SHALLOW WATER MAPPING CORAL A CASE STUDY FROM THE GREAT JIL.IIrY. ............... J........ R.E. Reichelt and S.J. Bainbridge Australian Institute of Marine Science PMB 3, Townsvnle~ Queensland 4810. Australia Commission VII ABSTRACT A ground truth study at John Brewer Reef, Great Barrier (Australia), was carried out in order to test whether the already established geomorphological mapping of coral reef zones can be extended to the mapping of the living conlponents of the substrata. Present results indicate that the spectral and spatial resolution of Landsat MSS data does not pelmit the distinction of hard coral substrata from algal substrata, which are the two dominant classes of living cover. Shallow cora] reef habitats are very heterogeneous over SPOT XS was a wide range of spatial scales and therefore the higher spatial resolution evident in a classified image. A detailed analysis of feature and separability for SPOT is in progress. In addition~ the increased spectral resolution of Landsat TM data at the important blue end of the visible spectrum is presently under INTRODUCTION Cora] reefs are often remote and difficult to visit regularly for resource monitoting purposes. They are usuaHy large, sometimes many kilometres in length, which adds to the cost of sampling by field observation. The (OBR) of Australia is an archipelago of approximately 3000 individual reefs, extending over 2000km. In the last 25 years, many reefs in the central region of the GBR have been altered greatly by outbreaks of the Crown-of-Thorns starfish, Acanthaster planci, eats the live coral allowing subsequent growth of algae on the bare skeletons of the coral. Reefs that have been severely disturbed by starfish outbreaks have been observed to change from being coral dominated to algal dominated (Moran, 1986). Existing methods for monitoring the effects of this starfish all involve underwater sampling by divers which is very costly. Consequently remotely sensed data are being evaluated as a Olore cost-effective alternative. Landsat MSS imagery has already been established as an valuable tool for broadscale mapping of coral reef zones (Smith~ 1975; Jupp et al. 1985b) at the geomorphological scale. A reef cover model with classes such as ~reef fronf ~ ~outer reef fIat' and 'lagoon' was generated by Jupp et al. (1985b) using Landsat data and these classes matched closely the major geonl0rphological zones for coral reefs defined by Hopley (1982). In order to study effects of stalfish outbreaks (COT -CCEP Crown-of-Thorns Study, 1986) the monitoring method must permit the distinction of coral from algae within 3;ny particular zone and, for ecological monitoring in general, the nlapping of substrata at scales finer than that of the geonl0rphological zone is required. Substratum mapping in shallow coral reef areas using satellite inlagery is complicated by a number of environmental factors. One of the most inlpoliant is the confounding effect of varying water depth~ both by topographic variation and by tidal variation (Bina and Ombac, 1979; Jupp et al., 1985b). Corals and algae are the main living occupiers of space on coral reefs separability remotely sensed data is not well known. The brown colour is detelmined to a large extent pigments zooxanthenae~ symbiotic algae tissue of helmatypic ] 986). An observer in a small boat over a shallow coral reef with relatively conditions must still actually in order to assess whether the green-brown is caused by high algal cover rather than coral cover. Unfortunately spectral signatures of the groups benthic organisms not yet been characterised. by cover temporal change in the imagery of a reef, the first step must to assess the information content of the imagery at any single point in time. In terms of ground taIth, the main question in this paper is: Which, if any, of the biological or topographic attributes measured by ground observers serve to distinguish the various spectral classes detected in the image data? The substantial histotical archive of Landsat MSS data, and the potential to compare these historical data with the past records of starfish outbreaks has meant that the initial emphasis in this study is on the Landsat data. approximately region John Brewer Reef is a crescentic of the Great Barrier Reef near Townsville. It is 5km on longest axis and was severely disturbed by starfish outbreak in 1984-5 (Moran et at., 1985). One Landsat MSS image and one SPOT in1age of this reef were obtained through the Australian Centre for Remote Sensing for analysis in this study. The Landsat MSS (Landsat 5) image was Path 94, Row 73 in Universal Transverse Mercator Zone 55 with the scene centre at 18°49'S 147°37'E, taken at approximately 0930hrs local tin1e on 20th June 1985. The subset analysed here was 128 pixels x 100 Hnes starting at pixel 2054 line 609 of the image. Local tide height was calculated as 1.4m using a Queensland Department of Harbours and Marine Co-tidal chaI1 for the Townsville region (Drawing no. 01-155Sh2). The SPOT XS (multispectral) image (processed to level 1b) was from scene K,L = 373,387 with a subset of 512 pixels x 480 lines, at pixel 600 in line 2000. The scene was recorded at approximately 1030hrs local time on 24th July 1987. tide height was calculated as 1. 2m, Unfortunately this image was acquired in Dec 1987, after the field work had been completed and a detailed analysis of the spectral data is still in progress. After subsetting the images from computer-compatible tape on a VAX, all image analysis was done using a microBRIAN system (Jupp et aI., 1985a). Both images were contrast enhanced and then spatially rectified using 20 control points located on an orthophoto map of John Brewer Reef prepared by the Australian Survey Office (now caned the Surveying and Land Information Group, Canberra). During rectification the images were Locating a both resampled onto a common grid producing 5 pixels x 400 pixel that represents a pal1icular ground control point in an image of a reef is sometimes difficult because there are no regular features and most are not sharply defined. Using a supervised, nearest-neighbour "" ....'vv~,.u. (Jupp et al. ~ 1985a) the Landsat data were into 25 themes. The supervision of the classified into 17 themes. and the image (raw data minus mean values of each classification involved use a regions of the image that have radiance values class - Jupp and Mayo. 1982) to widely separated from the means the existing classes. These regions were then selected an ainled at reducing the infornlation as new seeds for content vV.!,'-'U'U-l classes were on by surveyors who (April 1987). The 2. File :AJBREW SensorzLands~t MSS Date I 28-86-85 Bands '4-S,S-G#6=R H~me Im~Qe 1. Location :John Brewer Reef Processlng ' Rectlfled Raw Im~ge showing sample site locations. Each of these ten sites were sampled by a SCUBA diver using 4 contiguous benthic Hne transects (Reichelt et . 1986) each 50m in length with substratum and topographic vadation being recorded at every metre. Topographic conlplexity along the transects was vertical relief. to the nearest 10cm~ fronl one metre to the next. recorded as fluctuation Details of the sanlpling techniques are given in Bainbridge and Reiche1t( 1988). The sites vaIied in depth from 1m to 3m. 505 000 File Na.e :ASBREW SensarlSPOr XS Image Date I 24-07-87 Bands 11~BJ2.G,3.R Location IJohn Brewer Reer Processlng ' Rectlfied Raw Image ] AINS/COTSAC image of John Brewer Reef showing structure of reef flat habitats sand channels. et al. (l or color. (MA) ems above the coral from those with to its greatly enhanced spatial . led to a more complex Landsat MSS were MA 1 13 5 3 0 0 2 0 3 6 21 1 1 6 7 1 2 3 I 4 5 II II 1 50 3 5 6 7 8 I 9 17 0 ]0 10 5 10 5 5 5 5 10 31 31 27 3] 60 35 36 I 34 23 3 300 1 100 40 100 60 60 30 16 17 21 20 12 12 18 best resulting odginal transects indication of the success function analysis; Variable a HI I HI 3 0 13 0 IV 1 1 0 0 8 0 >50 MA 1 n Discriminating Variables 2 correct 0 0 0 2 Zone 83 outer reef flat inner reef flat crest + crest + slope 81 100 50 >50 F = 8.2. = 15, o P< A"-U,"".H...,1 et t:l<'I'"'I-.<:>I1" (1986) with (1988) with re>nor81 One of the first problems encountered in this study was the time consuming nature the sampling procedure which resulted only 10 ground control points being surveyed benthic cover. This is a relatively sman number of points even though the total transect was 200m, which covers 3 to 4 for Landsat l\ISS to 10 pixels for has to out """.....-".-r. V ......LIl ....' .... Another "IV,",'IJ,",' to crest and slope, areas are ecologically MSS. '.'''-Hl.",nT did separate biotic abiotic geomorphological zones also. and so .. ""'''''''''-,,'VA ......... .....,. the raw abundance of 2 has not been the subject of . . . ""' ..........'..... . ........ j''''''Jl •. ...., ..... 1 for imagery data is not terrestrial areas SPOT mapping, Forster et al., 1987). The lineaments evident in the outer reef flat zone are not detected in LandsatMSS data because of the difference spatial resolution. are sand channels between the strips of hard substrata which may ( 1982) .Il ('",,,,.. ,,. ... 01,,. 1 press. Dixon, W.J. and Brown, M.B. (Editors) 1979. BMDP-79 Bionledical computer P-series. University of California Press, Los Angeles, 880pp. Forster, B.C., Ttinder, J.C Can"olL ,FaITington. . Kwoh, L., Nicholson, Holden. G.F.H. 1987. Mapping research using SPOT digital data. Australasian Remote Sensing Conference, Adelaide, September 1987. 490-499. q Hill, G.J.E. and Kel1y. G.D. 1987. Landsat TM versus MSS for land cover southern Queensland. Septenlber 1987. Volume I: I Hopley, 1982. York,453pp. New geomorphology of the Jupp, D. ,Heggen, S ,Mayo, B.A. 1985a. The BRIAN handbook ............... Natural Resource Series 3, 51 pp. A .......... Jupp, D.L.B. and Mayo. K.K. 1982. The use of residual images analysis. Photogram. Eng. Rem. Sensing 48: 595-604. Landsat image • Kenchington, R.A. and Jupp, D.L.B., Mayo, K.K., Kuchler, D.A., Claasen. D.van Guerin, 1985b. Remote sensing for planning and managing the Great Reef of Australia. Photogrammetria 40: 2 I -42. 1986. Coral reef and Menor, Kuchler, D.A, Maguire, C" McKenna, A., PtiesL ITC 1 survey method for verification of LandsatMSS image 217-223. Moran, P.J., Bradbury, R.H. and Reichelt, R.E. 1985. Mesoscale studies of-Thoms/Coral interaction: A case history from the Great Bartier Int. Coral Reef Symp., Tahiti 5: 321-326. Reichelt, R.E. 1988. Space and stnlcture effects on coral Symp., Townsville. press. Reichelt, R.E .. Loya, Y. and Bradbury, R.B. benthic communities on two coral reefs 73-79. Ipp. Richards. J.A. 1986. Remote sensing digital analysis. Smith, V.E. 1975. Thematic 'Y"Q1'1Ih-r~.,.n.'" Institute Science 1Oth lnt. Symposium on Remote Sensing of 10pp. Veron, J.E.N. 1986. Corals of Australia and Australia, 644pp. Indo-Pacific. Angus