Conceived and designed the experiments: JB ES. Performed the experiments: JB ES. Analyzed the data: JB ES. Contributed reagents/materials/analysis tools: JB ES. Wrote the paper: JB ES.
The authors have declared that no competing interests exist.
A number of factors have recently caused mass coral mortality events in all of the world's tropical oceans. However, little is known about the timing, rate or spatial variability of the loss of reef-building corals, especially in the Indo-Pacific, which contains 75% of the world's coral reefs.
We compiled and analyzed a coral cover database of 6001 quantitative surveys of 2667 Indo-Pacific coral reefs performed between 1968 and 2004. Surveys conducted during 2003 indicated that coral cover averaged only 22.1% (95% CI: 20.7, 23.4) and just 7 of 390 reefs surveyed that year had coral cover >60%. Estimated yearly coral cover loss based on annually pooled survey data was approximately 1% over the last twenty years and 2% between 1997 and 2003 (or 3,168 km2 per year). The annual loss based on repeated measures regression analysis of a subset of reefs that were monitored for multiple years from 1997 to 2004 was 0.72 % (n = 476 reefs, 95% CI: 0.36, 1.08).
The rate and extent of coral loss in the Indo-Pacific are greater than expected. Coral cover was also surprisingly uniform among subregions and declined decades earlier than previously assumed, even on some of the Pacific's most intensely managed reefs. These results have significant implications for policy makers and resource managers as they search for successful models to reverse coral loss.
There is growing scientific and public awareness of the widespread depletion of marine habitat-forming species, such as mangroves, seagrasses, oysters, and corals [e.g., 1,2,3]. This loss inevitably leads to the decline of the plants and animals that live in the biogenic structures created by such foundation species, and contributes to the overall degradation of marine ecosystems
Scientists have recognized the ecological and economic value of coral reefs and the threats to reef-building corals for decades
Here we describe a comprehensive analysis of the timing, rate, and geographic extent of the loss of coral cover across the Indo-Pacific (
This study provides the first regional scale and long-term analysis of coral cover in the Indo-Pacific. Our results indicate that the loss of coral cover began earlier than assumed and that coral cover is currently very similar across the Indo-Pacific, suggesting that coral decline is a general global phenomenon.
Our analyses were based on quantitative surveys that measured the percentage of the bottom covered by living scleractinian corals on subtidal coral reefs (1–15 m depth, mean survey depth was 6.2 m) within ten subregions of the Indo-Pacific (
Coral cover data from the Australian Institute of Marine Science's (AIMS) Long Term Monitoring Program (LTMP)
Most surveys were based on the line transect technique or some variant to estimate coral cover. A transect (typically 10–30 m in length and usually a tape measure or chain) is placed on the reef, oriented either along a depth contour or down the reef slope. Coral cover is then estimated either
We also included data from manta tow surveys of coral cover
Most of the reefs in the database were surveyed only once, but a subset of 651 reefs were surveyed two or more times (
Linear repeated measures regression analysis was used to test the null hypothesis that there was no relationship between coral cover and time from 1968 to 2004. We were unable to perform a formal meta-analysis because several critical components (e.g., variance estimates, sample size, repeated sampling of each reef, etc.) were not available for all data sets. We used Stata (version 9.1, STATA Corp.) and performed two sets of analyses: (1) on the annual subregional means based on all 6001 surveys, and (2) on the data from the 651 monitoring sites. In both analyses, time (year) and coral cover were treated as continuous variables. Because locations were repeatedly sampled over time, coral cover estimates of a given subregion or reef in different years were not independent. This longitudinal structure was incorporated into the statistical model by using repeated measures of subregions or reefs. Thus, statistical estimates of the absolute net decline in coral cover were based on the individual trajectories of subregions or reefs and were not derived by pooling all the data for each year. For these and all other analyses, data were transformed when necessary to meet basic statistical assumptions.
In the subregion analysis, we used the mean cover in each subregion for each year as the dependent variable, rather than the individual reef means, in part because the sample size varied greatly among years, periods, and subregions. Performing this analysis on yearly subregional averages equalizes the influence of each subregion and prevents the results from being driven primarily by especially well-sampled subregions like the GBR and the Philippines (
For the analysis of the reef monitoring data, we performed four repeated measures regression analyses to test the null hypothesis and estimate the slope of significant linear functions during the entire 36 year range and for each of the three periods: 1970–1983, 1984–1996, and 1997–2004. The period delineations were based on the timing of major disturbance events and expected and observed trends in coral cover in the Indo-Pacific. For example, the beginning of the third period (1997–2004) coincides with a major global mass-bleaching event in 1998 and 1999
Estimates of the rate of coral loss could be influenced by year-to-year and period-to-period changes in the location of reef surveys. For example, if surveys initially focused on high cover reefs or subregions and then shifted focus to low cover reefs, the estimated rate of regional or subregional coral loss could be exaggerated. Alternatively, an initial overrepresentation of low cover reefs or subregions could underestimate the true rate of net coral loss. This problem is diminished in the monitoring sites analysis because individual reefs are monitored through time and reef identity is far less variable. Nevertheless, the identity of monitored reefs did change over time (e.g., when new reefs and subregions were added), so this potential source of bias was not entirely eliminated. A second potential bias in the analyses is the overrepresentation of the best-sampled subregions, mainly the Philippines and the GBR. Therefore, the regression results are not necessarily representative of all ten subregions, especially those that were not well monitored.
Because the effects of a variety of disturbances on coral cover are depth-dependent
Our results indicate that coral cover on Indo-Pacific reefs is currently lower and far more uniform than expected (
(a) Cover (means ± 1 SE) in ten subregions of the Indo-Pacific. Data are from 2003 for seven subregions and from 2002 for three subregions not adequately sampled after 2002 (Hawaiian Islands, Taiwan & Japan, and Western Pacific). Values above the bars are the number of reefs surveyed in each subregion. (b-i) Histograms illustrating percent coral cover in the Indo-Pacific and selected subregions during different periods. (d) is based on
The general absence of quantitative data on reef health has led to several misconceptions about the causes, patterns, and best remedies for global coral decline. For example, in 2003, coral cover on the Great Barrier Reef (GBR), considered the “best-managed”
Additionally, there are other important measures of reef degradation, in particular the abundance and diversity of reef inhabitants
Historically, i.e., 100–1000 y.b.p., average coral cover in the Indo-Pacific was probably approximately 50%
We may never know the precise Indo-Pacific coral cover baseline, but we now know that regionally, cover is currently at least 20% below the best historical reference points. Our results suggest that average Indo-Pacific coral cover declined from 42.5% during the early 1980s (95% CI: 39.3, 45.6, n = 154 reefs surveyed between 1980 and 1982) to 22.1% by 2003 (
(a) White bars represent Indo-Pacific coral cover and open symbols (right axis) are the number of reefs surveyed each year. (b) Coral cover in ten Indo-Pacific subregions in each of three periods. Plotted values are means±1 SE and values above each bar are the subregional sample sizes. * = no data were available
Analysis | n | df | F | p | R2 | Slope (95% CI) |
Subregional means (1968–2004) | 213 | 1,9 | 9.06 | 0.015 | 0.05 | −0.37 (−0.64, −0.09) |
Monitoring sites (1970–2004) | 2994 | 1, 651 | 25.94 | <0.0001 | 0.02 | −0.39 (−0.54, −0.24) |
Monitoring sites (1970–1983) | 186 | 1,109 | 16.07 | <0.0001 | 0.09 | 1.53 (0.77, 2.29) |
Monitoring sites (1984–1996) | 1299 | 1,395 | 0.95 | 0.33 | 0.0013 | −0.19 (−0.57, 0.19) |
Monitoring sites (1997–2004) | 1509 | 1,475 | 15.24 | <0.0001 | 0.01 | −0.72 (−1.08, −0.36) |
Monitoring sites except GBR (1984–1996) | 502 | 1,210 | 0.77 | 0.383 | 0.0027 | −0.29 (−0.93, 0.36) |
Monitoring sites except GBR (1997–2004) | 594 | 1,280 | 7.75 | 0.006 | 0.0162 | −0.90 (−1.53, −0.26) |
Analyses were based on the individual, independent trajectories of subregions or reefs (for the monitoring sites analyses). n = total number of observations
The estimated annual rate of coral cover loss in the Caribbean between 1977 and 2001 was approximately 1.5%, with the greatest decline occurring during the 1980s
Regional and subregional trends in coral cover during the 1970s are less clear than for more recent periods because fewer surveys were performed, few subregions were adequately sampled, and 77% of surveyed reefs prior to 1973 were on the GBR. Therefore, it is unlikely that the regional cover average of 30% between 1968 and 1972 (95% CI: 23.5, 35.5, n = 70) is representative of all subregions, particularly those that did not experience outbreaks of
Our analysis suggests that the regional-scale coral decline in the Indo-Pacific began several decades earlier than often assumed. For example, Pandolfi et al.
Comparing the timing and rate of coral decline among Indo-Pacific subregions is difficult because many were not adequately sampled until the early 1980s. Furthermore, historic baseline coral cover may have varied among subregions due to differences in disturbance frequency or the morphology of dominant species. For example, reefs dominated by plating acroporiid corals probably had higher baseline cover than reefs dominated by branching corals. Thus, similar current cover among subregions could actually reflect variability in the degree of coral loss. Additionally, the dependence of facilitation and other ecosystem functions on coral cover could vary among subregions (e.g., 20% cover might not be universally functionally equivalent).
Between 1984 and 1996, coral cover was slightly lower in east Indonesia than on the GBR (
Major storms, though not novel disturbances, are considered primary causes of recent coral loss in several locations including Hawaii and Moorea
Despite the well-documented effects of several causes of mass coral mortality, there is substantial evidence that coral communities remain resilient, often recovering in ten to thirty years after major disturbances
Average GBR coral cover has been consistently below 27% since 1986 (
The results of our analysis of 6001 quantitative reef surveys indicate that the degree, geographic extent, and duration of the Indo-Pacific coral decline have been significantly underestimated. Many coral reef scientists know of exceptions to the general pattern of reef degradation: there are currently many, perhaps hundreds or even thousands of high coral cover (i.e., >60%) reefs in the Indo-Pacific and Caribbean that resemble the presumed historical coral baseline [e.g.,59]. But our results indicate that such observations are anomalies and currently represent less than 2% of reefs in the Indo-Pacific. This study also highlights the urgent need for conservation policies to restore coral reefs and the ecosystem services they provide, estimated to be worth $23,100–$270,000 km−2 year−1
The loss of coral cover represents both an absolute loss and a reduction in the quality of reef habitat
Calculation of Indo-Pacific reef area
(0.03 MB DOC)
Published data sources
(0.04 MB DOC)
Analysis of potential effects of depth on coral cover estimates
(0.05 MB DOC)
Potential biases in survey techniques and site selection
(0.05 MB DOC)
The number of surveyed reefs in the ten Indo-Pacific subregions.
(0.05 MB DOC)
Characteristics of the eight basic sources of coral cover data.
(0.05 MB DOC)
Number of monitoring sites in the ten Indo-Pacific subregions during each of three periods (note most monitoring sites were surveyed for more than one period).
(0.05 MB DOC)
Results of linear repeated measures regression analyses on the relationship between coral cover and time in the Indo-Pacific. Unlike the results presented in
(0.04 MB DOC)
Patterns of coral cover decline in ten Indo-Pacific subregions. Black bars are mean coral cover ± 1 SE for each year (missing bars are years in which no data are available). Open symbols (right axis) are the number reefs surveyed in each subregion during each year (note changes in scale).
(0.73 MB EPS)
Locations of the 2667 surveyed reefs (green dots). (This KML file can be viewed with the Google Earth mapping system.)
(0.09 MB ZIP)
We thank R. Aronson, K. France, W. Precht and P. Reynolds for helpful comments on the manuscript, A. Melendy for assistance with statistical analysis, and K. France, R. Katz, L. Ladwig, S. Lee, M. O'Connor, C. Shields, G. Smelick and I. Vu for assistance with data collection. We are especially grateful to the many organizations that shared their published and unpublished data with us, including the Australian Institute of Marine Science, the Great Barrier Reef Marine Park Authority, the Hawaii Coral Reef Assessment and Monitoring Program, the National Oceanic and Atmospheric Administration, the Nature Conservancy, ReefBase, Reef Check, the World Resources Institute, and the thousands of divers who collected the data.