are shown in Table 1. Table 1: The characteristics of multispectral data from Pleaides and SPOT 6 imageries Spectral Bands Wavelength Range Pleaides SPOT 6 Blue 0.430 – 0.550 µm 0.450 – 0.520 µm Green 0.500 – 0.620 µm 0.530 – 0.590 µm Red 0.590 – 0.710 µm 0.625 – 0.695 µm Near Infrared 0.740 - 0.940 µm 0.760 – 0.890 µm Ground truth data of coral reef substrates jointly collected by the Malaysian Remote Sensing Agency and The Board Trustee of Sabah Parks on 10-11 April 2018 was used to verify image classification of coral reef substrates in Selingan Island. 2.3 Methods The study used Erdas Imagine Version 2016 image processing system to process multispectral data of Pleaides and SPOT 6 satellite images. Several image processing techniques including radiometric correction, atmospheric correction, water column correction, sun glint correction, image fusion, image classification and accuracy assessment were applied on the individual multispectral data of Pleaides and SPOT 6 and also on the fusion image of both satellites. 2.3.1 Radiometric Correction Remote sensing signals are essentially the amount of energy received at the sensor from the target in a given spectral width of the satellite sensor. However, signals received by the sensors usually contain noise. There are two types of radiometric noise namely, internal and external noise. Internal noise comes from the satellite sensor while external noise comes from the atmosphere and areas adjacent to the target. In order to get accurate information from the satellite imagery, radiometric correction should be applied on the satellite imagery before applying other image

processing techniques. The purpose of radiometric correction is to reduce or correct errors in the digital numbers of image. The radiometric correction process improved the interpretability and quality of remotely sensed data. The capability of remote sensing in identifying and mapping coral reef substrates can be influenced by radiometric factor such as atmospheric scattering, the effect of ocean sun glint and water column. These factors will influence the amount of radiation reaching the sensors and therefore reduce the accuracy of remotely sensed data. Therefore radiometric limitation of sensors should be considered and corrected before applying any imaging analysis. There were several types of radiometric correction process that involved in the study as mention below:2.3.1.1 Conversion Digital Number to Radiance The generic term of pixel values is Digital Number where it is commonly use to describe pixel value that has not been calibrated into physical meaningful units. Radiance is the amount of radiation coming from an area. In order to derive a radiance image from an uncalibrated image, a gain and offset must be applied to the pixel values. These gain and offset values are typically retrieved from the image metadata. The digital numbers of Pleaides and SPOT 6 data were converted to radiance using algorithm provided by Astrium (2013) as shown below: Lb(p) = DC(p) + BIAS(b) ……………………………… (1) GAIN(b) where Lb(p) is Top of Atmosphere (TOA) Radiance and DC(p) is a digital count or digital number. 2.3.1.2 Conversion Radiance to Reflectance Reflectance is the proportion of the radiation striking a surface to the radiation reflected off it. Some material can be identified by their spectral reflectance so it common to correct image to reflectance as a first step toward identifying features in an image. The radiance of Pleaides and SPOT 6 data were converted to reflectance using algorithm provided by Astrium (2013) as shown below: Pb(b) =

𝜋. 𝐿𝑏(𝑝) …………….………………………… (2) Eo(b).cos(𝜃s)

where Eo(b) is Solar irradiance of the band, 𝜃s is Sun Zenith Angle. 2.3.1.3 Atmospheric Correction In order to make a meaningful measure of radiance at the earth’s surface, the atmospheric interferences such as moisture contents, haze and others must be removed from the satellite imagery. The process is called atmospheric correction. The main purpose atmospheric correction was carried out in this study is to remove the contribution of scattering in the atmosphere and reflection from the sea surface from the top of atmosphere radiances measure by a sensor in the visible region of the spectrum. In this study, dark object subtraction method was used to remove atmospheric attenuation and scattering on the reflectance data of Pleaides and SPOT 6 imageries. Dark object subtraction searched each band for the darkest pixel value. The scattering in the reflectance data was removed by subtracting the value from every pixel in the band. 2.3.1.4 Water Column Correction A water column is a conceptual column of water from the surface of a sea to the bottom sediment. According to Mumby and Edwards (2000), the intensity of light decreased exponentially with the increasing depth when light penetrates water which is known as attenuation. Therefore, coral reef substrates that are located under water are covered by light attenuating water column obscuring object discrimination. In order to improve the visual interpretation of satellite imagery and the accuracy of coral reef substrates classification, the effect of light attenuation in the water column has to be reduced by applying depth-invariant index algorithm developed by Mumby and Edwards (2000) on the atmospherically corrected image. In relatively clear water, the intensity of light will decay exponentially with increasing depth. In this study, the pixel values of the sea bed (sand) for each band were selected with increasing the depth. The pixel values of the sea bed (sand) were then transformed to create linear relationship between depth and radiance using the following natural logarithm: Xi = ln (Li) ………………………………………………..…… (3)

where Li is the pixel radiance in band i. The ratio of attenuation coefficients for band pairs were then calculated from the slope of the bi-plot made of log transformed radiances for the sand substratum at differing depths. Then depth-invariant index of bottom type were generated using the following equation: Depth-invariant indexij = ln(Li) - ki ln(Lj) …………………………… (4) kj where ki is the ratio of attenuation coefficient. kj 2.3.1.5 Sun Glint Correction Sun glint is a phenomenon that occurs when sun light reflects off the surface of the ocean at the same angle that a satellite or sensors viewing the ocean surface. This is called as sun glint effect where the direct sun light was reflected as a specular reflection. In the affected area of the image, smooth ocean water becomes a silvery mirror, while rougher surface waters appear dark. Sun glint may cause misclassification and poor accuracy for mapping coral reef substrates. There are several methods for glint removal from high resolution imagery. In this study, glint removal was performed on atmospherically corrected image of Pleaides using a technique developed by Hedley et.al (2005). In order to apply this technique, a linear regression was performed between the near infrared (NIR) brightness and the brightness in the visible band using a sample set of pixels. The slope of regression was then used to remove glint using the following equation: Ri’ = Ri – bi (RNIR – Min NIR) ………………………………………... (5) where Ri’ is the sun glint corrected pixel brightness in band i, Ri is the pixel value in band i, bi is the regression slope, RNIR is the pixel NIR value and Min NIR is the minimum NIR brightness of a pixel with no sun glint. Sun glint correction has not performed on the SPOT 6 data due as the surface of ocean was free from the sun glint coverage. 2.3.2 Image Fusion Multispectral image fusion is the process of combining relevant information from two or more images into a single image. The fused image is more informative and accurate than any single source image, and it consists of all the necessary information. The fused image can have the complementary spatial and spectral resolution characteristics. The purpose of image fusion is not only to reduce the amount of data but to construct a single image that will enhance the information extraction. The advantages of image fusion include sharpened image, feature enhancement and improved classification. In the study, we performed High Pass Filtering (HPF) fusion techniques on the sun glint corrected multispectral image of Pleaides and water column corrected multispectral of SPOT 6. 2.3.3 Image Classification The fusion image of Pleaides and SPOT 6 was subsequently classified into five classes of coral reef substrates using maximum likelihood classifier. In order to compare which satellite imagery gives higher accuracy in the classification of coral reef substrates, this study also performed maximum likelihood classification on the individual sun glint corrected multispectral image of Pleaides and water column corrected multispectral image of SPOT 6. 2.3.4 Accuracy Assessment Accuracy assessment is the most important part of image classification. It compares the classified image with ground data in order to assess the accuracy of classified map. The accuracy assessment was performed on the classified fused image and on the individual classified Pleaides and SPOT 6 images. The study created a set of 250 of random points for each classified image in order to generate confusion matrix of the accuracy.

3. RESULTS AND DISCUSSION The results of radiometric correction showed that after calibrated the multispectral of Pleaides data in digital number (Figure 1(a)) to radiance (Figure 1(b)), the features in the radiance image of Pleaides look brighter, sharp and more details due to the process improves the interpretability and quality of remotely sensed data. The data in radiance which was then converted to the reflectance value was gave a result with the surface of ocean became more smoothly as shown in Figure 1(c).

Figure1(a): The original multi spectral image of Pleaides in digital number values

Figure 1(b): Multi spectral image of Pleaides after the digital number converted to radiance

Figure 1(c): Multispectral image of Pleaides after conversion from radiance to reflectance

Atmospheric correction procedure using Dark Object Subtraction was then performed on the reflectance multispectral image of Pleiades. The result showed that the ocean surface became more smoothly as shown in Figure 1(d) compared with the image before atmospheric corrected (Figure 1(c)). In this study, the result of exponential attenuation of radiance with depth linearized for Pleaides bands using natural logarithms is shown in Figure 1(e). The results of linear relationship between depth and radiance using the natural logarithm are shown in Figure 1(e), Figure 1(f), Figure 1(g) and Figure 1(h). The results of linear regression showed that bi-plot of log transformed Pleaides bands 2 and 3 gave the significant correlation with higher R2 = 0.929 compared to other bands. Therefore, the study used regression slope 0.8689 as a ratio attenuation coefficient that generated from this linear regression to run water column correction on the atmospheric corrected image of Pleaides data. Figure 1(i) showed the result of the reflectance image of Pleaides after applied water column correction. The resultant image showed that it increased the visibility of the substrate features under water coverage and look details. Water column correction minimise the confusing effects caused by different depths of water, therefore it improves the visibility of coral reef substrates in the corrected image.

Figure 1(d): Multispectral image of Pleaides after done atmospheric correction

Figure 1(e): Graph of exponential attenuation of radiance with depth linearized for Pleaides bands using natural logarithms

Figure 1(f): Graph of bi-plot of log-transformed Pleaides bands 1 and 2 for a unique substratum at various depth

Figure 1(g): Graph bi-plot of log transformed Pleaides bands 2 and 3 for a unique substratum at various depth

Figure 1(h): Graph bi-plot of log-transformed Pleaides bands 3 and 4 for a unique substratum at various depth

In this study, several samples of pixel values for sun glint brightness in each visible band and samples of pixel values for deep water region in near infrared (NIR) band were selected and transformed into linear regression. Sun glint correction was carried out by establishing the linear relationship between near infrared (NIR) brightness and the amount of sun glint in each visible band as shown in Figure 1(j), Figure 1(k) and Figure 1(l). The results showed that linear regression between visible band 1 and near infrared band 4 gave the significant correlation with higher R2 = 0.1111 compared to other bands. Therefore, the study used regression slope -0.6009 to run sun glint correction using water column corrected image of Pleaides. These effects of sun glint were clearly observed in the wave pattern before sun glint correction procedure as shown in Figure 1(i). Figure 1(m) showed the visual effect of the removal sun glint. As a result of glint corrected image, the ocean surface looks very smooth and the subsurface features were enhanced detail. The proposed glint correction method improved the visual interpretation of the image in the region cover with water.

Figure 1(i): Pleaides image after water column correction

Figure 1(j): Graph of linear relationship between visible band 1 and near infrared band 4

Figure 1(k): Graph of linear relationship between visible band 2 and near infrared band 4

Figure 1(l): Graph of linear relationship between visible band 3 and near infrared band 4

Figure 1(m): Pleaides image after sun glint correction

The study also applied radiometric correction on SPOT 6 multispectral image before apply fusion technique on both Pleaides and SPOT 6 images. The results of radiometric correction showed that after calibrated the multispectral of SPOT 6 data in the digital number (Figure 2(a)) to radiance (Figure 2(b)), the features in the radiance image of SPOT 6 looks brighter, sharp and more details due to the process improves the interpretability and quality of remotely sensed data. The data in radiance which was then converted to the reflectance value was gave a result with the surface of ocean became more smoothly as shown in Figure 2(c).

Figure 2(a): The original multispectral image of SPOT 6 in digital number values

Figure 2(b): Multi spectral image of SPOT 6 after the digital number converted to radiance

In this study, the result of exponential attenuation of radiance with depth linearized for SPOT 6 bands using natural logarithms is shown in Figure 2(e). Figure 2(e) showed the exponential attenuation of radiance with depth linearized for SPOT 6 bands using natural logarithms. The results of linear relationship between depth and radiance using the natural logarithm are shown in Figure 2(f), Figure 2(g) and Figure 2(h). The results of linear regression showed that bi-plot of log transformed SPOT 6 bands 3 and 4 gave the significant correlation with higher R2 = 0.3828 compared to other bands. Therefore, the study used regression slope 0.3632 as a ratio attenuation coefficient that generated from this linear regression to run water column correction on the image of SPOT 6. Figure 2(i) showed the result of the image of SPOT 6 after applied water column correction. The resultant image showed that the substrate features under water coverage became brighter, clearly and looks details due to water column correction minimise the confusing effects caused by different depths of water. It improves the visibility of coral reef substrates in the corrected image.

Figure 2(c): Multispectral image of Pleaides after conversion from radiance to reflectance

Figure 2(d): Multispectral image of SPOT 6 after done atmospheric correction

Figure 2(e): Graph of exponential attenuation of radiance with depth linearized for SPOT 6 bands using natural logarithms

Figure 2(f): Graph of bi-plot of log-transformed SPOT 6 bands 1 and 2 for a unique substratum at various depth

Figure 2(g): Graph of bi-plot of log-transformed SPOT 6 bands 3 and 3 for a unique substratum at various depth

Figure 2(h): Graph of bi-plot of log-transformed SPOT 6 bands 3 and 4 for a unique substratum at various depth

The study has not applied sun glint correction on the SPOT 6 image as the ocean surface was free from the sun glint coverage. The resultant of radiometric corrected multispectral images of Pleaides and SPOT 6 were subsequently combined together using image fusion technique. Image fusion was carried out using multispectral of Pleaides and SPOT 6 data with different spectral and spatial resolution to produce an image with enhanced spectral and spatial resolution. Spatial resolution for Pleaides data is 2 m while SPOT 6 data is 6 m. The study was applied High Pass Filtering fusion technique on the individual band of higher resolution image (Pleaides) with the all bands of lower resolution image (SPOT 6). The results of fusion image between Pleaides and SPOT 6 are shown in Figure 3(a), Figure 3(b), Figure 3(c) and Figure 3(d).

The results showed that fused image generated by the individual band 2 of Pleaides multispectral data with the all band of SPOT 6 multispectral data gave the best result compared with the results generated by band 1, band 3 and band 4 of Pleaides. It shows that the interpretability and visibility of coral reef substrates in the fusion image of band 2 Pleaides was increased compared to images fusion of other bands that reduced the visibility and interpretability of coral reef features. However, the resultant fused image mostly contains the spectral characteristics of SPOT 6 compared with the spectral and spatial characteristics of Pleaides data. Normally, the spectral information contained in the low resolution information is preserved by fused method. Therefore, low resolution of multispectral image SPOT 6 gave the colour information of the fused image.

Figure 2(i): Multispectral image of SPOT 6 after applied water column correction

Figure 3(a): Fused multispectral image of SPOT 6 with band 1 of Pleaides data

Figure 3(b): Fused multispectral image of SPOT 6 with band 2 of Pleaides data

Figure 3(c): Fused multispectral image of SPOT 6 with band 3 of Pleaides data

Figure 3(d): Fused multispectral image of SPOT 6 with band 4 of Pleaides data

The fused image was then performed sun glint correction in order to enhance the detail of subsurface features or coral reef substrate. As a result, the pixels which contain sun glint information were removed from the fused image. The ocean surface in the resultant image looks very smooth and the subsurface features were enhanced detail. The sun glint corrected of Pleaides with SPOT 6 fusion image was then classified coral reef substrates using supervised classification with the maximum likelihood classifier into five classes, namely, live coral, dead coral, rubble, sand and unclassified. In order to evaluate the efficiency of Pleaides and SPOT 6 fusion image in mapping coral reef substrates, the study also applied coral substrates classification using the same method to the individual radiometric corrected image of Pleaides and SPOT 6.

Figure 3(e): The result of sun glint correction applied on fused multispectral image of SPOT 6 with band 2 of Pleaides data

The results of supervised classification for the individual Pleaides image, the individual SPOT 6 image and the fusion image of Pleaides with SPOT 6 are shown in Figure 4(a), Figure 4(b) and Figure 4(c). The accuracy

assessment was subsequently carried out on each of the resultant classification images using the ground-truth data dated on 10-11 April 2018. The ground truth data was used as a reference in the verification of 250 random point samples that generated by computer in order to create the accuracy of coral reef substrates classification map.

Figure 4(a): Coral reef substrates classification of Pleaides image before fusion using maximum likelihood classifier

Figure 4(b): Coral reef substrates classification of SPOT 6 image before fusion using maximum likelihood classifier

Figure 4(c): Coral reef substrates classification of Pleaides and SPOT 6 fusion image using maximum likelihood classifier

The results of the accuracy assessment for the individual Pleaides classification image, the individual SPOT 6 classification image and the classification fusion image of Pleaides and SPOT 6 are shown in the Confusion Matrix in Table 2, Table 3 and Table 4. Table 2: Confusion Matrix for coral reef substrates classification of Pleiades image before fusion using maximum likelihood classifier Class Name Live Coral Dead Coral Rubble Sand Unclassified Total Producer User Accuracy Accuracy Live Coral 1 6 0 0 64 82.61% 89.06% 57 Dead Coral 0 0 2 0 15 86.67% 86.67% 13 Rubble 12 1 1 0 50 80.00% 72.00% 36 Sand 0 0 3 0 17 77.78% 82.35% 14 Unclassified 0 0 0 1 104 100.00% 99.04% 103 Total 69 15 45 18 103 250 Overall Classification Accuracy = 89.20%, Overall Kappa Statistics = 0.8486 Table 3: Confusion Matrix for coral reef substrates classification of SPOT 6 image before fusion using maximum likelihood classifier Class Name Live Coral Dead Coral Rubble Sand Unclassified Total Producer User Accuracy Accuracy Live Coral 1 5 1 3 34 70.59% 72.73% 24 Dead Coral 6 2 3 2 21 38.10% 80.00% 8 Rubble 2 0 1 0 17 82.35% 63.64% 14 Sand 1 1 1 0 17 82.35% 63.64% 14 Unclassified 3 0 0 0 161 97.52% 96.91% 158 Total 36 10 22 19 163 250 Overall Classification Accuracy = 86.80 %, Overall Kappa Statistics = 0.7597 Table 4: Confusion Matrix for coral reef substrates classification of Pleaides and SPOT 6 fusion image using maximum likelihood classifier Class Name Live Coral Dead Coral Rubble Sand Unclassified Total Producer User Accuracy Accuracy Live Coral 0 1 0 0 29 62.22% 96.55% 28 Dead Coral 0 0 0 1 4 37.50% 75.00% 3 Rubble 9 0 0 0 12 27.27% 25.00% 3 Sand 5 5 7 1 34 100.00% 47.06% 16 Unclassified 3 0 0 0 171 98.82% 98.25% 168 Total 45 8 11 16 170 250 Overall Classification Accuracy = 87.20 %, Overall Kappa Statistics = 0.7454

The study had found that Pleaides data produced the most accurate results of coral reef substrates classification with 89.20% of overall accuracy. The maximum likelihood classification clearly distinguished coral reef substrates within the study area into five classes, namely, live coral, dead coral, rubble, sand and unclassified. This was followed by fusion image of Pleaides with SPOT 6 which produced 87.20% of overall accuracy for coral reef substrate classification. The study found that fusion image of Pleaides with SPOT 6 could not classified dead coral and rubble accurately where it only gave Producer Accuracy 37.50% for dead coral class and 27.27% for rubble class respectively as shown by Confusion Matrix in Table 3. This was happened due to the majority of dead coral class and rubble class had found mixed with the other classes. However, fusion image of Pleaides with SPOT 6 data classified accurately sand class where it produced 100% Producer Accuracy. The study found that SPOT 6 data produced the lowest accuracy of coral reef substrates classification with 86.80% of overall accuracy. Confusion Matrix in Table 3 showed that SPOT 6 data could not classified dead coral accurately where it only gave Producer Accuracy 38.10% due to the majority of dead coral class had found mixed with the other classes.

4. CONCLUSION The study discusses the efficiency of fused multispectral image Pleaides with SPOT 6 in the mapping coral reef substrates compared with the original image of Pleaides and SPOT 6. The study concludes that the fused multispectral image Pleaides with SPOT 6 did not improved the accuracy level of coral reef classification which it only produced 87.20% of overall accuracy compared with the original image Pleaides which produced the highest overall accuracy 89.20%. However, the fused image of Pleaides with SPOT 6 produced coral reef classification substrates map more accurately when it compared with the original image of SPOT 6 which only produced 86.80% overall accuracy. The study finally concludes that not all fusion images would be improved the accuracy of coral reef substrates classification compared with the high resolution of original image. The reason why fusion image of Pleaides with SPOT 6 gave poor classification accuracy was the spectral information contained in the low resolution information SPOT 6 was preserved by fused method. High Pass Filtering fusion technique used in this study maintain the spectral information from low resolution image well but the spatial details from high resolution image are not satisfactorily injected, thus the resultant image not achieving a good spatial enhancement. Therefore, low resolution of multispectral image SPOT 6 gave the colour information of the fused image, that’s why both the fusion image and the original image of SPOT 6 produced the accuracy of coral reef classification not far from each other. Further study using different fusion techniques needs to be implemented in order to generate fusion image of Pleaides and SPOT 6 with the best spatial and spectral quality enhancement, thus it suppose will improves the accuracy level of coral reef substrates classification.

5. ACKNOWLEDGEMENT We would like to express our gratitude to The Board Trustee of Sabah Parks for their assistance provided us with the logistic and staffs during collection of ground truth data in Turtle Island Park, Sandakan, Sabah. We are also thanks to Malaysian Remote Sensing Agency who gave us the opportunity to do this research paper.

6. REFERENCES Astrium. 2013. SPOT 6 and SPOT 7 Imagery – User Guide: Spectral Modelling. Toulouse, France. Pp. 101-103. Edurne, I.U, Consuelo, G.M., Javier, M.R., Angle, G.P. and Dionisio, R.E. 2017. Fusion of High Resolution Multispectral Imagery in Vulnerable Coastal and Land Ecosystems. Published online 2017 Jan 25. doi: 10.3390/s17020228. Retrieved September 20, 2018 from http://creativecommons.org/licenses/by/4.0/ Hedley, J.D., Harborne, A.R. and Mumby, P.J. 2005. Technical Note: Simple and Robust Removal of Sun Glint for Mapping Shallow Water benthos. International Journal of Remote Sensing, Vol.26, No. 10, 20 May 2005, pp. 2107-2112. Maipol Spait, 2001. Marine Park Management: Issues and Challenges. 6th Sabah Inter-Agency Tropical Ecosystem (SITE) Research Seminar, Tanjung Aru, Kota Kinabalu, Sabah, Malaysia. 13-14 September 2001. Mumby, P.J., Edwards, A.J. 2000. Water Column Correction Techniques, pp. 121-128, in: Green, E.P., Mumby, P.J., Edwards, A.J., Clark, C.D., (Ed. A.J. Edwards). Remote Sensing Handbook for Tropical Coastal Management. Coastal Management Sourcebook 3, UNESCO, Paris.

SUSPENDED SEDIMENT CONCENTRATION MAPPING AT TEMENGGOR LAKE USING LANDSAT-8 TM DATA Hamzah Mohd Ali (1), Abd Wahid Rasib (1), Nur Amalina Aminuddin (1), Othman Zainon (1), Rozilawati Dollah (2), Abdul Razak Mohd Yusoff (1) and Khairulnizam M.Idris (1) 1

TropicalMap Research Group, Faculty of Built Environment and Surveying, Universiti Teknologi Malaysia, Johor Bahru, Johor, Malaysia. 2 School of Computing, Faculty of Engineering, Universiti Teknologi Malaysia, Johor Bahru, Johor, Malaysia. Email: [email protected]

KEY WORDS: Suspended sediment, Temenggor Lake and Landsat-8 TM ABSTRACT: Sediment is a natural occurring material that is broken down by the process of climate and soil erosion whereby transported by the action of wind, water or by the force of gravity acting on the particles. From the satellite data, the total concentration of suspended sediment in the inland water able to be mapped using remote sensing technique processing. Remote sensing is the acquisition of information about an object or phenomenon without making physical contact with the object and therefore not comparable to in-situ observations. In modern usage, the term generally refers to the use of aerial sensor technologies to detect and classify objects on earth by means of propagated signals. In this study, Landsat-8 TM satellite image is used in determining the concentration of suspended sediment at Temenggor Lake, Perak. Remote sensing techniques is used in processing and analysing the correlation of data sampling and satellite data, respectively. This study indicates that band 2 of Landsat-8 TM able to show the good correlation from polynomial equation at R2 = 0.54 in determining the suspended sediment concentration at Temenggor Lake. This study concluded that Landsat-8 TM is able to map the concentration of suspended sediment in inland water at Temenggor Lake.

INTRODUCTION Land management agencies are under increasing pressure to monitor the effects of their activities. One consequence has been a concern about the amount of sedimentation in streams near logging operations in forest stands. Sedimentation can adversely affect water quality as well as fish habitat (Thomas, 1985). Sedimentation that caused by suspended sediment can lead to water pollution in certain area. By analysing and interpreting the content and concentration of suspended sediment, the level of water pollution at the river can be known. Erosion of a river also can contribute to the concentration of suspended sediment. The erosion of a river is mostly caused by the rain and flood of a certain area. Most suspended sediment moves during infrequent high flows that collectively account for only a small portion of the measurement period (Thomas, 1985). The erosion of a river also can give a variation results in determining the concentration of suspended sediment in an area. Besides that, the residential at the nearby river can contribute to the variation and concentration of suspended sediment of the river. As more people live nearby the river area, more activities occurred at the river. The activities of residential such as farming can lead to the pollution of the river. More concentration of suspended sediment exists when more pollution occurs at the area. Sediment is a naturally occurring material that is broken down by processes of weathering and erosion, and is subsequently transported by the action of wind, water, or ice, or by the force of gravity acting on the particles. Seas, oceans and lakes accumulate sediment over time. The sediment could consist of terrigenous material, which originates on land, but may be deposited in terrestrial, marine, or lake environments; or of sediments originating in the body of water. Terrigenous material is often supplied by nearby rivers and streams or reworked marine sediment. Deposited sediments are the source of sedimentary rocks. Lake bed sediments that have not solidified into rock can be used to determine past climatic conditions. There are many types of sediment that existing at the bottom of the river. The various types of sediment can be determined by using remote sensing techniques. Remote sensing is the acquisition of information about an object or phenomenon without making physical contact with the object and thus in contrast to in situ observation. Furthermore, aerial sensor technologies such as air-borne and space-borne sensors have proved to be a useful method to such studies. As it provides an instantaneous and synoptic view of sediments that would otherwise be unavailable (Mobasheri et al., 2003). Suspended sediment is one of the parameter that are important in observing the water quality. Collecting information about (Suspended Sediment Concentration) SSC, in coastal waters and estuaries is vital for proper management of coastal environments (Mobasheri et al., 2003). The suspended sediment in a location have to be studied and analysed before taking any action to the location such as, planning and controlling to the location.

The study of suspended sediment in certain area is important especially to Jabatan Pengairan dan Saliran so that they know the erosion that occurs at the area. When the concentration of suspended sediment is higher, it shows that the area has a higher rate of erosion. The study of suspended sediment also can determine the level or rate of pollution of certain area especially river and lake. Using effective technique can give the best information to the Jabatan Alam Sekitar (JAS) in controlling water pollution. This study has been carried out for the purpose of mapping the concentration of suspended sediment at Temenggor Lake using remote sensing technique. The concentration of suspended sediment at Temenggor Lake extracted from remote sensing technique is also analysed. In this research, satellite image of Landsat 8 is used in determining the concentration of suspended sediment at Temenggor Lake. The result for this research is the concentration of suspended sediment map at Temenggor Lake. This study can save time and cost in mapping the concentration of suspended sediment at Temenggor Lake. Data or information that is obtained by this study can be used by Jabatan Pengairan dan Saliran, Jabatan Alam Sekitar and Tenaga Nasional Berhad.

LITERATURE REVIEW 2.1

Suspended sediment

Sediment is matter that settles to the bottom of a liquid. In Geology, sediment is a particulate matter that is carried by water or wind and deposited on the surface of the land or the seabed, and may in time become consolidated into rock. The word “sediment” is used for all deposition from rivers, lakes and sea. Sediment also can be defined as deposition or dirt that are deposited in the water (Fleming, 1977). According to (Doxaran, 2003), sediment is defined as contaminated impurities content. Sediment can be transported by water and wind. Sediment that is transported by water is called fluvial process whereas sediment that is transported by wind is known as aeolian process. Fluvial process includes the movement of sediment and erosion on the river bed. Aeolian process is a process that involves the ability of wind in transporting and depositing sediment. Mostly, the size of sediment affects the transportation and deposition of sediment. The mapping of the suspended sediment at the river is based on the characteristics of suspended sediment. The various type of suspended sediment at the river can be detected by remote sensing technique. Different sediment gives different results in determining the suspended sediment using remote sensing. Variations of sediment type (grain size and refractive index) and changing illumination conditions affect the reflectance signal of coastal waters and limit the accuracy of sedimentconcentration estimations from remote-sensing measurements (Doxaran, 2003).

2.2

Monitoring suspended sediment

Remote sensing is one of the techniques that has been used in determining the concentration of suspended sediment. Remote sensing is defined as the use of satellite – borne sensors to observe, measure, and record the electromagnetic radiation reflected or emitted by the Earth and its environment for subsequent analysis and extraction of information (Janssen et al., 2001). Remote sensing which is also called earth observation refers to obtaining information about objects or areas at the Earth’s surface without being in direct contact with the object or area (Aggarwal, 2003). In general, remote sensing is the science or method in collecting information and data about objects or areas from a distance, especially from aircraft or satellites. Remote sensors collect data by detecting the energy that is reflected from Earth. The satellite that has been used for this study is Landsat – 8 satellite. This satellite carries two instruments which are the Operational Land Imager (OLI) sensor and Thermal Infrared Sensor (TIRS). The Operational Land Imager (OLI) sensor includes refined heritage bands, along with three new bands which are a deep blue band for coastal or aerosol studies, a shortwave infrared band for cirrus detection and a Quality Assessment band. The Thermal Infrared Sensor (TIRS) sensor provides two thermal bands. Both sensors provide improved signal-to-noise (SNR) radiometric performance quantized over a 12-bit dynamic range. SNR enable better characterization of land cover state and condition. There are several studies has been done for monitoring suspended sediment using satellite images. (Harrington et al., 1999) used water quality data from Lake Chicot, Arkansas and a corresponding set of Landsat MSS data to compare the ability of satellite – based sensor systems to monitor suspended sediment concentration, Secchi disk depth, and turbidity. Secchi disk depth and nephelometric turbidity are both optical measures of water quality and differ from suspended sediment concentration, which is a measure of the weight of inorganic particulates suspended in the water column. (Nellis et al., 1998) had apply an existing physical model that uses at-satellite reflectance for TM Band 3 to

estimate variations in suspended sediment, turbidity, and Secchi depth throughout the reservoir in four dates. Remote sensing can be assist in documenting a relatively short-term environmental hazard such as flood. This research also demonstrates the value of Landsat Thematic Mapper data for mapping geographic variations in water area and quality in conjunction with a major flood event. (Schaap et al., 2002) states that the study of sediment accumulation will help in conserving the value of the lake as a water source and recreational area as we able to estimate the rate of sediment accumulation in the lake and able to describe the distribution of sediment in the lake. (Pimstein et al., 2014) shows that the satellite images can be used for estimating the extent of sediment dispersal in the Dead Sea in order to add more spatial information for understanding of the transport and deposition processes. Spatial anomalies were computed in order to characterize the sediment distribution along the year and during specific flash flood events. The preliminary results of this research also are intended to be validated during a future field campaign after the next flash-flood event.

RESEARCH METHODOLOGY 3.1

Study area

The study area is located at Temenggor Lake, Gerik, Perak. Temenggor Lake is the second largest lake in Peninsula Malaysia after Kenyir Lake in Terengganu, Malaysia. This lake is man-made located south of 1,533 m high Ulu Titi Basah peak in Hulu Perak district in the state of Perak. It was created after the construction of Temenggor Dam to generate electric power. The location of the lake is about 45 km from the Hulu Perak district capital, Gerik. There is a man-made island, Banding Island and Lake Temenggor Bridge on the East-West Highway which crosses the lake (refer Figure 1).

Figure 1 Study Area

3.2

Data acquisition

Data acquisition The in-situ data collection has been carried out at Tasik Temenggor, Royal Belum, Perak. 60 samples of lake water have been collected at 60 points of different region as shown in Figure 2. The other parameter is also collected during the in-situ data collection such as pH, conductivity, turbidity, D.O, water temperature, salinity, wind speed, air temperature and the depth of secchi disk. The parameter is collected with HORIBA U-10, secchi disk and wind speed (refer Figure 3).

Figure 2 Water sampling location

Figure 3 The utilization of HORIBA U-10 The water samples then being processed at environment laboratory of Civil Engineering Faculty UTM. The water samples are processed by vacuum filtration method (refer Figure 4). Vacuum water filtration method involved in the separation of suspended solid from water sample.

Figure 4 Vacuum filtration is set up In a vacuum filtration, the solution to be filtered is drawn through the filter paper by applying a vacuum to a filter flask with a side arm adaptor also known as a Buchner flask. Vacuum filtration is a fast and efficient way of filtering. The suspended solid is collected by swirling the mixture of the solid and liquid, then pouring it quickly into the filtration apparatus. This typically comprises a Büchner funnel fitted with the appropriate size filter paper, a clamped filter flask with conical filter adapter, and a vacuum applied to the side arm of the filter flask. Filter paper that has been used for this filtration process is GF/Whatman filter paper. This filter paper is made up of fibre glass and has a diameter of 47mm. The filter paper is suitable in filtering fine suspended sediment. After filtration, the filter paper is dried in an oven for 45 minute on the temperature of 103 Degree Celsius. Concentration of suspended sediment is obtained by calculating the difference between initial weight before filtration and final weight after filtration and drying in the oven. The difference is then being divided by the number of volume of water sample. Table 1 shows the concentration of suspended sediment for all the sample points after filtration.

Table 1 Concentration of suspended sediment of sample point Sample Point 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

3.3

Concentration of Suspended Sediment (mg/l) 55.1 55 53.8 53.6 52.2 52 52 50.3 47.7 47.2 46.5 46.3 46.1 40.9 40.7 38.2 38 36.5 36.2 36.1

Sample Point 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40

Concentration of Suspended Sediment (mg/l) 33.8 33.7 30.2 30.1 30 30 27.7 26.5 23.3 21 21 20.7 38.7 33.5 30.1 27.2 26.1 25.4 30.1 30

Sample Point 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Concentration of Suspended Sediment (mg/l) 29.9 29.3 29.2 27.6 27.3 26.8 26.4 26.1 26 24.8 24.6 23.4 23.3 21.7 23.4 23.3 21 19.9 19.7 19.6

Data processing

Atmospheric correction for satellite image is an essential process. Atmospheric correction retrieves the surface reflectance, which characterizes the surface properties from remotely sensed imagery by removing the atmospheric effects. Atmospheric correction has been shown to significantly improve the accuracy of image classification. Removing the influence of the atmosphere is a critical pre-processing step. To compensate for atmospheric effects, properties such as the amount of water vapour, distribution of aerosols, and scene visibility must be known. The images acquired by Earth observation systems cannot be transferred to maps directly, due to geometric distortions. These distortions are due to errors in the satellite’s positioning on its orbit, the fact that the Earth is turning on its axis as the image is being recorded, the effects of relief and others. They are amplified even more by the fact that some satellites take oblique images. Therefore, geometric correction has to be done at the satellite image. Some distortions, such as the effects of the Earth’s rotation and camera angles, are predictable. Thus it can be calculated and correction values will be applied systematically. Satellites also have sophisticated on-board systems to record very slight movements affecting the satellite. This information is used mainly to correct the satellite’s position, but it also can be used to correct the images geometrically. For Landsat 8 satellite image, geometric correction is not necessary as the satellite image is geometrically corrected. Masking is a process of removing digital number that not needed. Masking process allows to confine image processing to specific areas in the image or to ignore specific areas for processing (an inverse mask). Masking process involved in two steps which are building mask image and applying mask image (refer Figure 5). Building image masks can be done from specific data values including the data ignore value, ranges of values, finite or infinite values, ROIs, ENVI vector files (EVFs), and annotation files. Then the mask that has been build can be applied permanently to an image. Usually, Band 5 which are near infrared (NIR) is used in masking process as this band can differentiate between land and water.

Figure 5 Location image after masking process The value of reflectance is obtained after the process of atmospheric correction, geometric correction and masking is being done at the satellite image. Table 2 shows the value of reflectance with respect to the visible bands which are band 2 (0.45 - 0.51 µm), band 3 (0.53 - 0.59 µm), and band 4 (0.64 - 0.67 µm). Table 2 Value of reflectance of sample point Sample Point 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

Band 2 0.0878 0.0882 0.0874 0.0874 0.0866 0.0870 0.0874 0.0922 0.0844 0.0838 0.0859 0.0848 0.0863 0.0859 0.0909 0.0885 0.0865 0.0853 0.0863 0.0866 0.0842 0.0844 0.0865 0.0852 0.0830 0.0802 0.0800 0.0831 0.0811 0.0810

Reflectance Band 3 0.0696 0.0698 0.0687 0.0668 0.0672 0.0680 0.0684 0.0733 0.0654 0.0642 0.0663 0.0647 0.0664 0.0672 0.0694 0.0677 0.0658 0.0656 0.0660 0.0659 0.0633 0.0637 0.0653 0.0646 0.0626 0.0600 0.0597 0.0634 0.0604 0.0598

Band 4 0.0422 0.0397 0.0384 0.0382 0.0380 0.0382 0.0383 0.0455 0.0363 0.0362 0.0383 0.0366 0.0381 0.0395 0.0409 0.0397 0.0379 0.0368 0.0379 0.0381 0.0352 0.0361 0.0377 0.0369 0.0344 0.0323 0.0326 0.0354 0.0338 0.0335

Sample Point 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Band 2 0.0782 0.0785 0.0876 0.0849 0.0850 0.0843 0.0822 0.0806 0.0867 0.0855 0.0841 0.0837 0.0816 0.0815 0.0828 0.0812 0.0814 0.0815 0.0804 0.0797 0.0797 0.0825 0.0867 0.0839 0.0840 0.0844 0.0848 0.0841 0.0818 0.0809

Reflectance Band 3 0.0547 0.0551 0.0676 0.0645 0.0632 0.0614 0.0583 0.0579 0.0758 0.0644 0.0630 0.0634 0.0610 0.0610 0.0609 0.0591 0.0733 0.0582 0.0560 0.0535 0.0539 0.0617 0.0644 0.0646 0.0617 0.0644 0.0653 0.0639 0.0623 0.0613

Band 4 0.0310 0.0334 0.0393 0.0372 0.0362 0.0360 0.0345 0.0370 0.0450 0.0366 0.0353 0.0354 0.0338 0.0335 0.0343 0.0331 0.0463 0.0335 0.0320 0.0312 0.0315 0.0350 0.0367 0.0367 0.0355 0.0367 0.0369 0.0359 0.0341 0.0327

RESULT AND DISCUSSION The sample have been divide into two group which is 40 samples for dependant sampling whereas 20 samples has been used as independent sampling. The 40 dependant samples was used in the regression analysis in determining the best result of relationship between suspended sediment concentration and the value of reflectance for each band. Based on the regression analysis, suspended sediment concentration and band 2 shows a good correlation, by referring the value of R2 which is 0.54 as shown in Figure 6.

Suspended Sediment Concentration

70.0 60.0 50.0

mg/l

40.0

Poly. (mg/l)

30.0 20.0

y = 88306x2 - 12279x + 440.22 R² = 0.54

10.0 0.0 0.075

0.08

0.085

0.09

0.095

Reflectance of Band 2

Figure 6 Relationship between value of reflectance of band 2 and suspended sediment concentration The regression analysis result shows that band 2 has the best result compared to band 3 and band 4 of Landsat TM. Therefore, band 2 can be used in determining the concentration of suspended sediment in satellite image. Calculated suspended sediment concentration is computed using the independent sample point. Table 3 shows the value of suspended sediment concentration of image that has been applied with the best model. Table 3 Value of SSC Image and SSC Calculated Sample Point 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60

SSC Image 42.0073 39.5719 34.1078 34.2290 47.5128 36.7279 32.6222 36.6638 23.5130 25.6468 44.0336 32.5057 37.3748 31.4734 27.6807 25.9750 23.4285 32.5638 33.5070 24.8678

SSC Calculated 41.7228 40.6011 32.9157 34.2298 53.7561 35.5223 32.6225 35.2129 23.0935 25.0941 42.3656 32.7976 39.7092 31.3608 28.9657 26.0701 22.5635 31.7002 34.0483 24.8673 RMSE

Residuals 0.2845 -1.0291 1.1921 -0.0007 -6.2432 1.2056 -0.0003 1.4508 0.4195 0.5526 1.6679 -0.2919 -2.3343 0.1126 -1.2850 -0.0951 0.8650 0.8636 -0.5413 0.0004

Sq. Residuals 0.0809 1.0592 1.4212 5.8522E-07 38.9786 1.4535 9.4249E-08 2.1050 0.1759 0.3054 2.7819 0.0852 5.4492 0.0126 1.6513 0.0090 0.7482 0.7458 0.2930 2.4800E-07 1.693

Based on Table 3, the value of root mean square (RMSE) is calculated which is 1.693. It shows that band 2 of Landsat 8 is the most suitable band in determining the concentration of suspended sediment at Temenggor Lake. The model that develop is applied on the satellite image for mapping the concentration of suspended sediment of Temenggor Lake. The suspended sediment concentration is categorized to six class with the range of 10 mg/l. Figure 7 shows the map of suspended sediment concentration of Temenggor Lake. From on the map, it shows that Temenggor Lake have a low concentration of suspended sediment in most area of the river.

Figure 7 Suspended sediment concentration map of Temenggor Lake

CONCLUSION As conclusion, Landsat-8 TM is able to be applied to map the concentration of inland water suspended sediment at Hulu Perak River. From this study, band 2 is the most suitable in determining the concentration of suspended sediment at Temenggor Lake. The map of suspended sediment concentration provides information regarding the location and quantity of suspended sediment concentration in the study area. Besides that, it also proved that implementation of remote sensing technique able to map the concentration of suspended sediment for large coverage inland water area.

ACKNOWLEDGEMENT The author would like to thank the State of Perak and Perak State Park Corporation who has been directly involved and gives cooperation to ensure this study successfully. Deepest gratitude to Ministry of Higher Education and University Teknologi Malaysia for providing research fund under FRGS-VOT 4F336, GUP Tier 1 VOT 14H44 and GUP Tier 1 VOT 20H01 that makes this research well executed with good financial support.

REFERENCES Aggarwal, S., 2003. Principles of Remote Sensing. Satellite Remote Sensing and GIS Applications in Agricultural Meteorology, pp. 24-25. Doxaran, D., 2003. Remote-sensing reflectance of turbid sediment-dominated waters. Reduction of sediment type variations and changing illumination conditions effects by use of reflectance ratios. Paris: Pierre and Marie Curie University. Fleming, G., 1977. The Sediment Problem. Glasgow, Scotland: Department of Civil Engineering,The University of Strathclyde. Harrington,A. J. & Schiebe,R. F., 1999. Remote Sensing of Lake Chicot, Arkansas: Monitoring Suspended Sediments, Turbidity, and Secchi Depth with Landsat MSS Data. Remote Sens. Environ, 39, pp. 15-27. Janssen, L., & Huurneman, G., 2001. Electromagnetic spectrum. In Principles of Remote Sensing (Second ed) Netherlands: The International Institute for Aerospace Survey and Earth Sciences (ITC), pp. 58-60. Mobasheri, & M. Reza., 2003. Remote Sensing of Suspended Sediments in Surface Waters, Using MODIS Images. Tehran, Iran: Physics Department, KNT University of Technology. Nellis, M., A. Harrington, J., & Wu, J., 1998. Remote sensing of temporal and spatial variations in pool size, suspended sediment, turbidity, and Secchi depth in Tuttle Creek Reservoir, Kansas: 1993. Geomorphology 21, pp. 281-293. Pimstein, A., Bookman, R., & Tibor, G., 2014. Mapping the Spatial and Temporal Extent of Suspended Sediments Distribution in the Dead Sea using Satellite Remote Sensing Methods. Schaap, D. B., & Sando, K. S., 2002. Sediment Accumulation and Distribution in Lake Kampeska, Watertown, South Dakota. Water-Resources Investigations Report. Thomas, B. R., 1985. Measuring Suspended Sediment in Small Mountain Streams. California: Pacific Southwest Forest and Range Experiment Station, pp.1.