Determination of shallow substrate from satellite remote sensing data with bio-optic based algorithm

Abstract

Remote sensing techniques have been widely used for extraction of coastal information including the sea surface, in the water and within beneath of shallow substrates. The coastal water where the substrates are found is classified into ‘case 1’ and ‘case 2’. They are differs based on the water constituents such as phytoplankton, suspended particulate matter (SPM) and coloured dissolved organic materials (CDOM). The interactions of incoming radiance and water-leaving irradiance within the water in both these coastal types have been formulated in the bio-optic algorithm. The bio-optical algorithm has been previously used in mapping ocean-colour, mapping total suspended matter (TSM) and deriving water properties. In this study, the applicability of bio-optical algorithm was examined and analysed over coastal water for detection and mapping of shallow substrates using satellite remote sensing data. Two satellite data sets examined are: (i) the fine resolution Worldview-2, and (ii) medium resolution Landsat-8 OLI, with 0.5m and 30m spatial resolution respectively. The test sites were conducted in Pulau Tinggi and Pulau Merambong, Johor that representing the coastal type 1 and II as well as the shallow (less than 20m) and deep areas (less than 40m). In-situ samples consisted of seagrass, seaweed, coral, mud, sand and ancillary information on water depth were divided into two independent mutual sets and used as input to the algorithm and the respective validations. The results indicated that shallow substrates could be extracted at 91.6 percents overall accuracy with 0.55 of kappa coefficient (k), hence showing good agreement at Pulau Merambong. However, at Pulau Tinggi, the overall accuracy of substrates derived at 52.17 percent (k = 0.33) and 42.22 percent (k=0.26) for Worldview-2 and Landsat OLI, respectively. It is therefore concluded that the bio-optical algorithm has been identified as restricted on deeper water even on the clearer water (type 1) with less TSM. Hence, the potential of bio-optical algorithm for mapping shallow (less than 20m) substrates within Malaysian coastal water is very high with the improvement of water-leaving radiance from deep water model.