Mapping and Prediction of Coral Bleaching Using GIS Corals on coral reefs bleach when they lose their complement of intracellular symbiotic algae known as zooxanthellae. Coral color varies widely, but most if not all corals are pigmented primarily by the olive-brown colored zooxanthellae within their tissues. Thus, the loss of these symbionts is readily apparent and the coral will often appear white, or bleached. Because most corals in the tropics are entirely dependent on the photosynthetic energy provided by zooxanthellae, bleaching has dire consequences for the health of corals and, consequently, reef ecosystems. While recovery can occur, coral mortality is usually the result of intense and prolonged bleaching. Bleaching can be triggered by a variety of stressors, but the most pervasive and important cause of mass worldwide bleaching is heat stress. Widespread coral bleaching events were rare before the 1970s, but they now occur frequently and more severely, often coinciding with El Niño conditions. Climatic warming and the subsequent increase in bleaching frequency and intensity suggests that widespread bleaching will occur annually within the next half-century (Hoegh-Guldberg 1999). There are several reasons why GIS has particular promise for coral bleaching science. (1) Bleached corals reflect more light than non-bleached corals and other bottom types and thus give off a strong spectral signature that can be picked up by remote sensors. (2) Tropical coral reefs tend to occur in shallow, clear waters where remote sensing will be most capable of discerning benthic features. (3) Coral bleaching events often occur over wide geographic areas or even worldwide. Remote sensing coupled with GIS allows for detection, mapping, and analysis at such large spatial scales. (4) GIS can allow for integration of a multitude of geospatial data that either affects (sea surface temperature, currents, river discharge) or can be affected by (fishing areas, tourism areas, downstream reefs that receive larvae) coral bleaching. (5) GIS can help predict bleaching sensitive areas, and it can also determine what areas are resistant to bleaching and what might contribute to their resistance (Wooldridge and Done 2004). This will be crucial to the development of reserves that can protect areas that don t typically bleach from other potential stressors (West and Salm 2003). The current coral bleaching literature suggests that GIS use in the field is only in its infancy and is largely dependent on the scale and accuracy of currently available remote sensing technology. The literature is relatively bursting with assessments of bleaching detection limits and pixel resolution requirements using various remote sensors, and use of GIS has apparently been hampered by the lack of remotely sensed data with appropriate spatial resolution. While bleaching may occur over large areas, it tends to be patchy in nature, due to both variation in coral cover and variation between coral species and habitats. Satellites sensors such as Landsat TM with 30 meter pixel resolution will simply not pick up some of this patchiness (Yamano and Tamura 2004), which leads to underestimates of bleaching. Furthermore, remote detection of bleaching depends entirely on how white a coral is, so if corals are only partially bleached and other pigments are still present, remote sensing may underestimate bleaching (Berkelmans and Oliver 1999, Andrefouet et al. 2002). IKONOS images, with 4 or 1 meter pixel resolution, are certainly more accurate, and under the ideal conditions of high coral cover and extreme bleaching it may be possible to discern bleaching with high accuracy (Elvidge et al. 2004). Andrefouet et al. (2002) suggest that the optimum spatial resolution to detect bleaching is in the range of 40-80 cm if processing time and spatial resolution are considered together. At this level, it is possible to discern bleaching in individual coral heads. These recommendations are not surprising given that speciesspecific differences in bleaching susceptibility will often drive bleaching variability at sub-
meter scales. It is likely that future bleaching events will be mapped seamlessly into a GIS at or near such scales using satellite remote sensing. There are a handful of studies that have incorporated point-type aerial survey data into a GIS for large-scale analyses of bleaching (Berkelmans and Oliver 1999, Berkelmans et al. 2004, Wooldridge and Done 2004). All of these studies were performed on the Great Barrier Reef (GBR); widespread bleaching events in 1998 and 2002 were monitored worldwide, but Australia appears to be the leader in communicating and monitoring bleaching events as they happen. The punctuated nature of bleaching events necessitates a quick response from the scientific community, and the proximity of many coral reef scientists to the GBR no doubt allows such a response. Comparing the 1998 and 2002 events over nearly the entire GBR allowed for a robust conclusion that 2002 surpassed the 1998 event as the worst bleaching event on record (Berkelmans et al. 2004). Without aerial surveys and GIS, it would have been difficult to come to such a strong conclusion given the high degree of spatial variability observed. For example, in many cases, reefs that bleached in 1998 did not bleach in 2002, and vice-versa. Spatial analysis was also able to discern a trend of higher bleaching intensity closer to shore that was consistent over both bleaching events (Berkelmans and Oliver 1999, Berkelmans et al. 2004). Once these spatial patterns in bleaching have been mapped, spatial patterns in environmental variables as well as habitat variables can aid in determining factors that correlate with and perhaps contribute to coral bleaching (Wooldridge and Done 2004). In particular, the use of AVHRR satellite sea surface temperature (SST) data has proven to be indispensable for understanding spatial variation in coral bleaching. Using independent methods with AVHRR data, Wooldridge and Done (2004) and Berkelmans et al. (2004) came to similar conclusions that spatial variation in SST could predict observed bleaching with a high degree of accuracy (73%, Berkelmans et al. 2004). In fact, these sorts of large-scale bleaching assessments in which high-resolution SST data are used demonstrate the unequivocal role of elevated temperatures in the bleaching phenomenon. Incorporating other information such as proximity to cooling continental shelf waters and community type only improves these models (Wooldridge and Done 2004). The only good thing that will come from rising coral bleaching and degradation is better science. As each future coral bleaching event occurs, there will be new and improved methods and geospatial data to characterize it. Satellite remote sensing is very close to being able to capture bleaching with acceptable accuracy, and this will allow for much more resolved worldwide censuses of bleaching. For the GBR, it appears that there is already sufficient data to allow for implementation of protected areas. Though an increased volume of remotely sensed data will surely be welcomed for use in coral reef GIS, it seems there will always be a need for ground truthing and field surveys to determine such variables as habitat type, species composition, and even symbiont composition, to fully understand bleaching variability. Literature Cited Andrefouet, S, R Berkelmans, L Odriozola, T Done, J Oliver, and F Muller-Karger. 2002. Choosing the appropriate spatial resolution for monitoring coral bleaching events using remote sensing. Coral Reefs 21:147-154 Berkelmans, R, and JK Oliver. 1999. Large-scale bleaching of corals on the Great Barrier Reef. Coral Reefs 18:55-60
Berkelmans, R, G De ath, S Kininmonth, WJ Skirving. 2004. A comparison of the 1998 and 2002 coral bleaching events on the Great Barrier Reef: spatial correlation, patterns, and predictions. Coral Reefs 23:74-83 Elvidge, CD, JB Dietz, R Berkelmans, S Andrefouet, W Skirving, AE Strong, B Tuttle. 2004. Satellite observation of Keppel Islands (GBR) 2002 coral bleaching using IKONOS data. Coral Reefs 23:123-132 Hoegh-Guldberg, O. 1999. Climate change, coral bleaching and the future of the world's coral reefs. Marine & Freshwater Research 50:839-866 West, JM, and RV Salm. 2003. Resistance and resilience to coral bleaching: implications for coral reef conservation and management. Conservation Biology 17(4):956-967 Wooldridge, S, and T Done. 2004. Learning to predict large-scale coral bleaching from past events: A Bayesian approach using remotely sensed data, in-situ data, and environmental proxies. Coral Reefs 23:96-108 Yamano, H, and M Tamura. 2004. Detection limits of coral reef bleaching by satellite remote sensing: Simulation and data analysis. Remote Sensing of Environment 90:86-103 Annotated Bibliography Andrefouet, S, R Berkelmans, L Odriozola, T Done, J Oliver, and F Muller-Karger. 2002. Choosing the appropriate spatial resolution for monitoring coral bleaching events using remote sensing. Coral Reefs 21:147-154 Aerial photographs were taken of various reefs on the GBR during the 1998 bleaching event and resampled at spatial resolutions ranging from 10 cm to 5 m to determine the resolution which most accurately captured the bleaching. Results were compared to bleaching estimates ground truthed from transect surveys. Images were binarized (converted to grayscale) so that bleached colonies would be more apparent. Bleaching detection decreased rapidly from 10 cm and 1 m resolution; detection changed less between 1 and 5 m. Corals that were completely bleached (totally white) and in shallow water had a much higher detection probability than those that were partially bleached and/or in deeper water. Also, reefs with higher coral cover exhibited less variance in bleaching estimates. Therefore, detection of bleaching was very sensitive to spatial resolution, coral cover, depth, and bleaching severity. The authors suggest a resolution between 40 and 80 cm for the best compromise between resolution and data processing effort. This is just below the resolution of the best satellite sensors to date. Berkelmans, R, and JK Oliver. 1999. Large-scale bleaching of corals on the Great Barrier Reef. Coral Reefs 18:55-60 Approximately 23% of reefs on the GBR were aerially assessed for bleaching during the 1998 bleaching event and a subset of these were ground-truthed by divers. A striking pattern of higher bleaching on the inshore reefs compared to the offshore reefs was observed, with 87% of inshore reefs bleached and only 28% of offshore reefs bleached. Moreover, the inshore reefs that were bleached showed more intense bleaching (i.e. higher prevalence and intensity) than the offshore reefs. The point data was incorporated into a GIS map of the GBR and shows this pattern well. The ground truthed data suggests that the aerial surveys underestimated bleaching in most cases, particularly in areas with low levels of bleaching. Elevated temperature (1-2C) is singled out as the primary cause of bleaching, though conditions may have been exacerbated by
high solar radiation and by lower salinity inshore. Higher bleaching in the central and southern GBR relative to the northern GBR is consistent with temperature anomalies in the two regions. This study was timely and set the stage for things to come. Berkelmans, R, G De ath, S Kininmonth, WJ Skirving. 2004. A comparison of the 1998 and 2002 coral bleaching events on the Great Barrier Reef: spatial correlation, patterns, and predictions. Coral Reefs 23:74-83 This paper is a follow-up to the 1998 study and covers considerably more ground. Aerial surveys were performed in a nearly identical manner to the 98 study to allow for comparison. A major advance is the use of AVHRR temperature data for the 2002 bleaching event. SST is superimposed in a GIS over both 1998 and 2002 bleaching data, categorized in six classes of bleached coral cover. By testing the predictive ability of several different temperature proxies (e.g. days above threshold, degree-days, etc.) they found that the maximum temperature over any 3-day period during the bleaching event had the highest predictive accuracy of 73%. Thus, they conclude that it is short periods of extreme temperatures that are most detrimental to corals. Predicted bleaching for both 1998 and 2002 was spatially interpolated using nearest-neighbor analysis and classified as low, medium or high. 2002 proved to be the worst coral bleaching event on record. Spatial patterns of bleaching were largely consistent between years, and the 2002 event also showed higher incidence of bleaching inshore, though the trend was a bit weaker. Elvidge, CD, JB Dietz, R Berkelmans, S Andrefouet, W Skirving, AE Strong, B Tuttle. 2004. Satellite observation of Keppel Islands (GBR) 2002 coral bleaching using IKONOS data. Coral Reefs 23:123-132 Before-and-after 4 m resolution, 4-band IKONOS images of the 2002 bleaching event were processed for the Keppel Islands, GBR. The radiance of the image taken during the bleaching event was normalized to the one taken previously, and bleaching was detected using the blue and green bands. Bleaching was detected as deep as 15 m. Cloud cover was unfortunately heavy for the image that was commissioned for the bleaching event, but bleaching was nonetheless detected. This was helped by the fact that the area had high coral cover and extensive bleaching. The authors suggest that IKONOS imagery, though expensive, is an option that should be further explored for coral bleaching studies. They also suggest that IKONOS images of coral bleaching may allow very detailed (down to the level of individual coral colonies) coral reef mapping due to the high reflectance of bleached corals. However, their method does require before and after images. Wooldridge, S, and T Done. 2004. Learning to predict large-scale coral bleaching from past events: A Bayesian approach using remotely sensed data, in-situ data, and environmental proxies. Coral Reefs 23:96-108 A number of different environmental and ecological variables were used to model the potential dependency of bleaching and coral mortality on them. This paper is probably the heaviest on GIS of all. They first used GIS to select sites to survey during the 2002 bleaching even on the GBR, and quantified bleaching frequency, mortality, and community type at each. This site data was then incorporated into a GIS along with average summertime SST (used as a proxy for acclimatization temperature ), tidal mixing, and heat stress (which is the same 3-day function used by Berkelmans et al.
2004). They present several different scenarios, but it appears that all of these variables contribute legitimate predictive value to the model. Each has some sort of dependency on the another, but the tidal mixing function and the heat stress function are probably the most important, just as in Berkelmans et al. (2004). Interestingly, bleaching did not appear to be dependent on community type, but mortality was dependent on this variable. This seems to relate directly to studies that show that while all species of coral bleach, certain ones are more likely to recover. The paper is strongly oriented towards developing a framework for protected area design. Yamano, H, and M Tamura. 2004. Detection limits of coral reef bleaching by satellite remote sensing: Simulation and data analysis. Remote Sensing of Environment 90:86-103 The ability of Landsat Thematic Mapper images to detect coral bleaching is examined. Landsat TM has a relatively low spatial resolution (30 m), but the authors show that it can detect coral bleaching when bleaching is extensive. When bleaching was 25-50% of coral cover, bleaching could be picked up by Landsat TM blue and green bands (just as in IKONOS images), but bleaching could not be detected when actual bleaching was 15%. Part of the problem with the images is the high degree of spatial misregistration that occurs when pixels include sand, which has similarly high reflectance to corals. The depth of detection (17 m) was similar to that reported for IKONOS images by Elvidge et al. (2004). The authors make a distinction between spectral resolution and spatial resolution, with the latter being the primary concern with the Landsat TM images. Overall, the authors do not seem to promote the use of Landsat TM except under conditions of very high coral cover and extensive bleaching. They do, however, suggest that satellite remote sensing of coral bleaching is sure to become more useful in the near future. I m not clear on why they spent so much time with this since it is obviously not