USDOC/NOAA/NESDIS NOAA/CSC ICLARM

Resolution for Action (Final Document)

Agenda
Evaluation Results (downloadable file)
Participants in Attendance (downloadable file)
Presentations
Pictures from the Workshop

Menu of Abstracts

Al Strong "Coral Reef Watch: NOAA's Early Warning System For Coral Reef Health"

Christopher Brown "Application of Satellite Ocean Color Imagery To Coral Reef Mapping & Monitoring"

Serge Andrefouet "Using SPOT-LANDSAT images for mapping, inventory and monitoring of reefs"

Heather Holden "Remote Sensing Of Coral Reefs"

John Klein III "Exploration of ASIA Hyperspectral Remote Sensed Data for Coral Reef Mapping"

Jim Hendee "A Prototype Expert System for Comparing In Situ Data with Satellite-Derived Data in Near Real-Time"
(*See footnote)

Jack Hardy "Coral Reef Monitoring with Airborne LIDAR"

Jack Hardy "Fluorscence Transect" (Excel file)

Rod Low, Jonathan C. Gradie, Kevin T. C. Jim "Marine Remote Sensing and GIS in Hawaii and the Pacific"

T.J. Done, J.K. Oliver, R. Berkelmans, J. Lough, W. Skirving "Can satellites help to identify local scale variability in coral bleaching and coral death? Goals and approaches of an international collaboration"

Coral Reef Watch: NOAA's Early Warning System For Coral Reef Health

A. E. Strong
NOAA/NESDIS/ORA/ORAD -- E/RA3
NOAA Science Center -- RM 711W
5200 Auth Road
Camp Springs, MD 20746-4304
Alan.E.Strong@noaa.gov

Like the rest of the world, most of the U.S. coral reef systems are threatened due to pollution, over-fishing, and thermally induced bleaching. Affected reefs include those of Florida and Puerto Rico, nearly half of Hawaii's, and an unknown but significant percentage of reefs in U.S. Pacific Territories. The geographically scattered and often remote locations of many coral reefs preclude routine or scheduled monitoring. In 1998, NESDIS, using NOAA (AVHRR) satellites, established an experimental capability to conduct thermal bleaching surveillance of coral reefs on a world-wide basis. This experiment demonstrated remarkably successful results for early warning of coral reef bleaching due to accumulated thermal stress over all tropical oceans, resulting in this Workshop.

A Coral Reef Watch Program is proposed by NOAA to (1) transition existing experimental satellite reef monitoring capabilities into a viable operational capability, (2) formalize the existing U.S. leadership in the emerging global "Virtual Coral Reef Ecosystem Monitoring Laboratory," (3) provide for a solid scientific basis for future monitoring, assessment products/capabilities, and understanding of the role of the tropics in global climate change, and (4) build upon existing NOAA international coral reef information systems for quick and efficient transfer of near real-time information and data products. This initiative, as applied to coral reefs, complements the Integrated Ocean Observing System. NOAA has established itself as a world leader in monitoring the oceanographic conditions potentially affecting reef health using satellite reconnaissance techniques. Coral Reef Watch solidifies that position as the world leader in operational environmental monitoring and early warnings and represents NOAA's commitment to U.S. tropical territories, who by their distance from the mainland, frequently are under-represented by NOAA's activities.

Last year, 1998, was the warmest year since temperature recordings started some 150 years ago. Similarly, the 1990s have been the warmest decade. In addition, 1998 saw the strongest El Nino ever recorded, at least for this century. Consequently, very high water temperatures were observed in many parts of the oceans, particularly in the tropical Indian Ocean, often with temperatures peaking 3 to 5 degrees C above maximum expected summertime values. Many Indian Ocean corals bleached and subsequently died, mainly due to the high water temperatures, but possibly in combination with other meteorological and climatic factors. Bleaching is a stress syndrome by which hard corals expel their symbiotic algae, and the corals appear white or transparent. Bleached corals may survive for weeks or months, but will die if conditions do not return to normal or the stress is prolonged and/or severe.

Massive mortality was recorded on reefs of Sri Lanka, Maldives, India, Kenya, Tanzania, and Seychelles, approaching 90% in many shallow areas. Reefs in other parts of the Indian Ocean, especially at depths below 20 m, experienced up to 50% mortality (Wilkinson, C., Linden, O., Cesar, H., Hodgson, G., Rubens, J. and Strong, A. E., 1999). Hence coral death during 1998 was unprecedented in its severity. The cultural and economic effects for coastal communities of the Indian Ocean are likely to be long lasting. In addition to potential decreases in fish stocks and negative effects on tourism, erosion may become an even more acute problem, particularly in the Maldives and Seychelles. If the observed global trend in rising ocean temperatures continues, there is an increased probability of a recurrence of the catastrophic bleaching observed in all areas of the tropical oceans during coming years. Coral reefs of the Indian Ocean have provided an important paradigm for the potential and devastating effects of continued global climate warming.

Since 1990, largely through the use of satellite data, we have been able to "predict" virtually all known large scale coral reef bleaching events in real time using the bleaching "HotSpot" criterion (Goreau and Hayes, 1994). We identify regions of the tropical ocean where satellite-derived sea surface temperatures (SST) exceed a threshold of 1.0 degrees Celsius above the highest expected summertime mean temperature. Maps showing monthly distributions of HotSpots are available on the world wide web at:

http://psbsgi1.nesdis.noaa.gov:8080/PSB/EPS/SST/climohot.html. Based on this technique satellite hindcasts have been produced for all years since 1982: http://http://manati.wwb.noaa.gov/orad/al/hot_anual82_97.html.

The global distribution, duration, and extent of HotSpots from January to December 1998 shows that ocean temperatures sufficient to produce coral bleaching have exceeded all previous levels since 1982, eclipsing the previous high observed in 1988.

Based upon the very strong correlation between the distribution and duration of oceanic HotSpots and the thermal thresholds for coral reef bleaching during the warmest season of the year, we have developed the prototype of an early alert system for coral bleaching. We envision a coral bleaching "warning" to indicate that SST anomalies over a specific reef tract are sufficient for coral bleaching to develop. Similarly, a coral bleaching "watch" would indicate that SST anomalies approaching the thermal threshold for bleaching are developing in a coastal area and that monitoring for coral reef bleaching should begin. These proposed NOAA watches/warnings will be communicated to potential observers at sites expected to show coral reef bleaching as predicted by the distribution and degree of thermal anomalies.

References:
Goreau, T.J., Hayes, R.L. ,1994. Coral bleaching and "ocean hot spots." AMBIO 23:176-180.

Wilkinson, C., Linden, O., Cesar, H., Hodgson, G., Rubens, J. and Strong, A. E., 1999. Ecological and socioeconomic impacts of 1998 coral mortality in the Indian Ocean: An ENSO impact and a warning of future change? AMBIO, 28(2), 188-196.

top of page

Application of Satellite Ocean Color Imagery To Coral Reef Mapping & Monitoring

Christopher W. Brown
NOAA/NESDIS
Office of Research & Application
Camp Springs, MD

Satellite imagery offers the potential to routinely observe large geographic regions appropriate for the detection and monitoring of events over an extended period of time. Satellite ocean color imagery, in particular, has proven useful in estimating the type and concentration of radiatively active constituents within the water column and assessing attributes of the bottom in shallow regions.

Several geophysical parameters derived from ocean color observations are of relevance in mapping and monitoring coral reefs. These parameters include estimates of water-leaving radiances (at various wavelengths), chlorophyll concentration, total suspended solids concentrations, and diffuse attenuation coefficients, and the detection and identification of algal blooms. Chlorophyll concentrations may be used to estimate regional phytoplankton biomass and primary productivity, assess regional eutrophication, and predict anoxia. The diffuse attenuation coefficient, i.e. the rate at which light attenuates with depth in the water column, is useful for estimating light availability at depth and in detecting and tracking turbidity plumes. The ability to remote characterize the taxonomic composition of an algal blooms may be employed to detect the invasion and expansion of macro algal onto coral reefs, as well as identify benign and deleterious phytoplankton blooms.

Data from existing and planned satellite ocean color sensors will be of use in various aspects of mapping and monitoring coral reefs. There are, however, limitations. The limitations and applications of ocean color observations will be determined by 1) the ability to accurately estimate the relevant geophysical parameters from the radiances acquired at the sensor, and 2) the spatial and temporal resolution of the acquired imagery and that required to map and monitor coral reefs. The former factor is beyond the scope of this note, and I concentrate here on the latter factor.

In general, the ocean color sensors provide global, multispectral visible - near infrared observations every 2-3 days at a spatial resolution of 1 km. Several sensors planned over the next few years will acquire finer spatial and spectral resolution imagery. The use of these data is dependent upon the overlap between the spatial and temporal characteristics of the available imagery and of the scale of the problem.

The spatial resolution and temporal frequency deemed necessary to address various aspects of the mapping and monitoring of coral reefs varies considerably in space and time (Table 1). The example, daily to weekly observations at mesoscales are necessary to monitor environmental parameters pertinent to coral reef health, such as the extent and duration of turbidity plume over a coral reef. On the other hand, detailed mapping and cover discrimination requires much finer spatial resolution and less frequently. The various applications with the temporal and spatial scales that overlap those available of existing and planned ocean color imagery are colored red.

Table 1. Spatial and temporal resolution necessary to map and monitor coral reefs. Text colored red indicates the applications that can be addressed by existing and planned ocean color observations.

Given present technologies and the stated criteria (Table 1), it is concluded that satellite ocean color imagery can be used to 1) detect undocumented coral reefs in remote regions; and 2) routinely monitor environmental parameters relevant to coral reef stress and health, such as turbidity. Satellite ocean color imagery will, however, be unlikely to 1) furnish detailed (< 10 m resolution) maps, and 2) directly assess coral reef health.

New technologies may furnish additional information useful for coral reef mapping and monitoring. For example, the hyperspectral Coastal Imaging Spectroradiometer (COIS), which will possess an IFOV of 30 or 60 meters, will improve the spectral and spatial resolution of existing ocean color imagery, and consequently extend the potential use of ocean color data for more detailed mapping and monitoring. The proposed Special Events Imager (SEI), which would acquire ocean color imagery with a spatial resolution of 300 m from a 300 km x 300 km region as frequently as every 15 minutes, could measure the spectral reflectance of a reef throughout the tidal cycle and be used to remove the influence of the overlying water column.

top of page

Using SPOT and LANDSAT images for mapping, inventory and monitoring of reefs

Serge Andréfouët, Frank Müller-Karger, Chuanmin Hu

Dept. Marine Science - University of South Florida
Remote Sensing Laboratory / Biological Oceanography Laboratory

The presentation aims to make a reminder of the characteristics of the multispectral images provided by the satellites LANDSAT and SPOT and how they have been used for the assessments of coral reefs environments.

We emphasize the intrinsic limitations of the spectral and spatial resolution of these images. The limitations strongly restrict the potentials of these images for reef studies. In favorable conditions (clear waters, clear sky, good geometry of the scene) the bottom remains visible at 25 meters with XS1 (green band) and 40 meters with TM1 (blue band). However, the reconnaissance of the different bottom classes or habitats using multispectal methods is limited to few meters (7-8 m for SPOT, 12-15 m for LANDSAT). The definition of the habitats is coarse and can be related to only a small set of environmental descriptors. The minimum discernable unit is an area of 60x60 meters for SPOT-HRV images, 90x90 meters for LANDSAT-TM. The coarse habitats can be mapped with a good accuracy (~70%) in shallow waters if a correction of the influence of the water column is performed. Effective and objective change detection assessments depend on the quality of the bathymetric and atmospheric corrections. Indeed, without accurate corrections, the very similar spectral signatures of the living communities such as corals and macro-algae prevent to detect subtle or vigorous changes in reef communities.

Despite these limitations, the products derived from LANDSAT and SPOT images can be very informative if they are used objectively within the right limits. At the scale of an entire reef system, maps of coarse habitats are accurate in shallow waters. They provide a synoptic view over an entire reef system useful to understand the spatial structure of the reef, to stratify in situ investigation and select new monitoring sites. For the characterization of the water column, the synoptic vision provided by the images helps to characterize the boundaries of the reef systems. Coupled with adequate in situ measurements, images inform of the exchanges between the reefs and their connected systems (ocean, land and other reefs) and allow, for example, estimates of residence time in atoll lagoons.

At the scale of the biosphere, the recent launch of LANDSAT 7 and the Long Term Acquisition Plan managed by NASA offers the perspective to get a complete coverage of reef areas worldwide. The images acquired for LTAP can be used at different levels. Without any ground-truth and corrections, a simple analysis of the images can separate soft bottom and hard-bottom, thus offer the perspective of a global inventory of reef worldwide, but without any precise benthic features. To study benthic features with higher precision, ground-truthing is required and only a limited number of areas can be investigated. Therefore, it seems necessary to take advantage of the increasing number of sites surveyed by the monitoring organizations (ReefCheck, CARICOMP, CRAMP, GCRMN…) for a fruitful exchange. The definition of a common protocol, defined as a "remote sensing layer", will allow the exchange of consistent data useful for both image interpretation and in situ monitoring.

top of page

Remote Sensing Of Coral Reefs

Heather Holden
National University of Singapore

As global pressures on coral reefs and related ecosystems grow as a result of developing economies and increasing coastal populations, the need for careful monitoring, planning and management becomes essential (Knight et al., 1997). According to Bryant et el. (1998), 58% of the world's coral reefs are threatened by human activity such as coastal development, destructive fishing, overexploitation, marine pollution, runoff from deforestation and toxic discharge from industrial and agricultural chemicals. SPOT and Landsat TM satellite sensors provide synoptic medium spatial and low spectral resolution imagery that covers large geographic areas consistently and repetitively. This could aid global estimates of coral reef health if proper corrections and identification procedures are established. We have been unable, however, to discriminate between various benthic habitats using in situ spectra with equivalent spectral characteristics to TM and SPOT due in part to the broad wavelength bands.

The prospect of hyperspectral satellite-borne sensors is exciting since it will allow large geographic areas to be covered with high spectral resolution at potentially lower cost than higher spectral resolution airborne sensors. The spatial resolution will be inferior to that possible with airborne remote sensing, but sub pixel analysis techniques, such as spectral mixture analysis, may alleviate associated problems. The U.S. Navy and Space Technology Development Corp. have entered a partnership to launch NEMO (Naval Earth Map Operation) in 1999 with high spectral (210 bands with 10nm bandwidth) and medium spatial (30m) resolution. Similarly, ARIES (Australian Resource Information and Environment Satellite) is planned for launch in 2000 with hyperspectral (approximately 200 bands with 20nm bandwidth) capabilities and 30m spatial resolution.

When using airborne data, there is a compromise between spatial and spectral resolution such that data storage issues dictate that a user can specify either hyperspectral or hyperspatial resolution, but rarely is it feasible to have both. A user might have to accept 10m spatial resolution in exchange for 100 narrow and contiguous spectral bands, or conversely, accept only 20 broader and non-contiguous spectral bands in exchange for 1m spatial resolution. In the future, these limitations may be lessened with cheaper and readily available data storage. The implications of this data storage limitation include a tradeoff between the ability to identify small features with hyperspatial resolution and the ability to discriminate different features with hyperspectral resolution

Before accurate remote sensing of submerged coral reefs is a feasible proposition, however, two fundamental issues need to be addressed. First, the water column significantly affects the remotely sensed signal through wavelength-specific augmentation and attenuation, which does not always follow Beer's Law of logarithmic extinction of energy. Radiative transfer models sensitive to varying substrate brightness, water depth, and water quality are needed to account for the variable effects of the water column. Secondly, several common coral reef features have optically similar hyperspectral reflectance characteristics, so the subtle differences need to be identified in order to enable remote discrimination.

We acknowledge that the overall health or vitality of a coral reef ecosystem can be evaluated using various measures such as growth rate, reproduction rate, or number of fish present. The only remotely detectable means of reef evaluation, however, is through the presence of bleached or algae-covered coral because of the change in visible spectral reflectance in water penetrating wavelengths (Holden and LeDrew, 1998a). The transition from a healthy coral to a stressed or unhealthy bleached coral involves the loss of photosynthetic algae (zooxanthellae) or loss of pigments from within the zooxanthellae (Muller-Parker and D'Elia, 1997). We therefore use the presence of bleached coral and algae-covered dead coral as a remotely detectable means of evaluating the overall health or vitality of a coral reef ecosystem (Holden and LeDrew, 1998b).

We believe that it is important to focus on fundamental issues such as spectral distinction between optically similar substrates using in situ data rather than using remotely sensed imagery immediately for several reasons. First, it is difficult to perform an assessment of the accuracy of a classification of a remotely sensed image due to the large geographic area covered and the difficulties involved with geopositional accuracy. Second, the spatial resolution or pixel size available from satellite imagery is large (i.e. 30m for Landsat TM), which presents problems of mixing of reflectance signatures within each pixel such that it is difficult to determine the "pure" reflectance characteristics of only healthy coral or only sea grass. If high spatial resolution imagery is captured using an airborne sensor, then the pixel size could be smaller than 1m, but there can still be several substrate types present within this small area in a typical coral reef environment. Finally, remote sensing is espoused as an ideal tool for resource management and ecosystem monitoring, but this is before fundamental research is complete in the areas of water column correction and substrate identification, so its capabilities may have been oversold thus disappointing potential users (Green et al., 1998).

Data collection has been ongoing each year since 1996 in Fiji, Indonesia and the U.S. Virgin Islands to create a large global database of reflectance measurements of coral reef features in an effort to establish a replicable and objective method of remotely monitoring changes in coral reef degradation and recovery. A hyperspectral radiometer (Analytical Spectral Devices Personal Spectrometer II) is used to measure reflectance of submerged features while scuba diving. Underwater photographs are taken of each of the features measured to add to the spectral and photographic database or library. Additionally, notes are taken describing the depth, feature type, surrounding substrate, water quality, feature size and morphology as well as any other pertinent information. There are over 700 spectra available in our library where each spectral measurement has 205 contiguous waveband channels with 1.4nm bandwidths.

Both Pearson correlation and cluster analysis reveal that when the entire spectral curve is considered, there is a strong correlation between and within populations of substrate types. Principal components analysis is used as a data reduction tool to identify one spectral reflectance curve for each of the broadly defined populations: healthy coral, bleached coral, algae-covered surfaces, sand and sea grass, as in Figure 1. These representative, or training, spectra are used to devise a classification procedure that would allow identification of populations based on the slope and change in slope of the spectral reflectance curves in narrow wavelength regions. The first derivatives between 605 and 625nm as well as between 585 and 605nm enabled identification of sand substrates and bleached corals, respectively. The second derivative between 506 and 566nm enabled discrimination of healthy corals. Finally, the maximum magnitude of reflectance was used as a means of discriminating algae or rubble surfaces from bleached branching coral. This 4-step procedure using first and second derivatives and magnitude of reflectance was tested on the entire data set in order to test its accuracy. The overall accuracy of the classification of in situ spectra is over 80%, which is appropriate for many applications such as coral reef ecosystem management on a regional scale. This analysis, however, does not consider problematic issues such as spectral mixing within a given pixel and water column correction of the remotely sensed imagery.

Furthermore, there appear to be subtle spectral differences in reflectance in narrow wavelength ranges that may allow discrimination of coral species (for example, see Figure 2). In general, healthy coral spectra display a reflectance minimum at 670nm with a steeply decreasing slope beginning at 650nm. Additionally, most of the healthy coral spectra have peaks in reflectance at 575 and 605nm. Porites sp. spectra have a broad, flat reflectance maximum between 610 and 650nm, and Acropora latistella spectra show a sharp increase to a peak at 518nm. Acropora palmata spectra have a steep slope between 540 and 570nm while Porites porites spectra reveal a gradual slope in this same wavelength range. Finally, Sinularia species #1 spectra display a steep slope to a reflectance peak at 482nm.

Whether airborne or spaceborne hyperspectral sensors are used, the research presented here represents the fundamental first steps in enabling remote discrimination of submerged coral reef ecosystem features. We have shown that in narrow wavelength regions, discrimination is indeed possible for broadly defined populations and that there does not appear to be a benefit to capturing contiguous hyperspectral spectra (Holden and LeDrew, in press). Therefore, a user interested in creating a thematic map to identify change over time in algae cover, for example, could limit data collection to fewer spectral bands and increase the spatial resolution. Without deconvoluting the hyperspectral reflectance response of optically similar features, spectral confusion results in low classification accuracy on a remotely sensed image. The spectral library presented may provide the opportunity for confident classification of submerged coral reef ecosystems thus reducing the cost of the monitoring project. It may also aid scientists and managers in selection of optimal band location and bandwidth characteristics for their airborne remote sensing missions.

Acknowledgments
Data collection in Fiji was supported by the National Geographic Society and the International Development and Research Council. The National Oceanic and Atmospheric Administration, National Ocean Service Center supported data collection in the US Virgin Islands for Coastal Monitoring and Assessment. This research was also supported by an NSERC operating grant to Dr. Ellsworth LeDrew and an NSERC graduate scholarship to Heather Holden. We owe great thanks to the field research assistance of Chris Derksen, Bill Kalbfleisch, Candace Newman, and Marnie Laing, Dr. Richard Murphy, and the support staff at Buck Island National Monument.

References
Bryant, D., L. Burke, J. McManus and M. Spalding. 1998. Reefs at Risk: a map-based indication of threats to the world's coral reefs. World Resources Institute.

Holden, H. and E. LeDrew. 1998a. The scientific issues surrounding remote detection of submerged coral ecosystems. Progress in Physical Geography. 22 (2), 190-221.

Holden, H. and E. LeDrew. 1998b. Spectral discrimination of healthy and non-healthy corals based on cluster analysis, principal components analysis and derivative spectroscopy. Remote Sensing of Environment. 65, 217-224.

Holden, H. and E. LeDrew. In Press. Hyperspectral identification of coral reef features. International Journal of Remote Sensing.

Knight, D., E. LeDrew and H. Holden. 1997. Mapping submerged corals in Fiji from remote sensing and in situ measurements: applications for integrated coastal zone management. Oceans and Coastal Management. 34 (2), 153-170.

Muller-Parker, G. and C. D'Elia. 1997. Interactions between corals and their symbiotic algae. In Life and Death of Coral Reefs. Ed. C. Birkeland. New York: Chapman & Hall.

Green, E., C. Clark, P. Mumby, A. Edwards, and A. Ellis. 1998. Remote sensing techniques for mangrove mapping. International Journal of Remote Sensing. 19 (5), 935-956.

Figure 1. Principal components analysis was used as a data reduction tool to identify representative or training spectra from the in situ data set.

Figure 2. Greater precision may be possible if spectral differences between types of corals can be found.

top of page

Exploration of ASIA Hyperspectral Remote Sensed Data for Coral Reef Mapping

By John Klein III.

PE NOAA/NOS/NCCMA
Silver Spring MD, 20910
301.7132.3000x160
john.klein@noaa.gov

Abstract

A recent mission was undertaken over the waters of the Caribbean to examine the effort required to examine and map the coral reef habitat around the United States Virgin Islands and Puerto Rico. The majority of the effort took place over the Island of St Croix of the USVI during the months of February and March. The Island is over 20 Miles long and 5 wide and required 14 flight lines for nominal coverage. Our mission was flown at an altitude of 10,000ft providing a ground spatial resolution 10ft (3m).

Utilizing the AISA hyperspectral imaging (HSI) device on board the NOAA Citation provided for a special opportunity to collect both photogrammetry and HSI simultaneously. Although the true potential of this data acquisition combination has yet to be fully exploited, there has been an opportunity to examine the benefits and drawbacks of the AISA data set. This paper presents a preliminary look at those issues surrounding a small portion of the data collected over St Croix and the Virgin Islands.

The area investigated concentrates over Buck Island National Park, which is located 2 miles northeast of Christanstead in St Croix. This island and its surrounding waters is the first National Park designated for a marine habitat. Once vibrant and teeming with abundant reef fish and invertebrate communities, the area has recently been under stress from a variety of sources, including numerous hurricanes in the past several years, assorted water quality threats including eutrophication, thermal anomalies, increased turbidity and possible hydrocarbon contamination. St Croix is an excellent test case area due to its distribution of both fringing and platform reef morphologies and the existence of both protected and unprotected reef habitat. The island has a robust economy consisting of approximately 50,000 residents and several large industrial refineries and manufacturing facilities. And, although not as popular as its neighbor islands St John and St Thomas, St Croix receives nearly 6,000 tourists from cruise ships per week.

All the original raw data was first processed using the proprietary software, Caligeo developed by the AISA manufacturer. Once a readable image was produce, subsequent data processing was executed in ENVI from Research Systems Inc. The results step through a typical processing sequence to determine the viability of the data set. First the overall image was used to identify 7 dominant target areas reflective of different coral reef environments. From these areas of the image, spectral information is gathered to provide input for the subsequent processing. Both supervised and unsupervised classifications techniques were employed. A k-means classifier was used which employs statistical characteristics of the spectral data values found in each pixel and groups them accordingly. A supervised methodology known as the spectral Angle Mapper was used to match the shape of the spectral profile in each pixel to those of the 7 target areas and ranks the pixel as to the certainty of the match. The last technique known as Minimum Noise Fraction is essentially performed by subsequent PCA analyses identifying and separating signal from noise with as many principal components as there are spectral bands. Of all the techniques, this one has the most notable distinction between bottom habitats. These results strongly suggest that a viable and useful map of the bottom types and habitats can be achieved through the use of HSI.

This examination of processing of the HSI revealed both the potential and problems associated with this technology. Experments such as this never proceed as expected and this effort was no different. The data show an exciting potential for mapping bottom types, extent and character of coral communities. Processing of the data, although computationally intense, takes significantly less time than the steps involved in digitizing and georectifying photogrammetry and has the added benefit of containing true radiance/reflectance data.

Now that this phase has been completed, subsequent efforts utilizing this technology will benefit from the lessons learned during this deployment and should be able to thoroughly exploit its true potential.

top of page

Coral Reef Monitoring with Airborne LIDAR

Jack Hardy
Center for Environmental Science
Western Washington University
Bellingham, WA 98225-9181
jhardy@cc.wwu.edu

Introduction
Airborne LIDAR (light detection and ranging) utilizes a laser light source to probe the characteristics of a surface target. Generally, a pulsed laser emission is directed downward from a low flying plane or helicopter platform. Information about the target is derived from back-scattered reflectance or fluorescence of the target. Both the variety of applications and the usefulness of LIDAR in environmental assessment and monitoring have grown rapidly over the past two decades. Airborne lasers have been successfully employed for terrain mapping and bathymetry, oil spill mapping, terrestrial vegetation mapping, assessment of stress in forest communities and quantification of phytoplankton chlorophyll biomass.

Marine benthic communities including seaweed, seagrass and coral contain photosynthetic chlorophyll pigments. When excited by shorter (blue or green) wavelength light, chlorophyll emits light at longer (red) wavelengths, i.e. it fluoresces. We previously demonstrated that coral fluorescence can be induced remotely by laser excitation and that the fluorescence decreases as coral bleach, i.e. lose their chlorophyll and/or zooxanthellae symbionts (Hardy et al. 1992).

Methods
We tested the feasibility of using airborne LIDAR to assess the distribution and health of coral on a Hawaiian fringing reef. The study site was located off the northwest coast of the Island of Hawaii between Waima Point and Puako. The site was chosen because it has a well-developed healthy coral cover and had been monitored and studied previously. We selected a sampling transect line 1.8 km long. At eight pre-selected stations along the transect, we measured coral cover by 10 m long line intercept transects using SCUBA. Depths along most of the transect were 8 to 22 m, but station 5 was <2 m. At 1 m intervals along the line intercepts coral samples were collected, frozen, and later analyzed for pigment content (Cottone 1995).

The NASA P-3 research aircraft flew three passes over the 1.8-km-long transect line (marked with buoys), at an altitude of 60 m and a speed of 123 m sec-1. Fluorescence was actively induced in the underlying substrate using pulsed lasers (frequency-doubled Nd:YAG lasers) which emitted 355- and 532-nm light every 10 nanoseconds. Both the upwelling fluorescence, and passive (solar) reflected light between 370 and 719 nm, were measured with a 32 channel radiometer (Hoge and Swift 1983). The pulsed excitation resulted in a fluorescence emission spectrum sample approximately every 5 m along the transect. The "footprint" was approximately 0.3 m in diameter.

Results
Bottom cover was virtually solid coral dominated by Porites compressa and Porties lobata. Mean pigment densities (per cm-2 of coral surface) were: 8.6, 5.0, and 0.29 ug cm-2, respectively for chlorophyll-a, peridinin, and B-carotene and were not significantly different between stations. Chl-a density showed a small but significant difference between species (P. compressa 7.8 ug cm-2 and P. lobata 10 ug cm-2).

Airborne-measured fluorescence spectra along the entire transect contained a dominant peak in the known chlorophyll region, i.e. about 680-700 nm. Additional peaks occurred at 483, 505, 527, and 630-640 nm. Laboratory and in situ field data suggest that peaks identical to or close to these characterize many species of coral (Mazel 1995). Reflected light spectra were typical of those previously measured in situ and exhibited a minimum (i.e. absorption peak) at 670 nm. Due to water column attenuation, the strength of the fluorescence emission signals were inversely related to bottom depth along the transect. After correction by a wavelength-specific water column attenuation coefficient, spectral peaks were much more prominent.

Conclusions
In situ-measured fluorescence and reflectance spectra differ between coral, algal encrusted reef, seagrass, sand and other substrates, and in the case of coral, change with pigment loss (bleaching) (Mazel 1995, Myers et al. 1999, Schmidt et al. 1999). Preliminary data suggest that specific changes in spectral fluorescence and reflectance can be related to changes in pigment density. Further quantification of these relationships will allow construction of appropriate algorithms for remote sensing. Despite the relatively high initial costs of data acquisition and processing hardware, the ability to quickly assess areas of tens to hundreds of square km should make LIDAR a cost-effective tool for monitoring coral reefs.

Literature Cited
Cottone, M. 1995. Distribution of coral reef pigmentation as measured by HPLC. Huxley College of Environmental Studies, Western Washington University. 95 pp.

Hardy JT, FE Hoge, JK Yungel, RE Dodge. 1992. Remote detection of coral bleaching using pulsed-laser fluorescence spectroscopy. Mar Ecol Prog Ser 88: 247-255.

Hoge FE. 1986. Active-passive correlation spectroscopy: a new technique for identifying ocean color algorithm spectral regions. Appl Optics Vol 25(15): 2571-2583.

Mazel CH 1995. Spectral measurements of fluorescence emission in Caribbean cnidarians. Mar Ecol Prog Ser 120:185-191

Myers, M., J.T. Hardy, C. Mazel and P. Dustan. 1999. Optical spectra and pigmentation of Caribbean reef corals and macroalgae. In Press. Coral Reefs 18(2).

Schmidt, C, T. Steely, J. Hardy, S. Strom, G. Muller-Parker, M. Bynagle, E. Macri, A. Miller, E. Peterson, and T. Rivers. 1999. Changes in optical spectra and pigmentation of the coral Montastraea faveolata in response to elevated temperature and ultraviolet-b radiation. International Conference on Scientific Aspects of Coral Reef Assessment, Monitoring, and Restoration. National Coral Reef Institute. Ft. Lauderdale Florida, 14-16 April.

top of page

Fluorscence Transect

Jack Hardy
Center for Environmental Science
Western Washington University
Bellingham, WA 98225-9181
jhardy@cc.wwu.edu

top of page

Marine Remote Sensing and GIS in Hawaii and the Pacific

More than 80% of the coral reef under the jurisdiction of the United States is found in Hawaii and the widely scattered islands of the Pacific Basin. The Northwest Hawaiian Islands, scattered in a chain several thousand miles in length, contain 50% to 60% of the reefs that are to be managed. Each of the remote and isolated sites may contain hundreds to thousands of sq. km. of reef systems which must be mapped in both location and distribution, as well as assessed in both geophysical and ecological content and status. Identifying and understanding the dynamic processes such as coral decline or expansion, marine life habitat alteration or man-caused damage requires timely revisits to each site. Clearly, the remoteness and isolation of these sites present significant logistical problems for traditional reef assessment and monitoring techniques: any visit to these sites requires special transportation considerations and supply preparations, both of which become expensive and time consuming. Management of these resources is clearly a significant challenge.

We propose that a multi-tiered approach based upon marine remote sensing can be cost-effective in providing a significant portion of the geophysical and ecological data required for a GIS-based coral reef mapping, characterization, and monitoring program. This tiered approach makes use of satellite, airborne and underwater remote sensing and GIS technologies and techniques developed, or being developed, in Hawaii and elsewhere. This strategy represents a balance between area covered and information content using different sensors from satellite, airborne and underwater remote sensing techniques. The data collection techniques can be categorized with different scales ranging from regional (thousands of sq. km.) to in situ (diver or ROV) data collection. Each scale has a range of cost-effective combinations of spatial resolution, area covered and information content that will translate into GIS-useable layers.

At the regional scale are the planned commercial 1-meter panchromatic and 4-meter multi-spectral satellite-based imaging systems. For example, the Ikonos-2, to be launched in December of 1999, will have 1 meter per pixel panchromatic imagery at 10-bit dynamic range and 4 meter multi-spectral capability at 8-bit dynamic range with real-time pointing of +/- 45 degrees along and cross track. With the appropriate use of pointing and sun angle, the panchromatic and bands 1 through 3 of the multi-spectral images can be used as a reconnaissance tool for coral reef mapping and monitoring in the remote and less accessible areas. Reconnaissance can mean separating light coral sand and sand channels from darker coral, algae and rock. A typical Ikonos "scene" is 10 km x 10 km but can be as small as 3 km x 3 km or as large as desired. Although this may not provide the most complete picture of coral health, it will certainly be a cost-effective way to alert resource managers of the need for a detailed look with more diagnostic remote sensing or in situ tools. We present examples of 1-meter panchromatic and 4-meter multi-spectral imagery of Hawaiian reefs areas taken with the Ikonos airborne simulator.

A significant advantage of using 1-meter class satellite imagery is the ability to provide essentially orthoscopic imagery that is easily geo-referenced to the meter scale. Standard satellite pointing is sufficient to provide geo-referencing to about 10 meters. The presence of a single identifiable scene-surveyed point (ground control point) reduces the error to that of the survey point or pixel size, whichever is largest. Because of the large area that can be imaged at once with these satellites, it is possible to geo-reference thousands of sq. km to this level of accuracy. This regional scale geo-referenced base imagery is fundamental for reducing the geo-referencing costs for all other remote sensing data. It serves as a context for all other GIS-related information layers. In fact, most of the other data collection techniques can then be geo-referenced with the satellite-acquired, geo-referenced map.

At the "local" scale, airborne photogrammetric (film and digital) and large-format, high resolution (0.1 to 1 meter/pixel), multi-spectral imagery provides an effective method for mapping areas in the 100 sq. km. range. Digital multi-spectral systems (such as the TS-1, a 4-band, 12-bit designed and manufactured in Hawai'i ) can be used with narrow band filters specific to coral reef monitoring. Some of these systems lend themselves to easy geo-referencing and data analysis. By cross-referencing known points to the satellite-acquired geo-referenced map, ortho-rectified data can be geo-referenced at least to the 1-m scale. The metrics for matching the narrow band multi-spectral filter sets with diagnostic metrics are being developed at the University of Hawai'i in conjunction with a number of Hawai'i -based remote sensing companies and state agencies. These techniques were used for coral mapping in Kaneohe Bay and Kailua Bay on Oahu and alien marine species mapping and monitoring along the coast of Maui. A new underwater, large area, mapping technique using this airborne multi-spectral technology for 0.01-meter to 0.1 meter resolution for depths >35 meters is currently under development.

Airborne digital multi-spectral mapping was used to produce the 1-meter/pixel geo-referenced base map of Kailua Bay (Oahu), which serves as the reference for all of the other science investigations in the area. Four narrow bands, specifically chosen for water penetration (> 35 m), material discrimination (coral, sand, algae, substrates, etc.), and bathymetry. This base map provides a unique view into the geological and morphological character of the region and serves as a reference for other specialized studies. As a result of this mapping effort, previously unknown coral structures and areas were discovered in an area considered to be well studied in comparison to most of the reefs of the Pacific. The narrow-band imagery, combined with detailed in situ spectral measurements of coral can yield a classified map of coral reef health over large areas.

The mapping products derived from high spatial resolution satellite and airborne multi-spectral imaging products could direct the application of specialized (and sometimes more costly) remote sensing techniques with much larger information contents (e.g., hyper-spectral, lidar, sonar imaging, fluorescence imaging, etc.). These specialized techniques expand the area over which results from ground truth sites can be applied. When combined into a comprehensive GIS, the relationship of various physical measurements can be compared from the smallest to the largest scales. The result is a tiered application of remote sensing techniques, which provide a cost-effective means for mapping, assessing and monitoring coral reefs in the Hawaiian Islands and the Pacific Basin.

top of page

Can satellites help to identify local scale variability in coral bleaching and coral death? Goals and approaches of an international collaboration

Countries and industries strongly reliant on coral reefs may suffer economically and socially if projections of more frequent and widespread death and bleaching of corals unfold as predicted during the 21st Century. In an attempt to assess the level of risk, Australia has developed national collaborations (between Australian Institute of Marine Science (AIMS); Great Barrier Reef Marine Park Authority (GBRMPA); CSIRO Atmospheric Research) and intentional collaborations (AIMS, GBRMPA and NOAA).

In Australia, AIMS and CSIRO are commencing a study to address these concerns by critically analyzing likely climate changes and coral responses for the GBR region. The study will bring together climate models, biological data on coral stress thresholds, and local scale environmental monitoring data. The climatic models have been developed by CSIRO Atmospheric Research, and information on biological thresholds and local variability in sea-surface temperatures is being compiled by AIMS and GBRMPA in collaboration with NOAA. The goal is to be able to say whether superimposition of a global-warming signal on spatial variability and natural annual and decadal variability will have any discernable effect on the frequency of bleaching events, and which areas if any within the GBR are likely to be most vulnerable.

The collaboration with NOAA is critical in addressing the issues of spatial variability in the risk of bleaching. NOAA's 'hot spots' have been sufficiently effective as broad scale predictors of the location of coral bleaching to warrant more detailed investigation of their ability to capture the local scale (5 - 50 km) variability of bleaching, which can be considerable. The Great Barrier Reef (GBR) and NW Australia are ideal locations for such an investigation, because of the availability of instrumental records of SST and other key parameters, and field observations on coral bleaching in the context of long-term and broad scale reef monitoring programs. A shore-hugging finger of a Pacific hot-spot sloshed into the southern half of the GBR in early 1998 and a large Indian Ocean hot-spot sat on top of northern offshore reefs off NW Australia. In both regions, in-situ records of SST and coral verified both the satellite-derived SST anomaly and the presence of bleaching and death of corals.

These locations also provide potentially important contrasts in geographic settings. The offshore NW Australian Reefs are oceanic, and their oceanographic settings and processes have been under investigation by AIMS for several years. The GBR setting is a complex jigsaw superimposed on a coastal-oceanic gradient. The GBR comprises 90% 'open water' occupied by ~3000 coral reefs that account for the other 10%. The reefs are distributed over 15 degrees of latitude (~2000 km) and sit on a shallow (10 m grading to 200 m) continental shelf up to 200 km wide. In terms of factors other than SST implicated in coral bleaching, the region is extremely variable in space and time: i.e. in the degree to which it is influenced by freshwater runoff from the land, by cool, enriched water upwelled from the deep ocean, and by water clarity, the latter being a function of coastal resusupension by currents and wind-waves, of sediments from rivers, and of biological productivity. Given this variability in the other factors implicated in bleaching, one might not expect too much local-scale predictive capacity for the 'hot spots', being based on just one predictor (SST anomaly). To the contrary, preliminary mapping of 'hot spots' on maps showing bleached and unbleached locations suggests a remarkable north-south and cross-shelf fidelity between the two maps.

The collaboration will investigate the utility of increased spatial resolution available from satellites and of improved SST algorithms. Division of responsibilities are currently being finalized, but its overall scope is as follows. AIMS has extensive archives of Local Area Coverage from AVHRR, and is also currently working on a "tropical" algorithm for retrieval of SST. Data from an automated radiometer system attached to a large tourist-carrying vessel will be used for ground-truthing the SST algorithms. The collaboration will produce improved spatial resolution of climatologies for key potential bleaching factors, viz. for temperature, rainfall, river flows. It will evaluate the performance of the LAC and the alternative climatologies and SST algorithms against the performance of the 1998 anomalies in predicting the location of bleaching on the GBR. NOAA will also generate SST maps that factor in results of a new GBRMPA study on the critical thresholds of temperature and duration that induce bleaching and death in indicator coral species in selected locations. Known rates and trajectories of recovery of disturbed coral communities will be used to guide projections of likely future status of bleached coral reefs at a variety of scales, based on various data and/or assumptions on the availability of recolonising larvae and/or survivors.

top of page