|
An Analysis for Detection of Bleaching in Shallow Coral Reef at Ishigaki Island using Multi Temporal Landsat TM Data
Ahmad Herodi, Yuasa Tetsuy School of Electronics Information Engineering, Yamagata University, 4-3-16 Jounan, Yonezawa-Shi, Yamagata, 992-8510 Japan Tel: +81-238-26-3323, Fax: +81-238-26-3323 E-mail: herodi@eatpost.yz.yamagata-u.ac.jp, yuasa@yz.yamagata-u.ac.jp Yamano Hiroyab National Institute for Environmental Studies, 16-2 Onogawa, Tsukuba-Shi, Ibaraki, 305-8506 Japan Tel: +81-29-850-2314 Email: hyamano@neis.go.jp Abstract The changes of the coral reef environments e.g. bleaching phenomenon is a big problem for the habitats living around the coral reef. During the last two decades, the coverage of health coral reefs have been degraded significantly, which mainly due to the natural anomaly such as global warming that caused the mass bleaching of coral reefs. To analyze and to conduct the assessment of the coral bleaching event, multi temporal Landsat TM image data was chosen with the consideration of Landsat satellite's ability to cover wide range of observed area and the viewpoint of price. The optical spectral signatures of coral reefs as they can be measured from the space have been reported in many papers. Thus, by using bleached coral's high reflectance in blue and green band spectra, the bleaching and the changes of coral reef have been examined. The work has taken place at shallow water-area of Ishigaki Island, Japan. The result shows that the severe and slight bleaching event due to the high sea surface temperature (SST) during summer 1998 can be detected and agreed well with in situ data taken from field investigation work. 1. Introduction Coral reefs are distributed widely in tropical and subtropical areas, and are fishery resources and tourism spots (Constanza et al., 1997), which have given many values for the environment. Recently, however, coral reefs have been significantly degraded and been destructed by both human disturbance activities (Hughes, 1994) and natural anomalies such as global warming (Hoegh-Guldberg, 1999) that caused high sea-surface temperature (SST). The change of SST monitored by NOAA satellite in August 1998 showing high temperature change at sea surface in Asia-Pacific regions gave an effect of coral reef bleaching. The mass bleaching, which caused a significant decreasing of health coral reef took place in 1997 and 1998 (Wilkinson, 2000). Bleaching occurs when coral losses the pigmented symbiotic algae, resulting in a bleached white appearance of living corals. The bleached coral results high reflectance in blue-green spectral bands of Landsat TM that is considered to enable for detecting the bleaching event. Holden and LeDrew (1998) and Clark et al. (2000) acquired high reflectance spectral resolution of coral and found that bleached coral is approximately 10% brighter in the visible spectrum than healthy coral, rising from about 10% reflectance to about 20%. Thus, bleach coral is approximately twice as brighter as unbleached coral. Since Landsat TM data is too coarse for mapping the coral in detail (Andréfouët et al., 2001) and coral reefs are more heterogeneous in variety, thus to assess the ability of Landsat TM data for detecting bleaching event should be limited to the bleached area with homogeneous coral and with wide range of bleaching spot (Yamano H., & Tamura M., 2004). Most of coral reef bleaching studies have assumed that it is difficult to detect coral bleaching using satellite sensor with low spatial resolution (~30 m) comparing to the scale of reef heterogeneity (Holden and LeDrew, 1998). Study of aerial photographs taken during the 1998 bleaching event shows that information on bleached corals can be obtained only for sensors with high (<2 m) spatial resolution (Andréfouët et al., 2002). This study is limited on using multi temporal Landsat TM images acquired from past 20 years (1984 ~ 2002) to obtain more dynamic change of coral reef. Moreover, the preference of Landsat TM was because its most effective sensors for mapping coral reefs in the viewpoint of price and the accuracy of mapping. With its observation spectral band e.i. blue and green bands that can penetrate water well for coral mapping and monitoring (Ahmad & Neil, 1994), the bleaching being analyzed. 2. Methods 2.1. Site Selection Hasegawa et al. (1999) showed that during 20 June and 10 August 1998, elevated in situ sea surface temperature associated bleaching event starting took place at Ishigaki Island, southwestern Japan (Fig.1).  Fig. 1. (A) Location of Ishigaki Island, in the Ryukyu Islands southwestern Japan. (B) Locations of reef of interest (Tamatorizaki and Urasaki reefs) are indicated. From 10 August to 10 September 1998, severe bleaching was observed. The proportion of healthy coral cover before summer 1998 exceeded 50% of the substrate (Environment Agency, 1996) based on survey taken from 1989 to 1992. The global-scale mass-bleaching event of 1997-1998 caused a significant loss of healthy coral in Japan's reefs (Kayanne et al., 2002). The status of coral reef before and after the bleaching event around Ishigaki Island was monitored at 72 sites by the Environment Agency of Japan (Environment Agency, 1999). After the 1998 bleaching, the status of corals at the same sites has been monitored continuously once a year (Environment Agency, 2000a, b). The selection of location for this study determined based on the state of bleaching level e.g. severely bleached and slightly bleached. The position of examined sites was determined by a GPS with an accuracy of ± 10 m. The existences and timing of bleaching prior to 1990 are unknown. In summer 2001, a coral bleaching event took place in the Ryukyu Islands due to high SST (Strong et al., 2002), with the proportion of dead coral was comparatively small. The ground truth data provided by the Environment Agency (1999) and Hasegawa et al. (1999) are used to validate the results of analysis using satellite imagery data (Table.1). 
As shown in Table 1, the number of sites of interest was selected to maximize the availability of clouds free images in order to have more clear changes of coral reef. 2.2 Satellite Images Selection In analysis work, we used Landsat TM images data (PATH 115, ROW 43) that acquired from past 20 years e.i. the images from 1984 to 2002 in summer season (Table.2). This is because the bleaching event is generally caused by high sea-surface temperature and strong incident light during summer season. The images data were geocoded in WGS-84 datum and geodetic projection with Latitude and Longitude coordinate type.
The work for normalized all the images should be based on four assumptions that: (1) the atmosphere is spatially uniform over a reef in a single image; (2) the incident light on the reef of interest is spatially uniform in an image; (3) the water-quality (extinction coefficient) over the reef of interest is the same in any two images; and (4) the reflectance of sand and deep water are constant in any two images (Yamano and Tamura, 2001). To normalize each DN values from three types of substrate i.e. shallow sand (DNs), deep water (DNd) and corals (DNc) are needed. These DNs should be obtained as close as possible to each other spatially to reduce the local change in atmospheric conditions and incident light due to the presence of clouds near the 3. Satellite Data Analysis The conversion of radiance L (in Wm-2 sr-1 µm-1) obtained from satellite sensor (Fig.2) to digital number (DNt) follows the form below: L = DNt / calibration coefficient (1) where the above equation can be changed to DNt = G[Lw + Lr + Le+ La] + B (2) Here, L = Lw + Lr + Le+ La; G and B are the gain and offset values for converting the radiance to DN. Lw is the water-leaving radiance from the substrate at the bottom of the water column and the radiance scattered by water column; Lr is the radiance reflected by water surface; Le is the radiance from adjacent points; La is the radiance scattered by atmosphere (called path radiance). The effect of multiple scattering was ignored. See Fig.2.  Fig. 2. Radiance components received by a satellite sensor Radiance from a substrate is described by: Lw = T.Ed.C.?b.exp[-2k.z] (3) where T is the light transmittance in the air, Ed is down welling irradiance above the sea surface, C is the factor that accounts for the loss of irradiance at the air-sea interface due to reflectance (Kirk 1994), ?b is the reflectance of bottom features, k is the extinction coefficient for water, and z is the water depth. Assuming the DN values from sand, deep water and coral (DNs, DNd and DNc) in two images , then based on equations (2) and (3), 
To remove the effects of the offset, path radiance and the influence of adjacent radiance, by subtracting DNc from DNs in the same image yielded: 
To conduct the normalization between two images, by introducing a coefficient a: 
the second image is normalized with respect to the first image. 4. Results and Discussion Fig. 3 and Fig. 4 show the results of the analysis of coral bleaching at five sites (Urasaki area are 3 sites and Tamatorizaki area are 2 sites respectively). The results indicate that bleaching was detected as a high reflectance in Landsat TM band 1 and 2, but not in band 3. It is because the band 1 and 2 has greater penetration of sunlight through the water column in these two spectral bands. In spite of the capability of Landsat TM image with middle spatial resolution (30 m) in detecting coral bleaching, it is limited to sites with large amount of coral coverage and severe bleached coral with up to 50% as shown in Table 1. Out of Urasaki Oki site (A1) as shown in Fig.3, all of four sites show a bleaching event took place one more time in summer season 2001. By comparison the radiance in this year event, the radiance of this site is clearly higher than those of previous and next year (Fig. 3 and Fig. 4). It considered that the result of analysis was suited well to the result taken from fieldwork in sites. While, in the Tamatorizaki sites where slight bleaching took place, the bleaching event still can be detected by comparing high reflectance of B1 and B2 in 1998 to those of the other years. However, in most cases, slightly bleached coral is difficult to detect (Yamano H.and Tamura M., 2004). Since the technology of earth observing satellite is on progress, it is important to improve the analysis using more advanced satellite sensor with high spatial resolution, because with 30 m in spatial resolution, the possibility of misinterpretation of the bleaching object would occur for the area with bleaching coverage under 23 % of the coral surface in a pixel (Yamano H.and Tamura M., 2004).  Fig. 3 Radiance change of multi temporal Landsat TM data bands 1-3 at severely bleached Urasaki sites.  Fig. 4 Radiance change of multi temporal Landsat TM data bands 1-3 at slightly bleached Tamatorizaki sites. Acknowledgement The author wish to acknowledge the support received from Dr. Tsuneo Matsunaga and Dr. Hiroya Yamano, Senior Researcher at Social and Environmental Systems Division, National Institute for Environmental Studies (NIES) Japan for contributing imagery data and sharing knowledge. References
Page 1 of 1
|