Classification of algal bloom types
From remote sensing reflectance
Soo Chin Liew, Leong Keong Kwoh, and Hock Lim
Centre for Remote Imaging, Sensing and Processing
National University of Singapore
Blk. SOC1 Level 2, Lower Kent Ridge Road, Singapore 119260
Tel: (65) 8745069 Fax: (65) 7757717
E-mail: liew_soo_chin@nus.edu.sg
SINGAPORE
Keywords : Ocean color, reflectance, algal blooms, phytoplanktons, classification
Abstract
A technique for classification of phytoplankton bloom types from remote sensing reflectance is described in this paper. Several minor algal bloom events were sighted and their characteristics reflectance signatures were collected using a handheld spectrometer during a series of sea-truth water sampling campaigns carried out from Dec 1996 to Dec 1999 in coastal waters around Singapore. Reflectance spectra of two additional algal bloom classes were also collected during two field trips to the Manila Bay. In order to assess the potential of utilizing satellite ocean color sensors for algal bloom detection and classification, reflectance data for the SeaWiFS and future MERIS sensor spectral bands were simulated from the in-situ radiance data. An algorithm based on the singular value decomposition (SVD) technique was then applied for classification of algal bloom types from the simulated satellite sensor reflectance data. The results show that all the eight algal bloom classes can be distinguished from the clear sea water reference sample. The average accuracy of classification using this technique for all the classes are 98.6% (for MERIS) and 96.6% (for SeaWiFS), in comparison to 87.5% (MERIS) and 73.8% (SeaWiFS) if the reflectance values are used.
1. Introduction
Phytoplantons (or algae) constitute the base of the marine food web. However, algal blooms may cause harm by shading other aquatic life, depleting the dissolved oxygen content, and causing paralytic or diarrhetic shellfish poisoning (Richardson 1997). It is important to monitor occurrences of algal blooms due to their strong social, economic and health impacts. Satellite remote sensing measurement of ocean color provides a tool complementary to in-situ sea-truth measurements for algal bloom monitoring. As the individual phytoplankton pigments are characterized by their unique light absorption features, this property allows detection and identification of algal blooms by ocean color remote sensing technique (Cullen et al. 1997, Kahru and Mitchell 1998, Sathyendranath et al. 1994),
Currently, the SeaWiFS sensor on board the Orbview 2 satellite (launched October 1997) provides ocean color data with about 1-km resolution. It has six bands in the visible region and two in the near-infrared region. Each band has a 20-nm bandwidth. The recently launched MODIS sensor on board the NASA's Terra satellite has eight ocean color bands in the visible spectral region, with 10 nm bandwidth. The other recent sensor was the OCTS onboard the Japan's ADEOS satellite launched in August 1996. ADEOS ceased operation in June 1997 when the ADEOS satellite stops its operation. It was the first second-generation ocean color sensor after the 10 years gap since NASA's CZCS. Future planned ocean color sensors include the NASDA's ADEOS2-GLI and ESA's ENVISAT-MERIS. These future sensors have more wavelength bands, all with about 10 nm bandwidths. With the availability of these ocean color satellites, it is foreseeable that satellite ocean color data will play an increasingly important role in the monitoring of algal blooms. Algorithms for algal bloom detection and for classification of algal bloom types will be required.
In this paper, we describe a method of detecting and classifying algal bloom types based on the singular value decomposition technique. Satellite remote sensing reflectance signatures for several types of algal blooms were simulated using in-situ reflectance data measured during a series of water sampling campaigns around Singapore and in Manila Bay. The satellite sensors tested are SeaWiFS and MERIS.
2. Algal Bloom Classes
Sea-truth water sampling campaigns were carried out from Dec 1996 to Dec 1999 in coastal waters around Singapore (Lin et al. 1999, Liew et al. 2000). In-situ reflectance spectra from sea water surface were acquired using a portable spectroradiometer. Several minor algal bloom events were sighted and their characteristic reflectance signatures were collected during this period. The classes of algal blooms observed include: Trichodesmium (a type of cyanobacteria); Chain forming diatoms; Chochlodinium (naked dinoflagellate); Dinoflagellates predominantly Dinophysis caudata; Diatoms (Rhizolenia Sp.); and mixture of chain forming diatoms (Skeletonema type) with some armoured dinoflagellates. Two additional algal bloom classes were collected during two field trips to the Manila Bay. One trip was carried out during the algal bloom episode (mainly Ceratium and Pyrodinium Bahamense) in Aug 1998, and the other in March 2000. Although there was no report of red tide during the later trip, results of water sampling in Manila Bay indicated that there were signs of increasing phytoplankton counts. A class of sea water reference spectra for sea water samples with low chlorophyll and low suspended solids was also collected during the regular water sampling field trips.
Altogether eight algal bloom classes and one reference sea water class are used in the analysis. The nine classes of spectra are tabulated in Table 1.
Table 1: The nine classes of reflectance spectra from algal blooms and reference sea water
| Class |
Description |
| 1 |
Clear sea water reference (Singapore) |
| 2 |
Trichodesmium (Singapore) |
| 3 |
Chain forming diatoms (Singapore) |
| 4 |
Cochlodinium (Singapore) |
| 5 |
Ceratium and Pyrodinium Bahamense (Manila Bay) |
| 6 |
Dinoflagellates (mainly Dinophysis caudata)(Singapore) |
| 7 |
Diatoms (Rhizolenia Sp.)(Singapore) |
| 8 |
Chain forming diatoms (Skeletonema) with some armoured
dinoflagellates(Singapore) |
| 9 |
Protoperidinium and Ceratum (Manila Bay) |
