Changes in tropical sea surface temperature (SST) are examined over the period 1950–2011 during which global average temperature warmed by 0.4°C. Average tropical SST is warming about 70% of the global average rate. Spatially, significant warming between the two time periods, 1950–1980 and 1981–2011, has occurred across 65% of the tropical oceans. Coral reef ecosystems occupy 10% of the tropical oceans, typically in regions of warmer (+1.8°C) and less variable SST (80% of months within 3.3°C range) compared to non-reef areas (80% of months within 7.0°C range). SST is a primary controlling factor of coral reef distribution and coral reef organisms have already shown their sensitivity to the relatively small amount of warming observed so far through, for example, more frequent coral bleaching events and outbreaks of coral disease. Experimental evidence is also emerging of possible thermal thresholds in the range 30°C–32°C for some physiological processes of coral reef organisms. Relatively small changes in SST have already resulted in quite large differences in SST distribution with a maximum ‘hot spot’ of change in the near-equatorial Indo-Pacific which encompasses both the Indo-Pacific warm pools and the center of coral reef biodiversity. Identification of this hot spot of SST change is not new but this study highlights its significance with respect to tropical coral reef ecosystems. Given the modest amount of warming to date, changes in SST distribution are of particular concern for coral reefs given additional local anthropogenic stresses on many reefs and ongoing ocean acidification likely to increasingly compromise coral reef processes.
 Sea surface temperature (SST) plays a major role in controlling climate and its variability on a range of time scales [Clement et al., 2005; Deser et al., 2010a]. Many organisms are limited to specific thermal ranges [Drinkwater et al., 2010] and SST is one of the primary controlling factors in the distribution of tropical coral reef ecosystems [Kleypas et al., 1999a]. The physical environment for tropical coral reefs is changing as the oceans warm due to increasing atmospheric greenhouse gas concentrations Other physical characteristics of the oceans are also changing, including increased stratification, acidification and large-scale changes in sea surface salinity associated with an intensified hydrological cycle [Sen Gupta and McNeil, 2012; Durack et al., 2012].
 Significant local stresses (the ‘human footprint’ of pollution, increased sedimentation through changed land use, over-fishing, destructive fishing practices) already affect many of the world's reefs. These local factors can affect their resilience to other environmental changes such as thermal stress [Mora and Ginsbur, 2008]. Progressive acidification of the oceans, as they absorb part of the extra atmospheric carbon dioxide, is also projected to eventually compromise the fundamental reef process of calcification [Kleypas et al., 1999b].
 Given the apparent temperature sensitivity of coral reef organisms, this paper examines how tropical oceans have warmed to date, how rates of change compare with average global warming, spatial variations in changes and what these changes mean for SST distributions at coral reef locations. Tropical SST is examined for both coral reef and non-coral reef locations over the period 1950–2011.
2. Data and Methods
 Monthly sea surface temperature (SST), 1950–2011, was extracted from the HadISST1 data set [Rayner et al., 2003] at 1° latitude by longitude resolution for the tropical oceans between 30.5°N and 30.5°S. Although there are some regional differences, HadISST data show similar patterns and trends as other SST data sets [Deser et al., 2010b; Kim et al., 2012; Tokinaga et al., 2012], so the use of this particular data set should not substantially influence the analyses and findings. Global average temperatures (annual anomalies from 1961 to 1990 average) were obtained from the HadCRUT3 land and sea temperature compilation [Brohan et al., 2006].
 The SST data were divided into two subsets: 1670 boxes containing coral communities (coral reefs) and the remaining 15,204 boxes (non-coral reefs). Presence/absence of coral communities in each 1° box was determined from ReefBase (www.reefbase.org) by Serge Andréfouët, IRD, Noumea. On this basis, coral reefs comprise 10% of the tropical oceans. Average annual SST anomalies (from 1961 to 1990 average) were calculated for reef and non-reef sites.
 Average SST and monthly standard deviations were calculated and mapped for the whole period, 1950–2011. Averages were also obtained for the first, 1950–1980, and last, 1981–2011, 31 years of the record period. Differences in means between these two periods were tested for significance using the t-test [Mitchell et al., 1966].
 Maps of average SST distribution, 1950–2011, were obtained by calculating, for each 1° box, the percentage of months within SST ranges: <17°C, 17°C–18°C, … 31°C–32°C, >32°C. A similar “binning” of SST has been used previously, though largely focused on distribution characteristics and changes of the Indo-Pacific warm pool [Cravatte et al., 2009; Lin et al., 2011]. The percentage of months within each temperature class was also calculated for the two time period 1950–1980 and 1981–2011 and the percentage change between periods was mapped. The sum of the absolute (i.e., positive and negative) percentage changes was also mapped to give an indication of locations experiencing the greatest changes in SST. The percentages of months within each temperature class were averaged separately for reef and non-reef areas. Finally, the differences in SST distribution between 1950–1980 and 1981–2011 were examined for five 1° boxes containing coral reefs along longitude 149.5°E at 5.5°N, 5.5°S, 10.5°S, 15.5°S, and 20.5°S.
 Significant warming of the tropical oceans largely tracked global average temperatures over the period 1950–2011 (Figure 1). The two series were significantly correlated (r = 0.90, p < 0.01), as were inter-annual variations (r = 0.82 for first differences,p < 0.01). The rate of warming of the tropical oceans was lower than global temperatures with significant linear trends of 0.08°C/decade and 0.12°C/decade, respectively. Both series were significantly warmer in 1981–2011 compared to 1950–1980 with temperature differences of +0.26°C and +0.38°C, respectively. The tropical oceans as a whole are, therefore, warming at about 70% of the global average rate.
 Coral reefs tend to occur in the warmer parts of the tropical oceans (Figure 2a), with 75% in areas with average SST > 27°C compared with 40% of non-reef areas. Median SSTs for coral reefs were 1.8°C warmer than non-reefs with maximum SST 1.3°C warmer and minimum SST 2.0°C warmer (Table 1). Coral reefs also tend to occur in areas of lower SST variability (Figure 2b) with 80% of months within a 3.3°C range compared to 80% of months within a 7.0°C range for non-reef areas (Table 1).
Table 1. Median (90th–10th Percentile) SST, 1950–2011 for All Months, Annual Maximum and Minimum SST for Reef and Non-Reef Sites (°C)
 Reef and non-reef areas showed a similar pattern of warming, 1950–2011 (Figure 3), with non-reef areas (90% of the tropical oceans) obviously closely matching variations of the whole tropical ocean (cf.Figure 1). Annual SSTs for reef and non-reef areas were significantly correlated (r = 0.84,p < 0.01) though the relationship was weaker for inter-annual variations (r = 0.52 for first differences,p < 0.01). The rate of warming was also similar, though marginally higher for reef (0.09°C/decade) compared to non-reef areas (0.08°C/decade). Average SSTs for 1981–2011 were significantly warmer for both reef (+0.28°C) and non-reef (+0.26°C) locations.
 Although average SST of the tropical oceans has warmed, the magnitude and significance of warming has varied spatially (Figures 4a and 4b). Sixty-five percent of the tropical oceans have significantly warmed, 34% show no significant change and <1% (in the northern tropical Pacific) have significantly cooled. Observed warming was greatest in the northwest and northeast tropical Pacific and southwest tropical Atlantic. Relative to inter-annual variability, however, the greatest changes occurred in the near-equatorial Indian and western Pacific and Atlantic Oceans (Figure 4c).
 Another way to consider thermal conditions for marine ecosystems, such as coral reefs, is how much time organisms are subjected to particular temperatures. Average percentages of months within each SST class (<17°C to >32°C) are provided in Figure S1. These illustrate that up to 26°C–27°C, most parts of the tropical oceans spend between 10 and 30% of the time within different SST ranges. Above 27°C, however, parts of the tropical Atlantic spend >50% of months within a given 1°C range (Figure 5a). This dominance within given 1°C ranges becomes evident for the near-equatorial Indo-Pacific for 28°C–29°C (Figure 5b) and for 29°C–30°C in the western equatorial Pacific (Figure 5c). There is also a small area northeast of Papua New Guinea where SST are within the 29°C–30°C range for >70% of months.
 Coral reefs tend to be in the warmer parts of the oceans with a narrower range of SST compared to non-reef locations (Figure 6a). Coral reefs spend ∼70% of the time within a 3°C SST range (27°C–30°C) compared to non-reefs which spend ∼70% of the time within a 6°C SST range (24°C–30°C).
 So, with the observed warming of much of the tropical oceans over the period 1950–2011 (Figures 3 and 4), how have these SST distributions changed? For non-reefs the loss and gain of SST within different 1°C ranges was generally <4% between 1950–1980 and 1981–2011. The changes for coral reefs due to warming were, however, more substantial with 2.7% and 5.7% losses of the 27°C–28°C and 28°C–29°C ranges, respectively, associated with a 10.1% increase of the 29°C–30°C range (Figure 6b). Spatially, losses and gains between the two periods occurred throughout much of the tropical oceans but have generally been small (5–10%) up to ∼26°C (see Figure S2). For the 26°C–27°C range, however, marked losses (30–40%) occurred in the near-equatorial Atlantic Ocean (Figure 7a). Similar magnitude losses occurred in the central near-equatorial Indian Ocean for the 27°C–28°C range (Figure 7b). SST in the range 28°C–29°C showed a complicated pattern of losses in the vicinity of the western Pacific warm pool (exceeding 50%) and gains in the near-equatorial Atlantic [cf.Lin et al., 2011] and parts of the central Indian Ocean (Figure 7c). For the 29°C–30°C range, substantial increases were concentrated in the near-equatorial Indo-Pacific region between ∼60°E and the dateline (Figure 7d). The areas of greatest total change (i.e., absolute sum of both losses and gains) affected <20–30% of months throughout much of the tropical oceans with substantial changes (affecting >70% of months) concentrated in the near-equatorial Indo-Pacific from the central Indian Ocean to the dateline and, to a lesser extent, the near-equatorial tropical Atlantic (Figure 8). These changes in SST distribution reach a maximum of >90% of months affected north of Papua New Guinea and just south of the equator in the central Indian Ocean.
 For coral reef locations, the percentage of months within the 28°C–29°C significantly decreased by 1.7%/decade (p < 0.01) and the 29°C–30°C range significantly increased by 3.1%/decade (p < 0.01) over the period 1950–2011. Although the change in the 31°C–30°C range was much smaller, the trend of 0.4%/decade was significant (p < 0.01; Figure 9). Annual variations in the 28°C–29°C range were significantly negatively correlated with both the 29°C–30°C (r = −0.96; r = −0.87 first differences, p < 0.01) and 30°C–31°C ranges (r = −0.81; r = −0.69 first differences, p < 0.01). This again suggests that as coral reef regions warm the decline of the 28°C–29°C range is due to increases in the higher SST classes.
 As a final illustration of how SST distributions are changing for coral reefs, five sites along 149.5°E from 5.5°N, north of Papua New Guinea, and 20.5°S in the central Great Barrier Reef, Australia were examined. These sites were selected as they encompass a range of average SST and variability (Table 2) and are located from the region of maximum change in SST distribution north of Papua New Guinea (>90%) to the central Great Barrier Reef where the changes are <30% (Figure 8). Average annual SST at all five sites was significantly warmer in 1981–2011 compared to 1950–1980. Significant warming was also evident between the two periods for most months of the year (Figures 10a–10e). The resulting changes in SST distribution were, however, most marked at the near-equatorial sites (Figures 10f–10j). At 5.5°N, 62% of months had SST within the 28°C–29°C range for 1950–1980 but in 1981–2011 only 15% of months were within this range and 80% of the time was in the 29°C–30°C range. This contrasts with the smaller changes spread over a wider SST range evident at 15.5°S and 20.5°S. So, although all sites have significantly warmed, the changes in SST distribution are substantially different. These smaller magnitude changes in SST distribution at higher latitudes of the tropical oceans, with losses and gains of less than 10%, are also illustrated in Figure S2.
Table 2. Average SST Characteristics for Five 1° Boxes ∼149.5°E, 1951–2011(°C)
Average ± SD
29.1 ± 0.5
29.0 ± 0.7
27.7 ± 1.3
27.0 ± 1.5
25.4 ± 2.3
4. Discussion and Conclusions
 The tropical oceans are warming at about 70% of the rate of average global temperatures. The magnitude and significance of recent warming of the tropical oceans varies spatially with 65% significantly warmer and 34% showing, as yet, no significant change. Less than 1% of the tropical oceans have significantly cooled between 1950–1980 and 1981–2011. Coral reefs occupy 10% of the tropical oceans and tend to occur in the warmer (+1.8°C) parts compared to non-reefs. Coral reefs also tend to occur in regions with lower SST variability (range 3.3°C) [Donner, 2011] compared to non-reef areas (range 7.0°C). SSTs for coral reefs tend, therefore, to show the skewed distribution typical of the warm pool regions [Clement et al., 2005; Williams et al., 2009] compared to the flatter distribution for non-reef sites. With the relatively modest amount (compared to projected changes for the end of this century [Meehl et al., 2007]) of global warming observed so far (+0.4°C) between 1950–1980 and 1981–2011, there have been substantial changes in the SST distributions for coral reefs. Reef and non-reef parts of the tropical oceans are warming at similar rates. The resulting changes in SST distributions for coral reefs are, however, more substantial with 8.4% loss of SST in the range 27°C–29°C and associated 10.1% gain of SST in the range 29°C–30°C between 1950–1980 and 1981–2011.
 A “hot spot” of change is evident in the near-equatorial (∼10°N–10°S) Indo-Pacific region from about 60°E in the central Indian Ocean to the dateline in the central Pacific. This also encompasses the region of greatest coral reef diversity as measured, for example, by the number of hard coral species in the Coral Triangle [Veron, 2008]. This “hot spot” is evident when the SST changes observed between 1950–1980 and 1981–2011 are scaled by SST variability (Figure 4c) and by summing the absolute differences in percentage of months within particular 1°C ranges (Figure 8). Substantial (affecting >50% of months) changes have already occurred in the thermal envelopes that tropical coral reef organisms are used to.
 The near-equatorial parts of the Pacific Ocean are projected to warm most over the 21st century [Meehl et al., 2007; Sen Gupta and McNeil, 2012]. Projecting levels of warming and aragonite saturation state over the 21st century, Meissner et al.  also found a “hot spot” of severe thermal stress (even for the most optimistic Representative Concentration Pathway projection by 2030) around Micronesia, northern Mariana Islands and Papua New Guinea. Analyses of land and sea temperatures, 1960–2009, highlights the spatial complexity of the “pace” of recent changes [Burrows et al., 2011]. These authors also identify the near-equatorial Indo-Pacific as one of rapid recent thermal changes and note that in such regions, if organisms are compelled to move to cooler sites, there are no communities adapted to even warmer temperatures to take their place. An additional constraint on the potential for coral reef communities to change their distribution is the availability of suitable shallow-water substrate [Kleypas et al., 1999a].
 Background variability appears important in modulating coral responses to thermal stress [Boylan and Kleypas, 2009; Carilli et al., 2012]. Using just a local thermal stress threshold, for example, suggests that bleaching stress is greatest in the central and equatorial Pacific but once historical variability is included this potential bleaching “hot spot” shifts westward to the Coral Triangle and parts of Micronesia and Melanesia [Teneva et al., 2012]. The lower temperature variability of tropical regions compared to higher latitudes also results in earlier emergence of significant climate change [Mahlstein et al., 2011].
 Tropical coral reefs are concentrated in the warmer parts of the tropical oceans with 75% in areas with average SST > 27°C, i.e., temperatures characteristic of the Indo-Pacific warm pools. Several studies have identified recent changes in warm pool SST characteristics. The near-equatorial Indo-Pacific and equatorial Atlantic regions, identified in the present study of greatest changes in SST distribution, also show significant decreases in sea surface salinity over recent decades [Durack et al., 2012]. The western Pacific warm pool is warming, increasing in size and extending further eastward [Cravatte et al., 2009] while the Indian Ocean warm pool is expanding westward [Williams and Funk, 2011; Kim et al., 2012]. The patterns of SST changes identified in this study, especially in the Indo-Pacific, are similar to those reported elsewhere [Lin et al., 2011] which are considered consistent with a weakening of the Walker and Hadley Circulations of the Pacific Ocean [Deser et al., 2010b; Tokinaga et al., 2012].
 These studies were primarily concerned with the causes and consequences of such relatively small changes in the surface ocean for climate variations and determining how the equatorial Pacific SST gradient has and may change in a warming climate [Vecchi et al., 2008; Karnauskas et al., 2009]. This thermal gradient is of particular importance for El Niño-Southern Oscillation (ENSO) events and hence, how their frequency and intensity may change in a warmer world. Consensus among climate models as to how ENSO could change was equivocal in the IPCC-AR5 assessment [Meehl et al., 2007] and appears still to be so with the new generation of climate modeling outputs that are starting to appear for IPCC-AR5 [Guilyardi et al., 2012]. Whether ENSO events become more intense and/or more frequent is also of significance for coral reef ecosystems as elevated annual maximum SST, conducive to coral bleaching, is extensive throughout the tropical oceans during El Niño events [Eakin et al., 2009].
 The importance of the magnitude of natural climate variability in a warming climate is of considerable significance for tropical coral reef ecosystems with reefs spending 80% of the time within a 3.3°C SST range, which is about half the thermal envelope of non-reef areas (Table 1). As identified in several recent studies, this lower thermal variability, both on land and at sea, in the tropics leads to earlier emergence of a climate signal associated with global warming [Burrows et al., 2011; Mahlstein et al., 2011; Hawkins and Sutton, 2012].
 How the warmer parts of the tropical oceans continue to warm is of interest to climate scientists because of the fundamental role these regions play in tropical and global climate dynamics and inter-annual variability, such as ENSO. How these regions warm and the rate of warming is also of concern for coral reef scientists in determining how tropical coral reef ecosystems have and will respond to a rapidly changing physical environment. Small temperature changes have already produced large differences in the thermal environments of some coral reefs. Can global climate models realistically simulate these changes in SST distribution already observed and thus provide greater confidence in future projections for the coming century?
 It is not just the magnitude of warming of average SST that is of concern for coral reef processes in a changing world. Coral reef scientists need information, for example, about likely changes in seasonal SSTs (e.g., annual maximum which is the most relevant indicator of thermal stress associated with bleaching), when and where absolute temperature thresholds are likely to be exceeded and possible changes in SST variability. This information also needs to be available for coral reef locations for, as shown in this study, the average thermal characteristics and observed changes for the tropical oceans are not entirely representative of coral reef environments. Projections of SST changes also need to be combined with similarly spatially resolved information about ongoing changes in sea surface salinity and aragonite saturation state. Such information will also help inform and refine experimental studies that manipulate physical environmental parameters for coral reef organisms. Such studies are already suggesting that there are fixed temperature thresholds which, once exceeded, may compromise various physiological processes of a range of coral reef organisms. The results presented here suggest that in some regions of the tropical oceans these potentially critical thermal thresholds are already being reached. We need to know whether these thresholds are indeed fixed or whether they will shift with continued rapid changes in the physical environment of tropical coral reef ecosystems. Coral reef ecosystems are not just corals. The structural complexity they create provides habitat and food for many thousands of reef-associated organisms and millions of people depend on coral reefs for their food and livelihoods. The vast majority of these have contributed almost nothing to the atmospheric compositional changes that are now starting to affect coral reefs [Donner and Potere, 2007].
 I am extremely grateful to Serge Andréfouët, IRD, Noumea for identifying the 1° boxes containing coral communities from ReefBase.