Rapid Recent Warming of Coral Reefs in the Florida Keys

Coral reef decline in the Florida Keys has been well-publicized, controversial, and polarizing owing to debate over the causative agent being climate change versus overfishing. The recurrence of mass bleaching in 2014, the sixth event since 1987, prompted a reanalysis of temperature data. The summer and winter of 2014 were the warmest on record. The oldest known in-situ temperature record of any coral reef is from Hens and Chickens Reef (H&C) in the Florida Keys, which showed significant warming from 1975–2014. The average number of days ≥31.5 and 32oC per year increased 2670% and 2560%, respectively, from the mid-1990 s to present relative to the previous 20 years. In every year after 1992 and 1994, maximum daily average temperatures exceeded 30.5 and 31°C, respectively. From 1975–1994, temperatures were <31 °C in 61% of years, and in 44% of the years prior to 1992 temperatures were <30.5 °C. The measured rate of warming predicts the start of annual bleaching between 2020 and 2034, sooner than expected from climate models and satellite-based sea temperatures. These data show that thermal stress is increasing and occurring on a near-annual basis on Florida Keys reefs due to ocean warming from climate change.

Scientific RepoRts | 5:16762 | DOI: 10.1038/srep16762 The recurrence of mass coral bleaching in 2014 prompted a reanalysis of the long-term in-situ temperature data that was initially studied from 1988-2005 for MLRF 15 . These data are supplemented with additional analyses of the H&C dataset with inclusion of new data from 2014 to determine how the warming shown by Kuffner et al. 18 corresponded to an increase in the number of days ≥ 30.5, 31, 31.5, and 32 °C each year. We focused on August and September because this is when data were most complete, as well as when temperatures reach seasonal maxima 15 . These temperatures were chosen because number of days ≥ 30.5 °C correlated with mass bleaching events in 1997, 1998, and 2005, one day ≥ 31 °C was predicted to cause bleaching at MLRF 15 , and Jaap 20 observed bleaching when temperatures reached 32 °C. Lastly, these long-term historical records are considered alongside in-situ temperatures from paired inshore and offshore sites in the upper, middle, and lower Florida Keys from 2010-2012 (Fig. 2).

Results and Discussion
The summer of 2014 and winter of 2013-2014 were the warmest on record for MLRF (Fig. 3). Prior to 2014, the warmest winter and summer on record were in 1997 at the start of the extreme 1997-1998 El Niño warming event. H&C exhibited a significant warming trend from 1975-2014 for number of days ≥ 30.5, 31, 31.5, and 32 °C, illustrating that the exposure of corals to thermally stressful conditions has increased significantly over the past 40 years (Fig. 4, Table S1). These trends held when the anomalous El Niño years of 1997-1998 were removed, except for days ≥ 32 °C. The lack of significance for days ≥ 32 °C is likely because only three out of the 26 years of data had non-zero values (1980,2005,2014). Seventeen of the past 27 years at MLRF were warmer than the climatology. In spite of this, there were no trends in: 1) the entire MRLF dataset (all data with time), 2) monthly mean temperatures (each month by year), or 3) days ≥ 30, 30.5, and 31 °C. However, when the El Niño years of 1997-1998 were removed, the trends for days ≥ 30.5 and 31 °C with time were positive and significant (Table S1). As previously mentioned, the MLRF dataset was initiated after the first mass bleaching event in the Florida Keys and temperatures at H&C warmed considerably in the early 1990 s (Fig. 4, Table 1). This coupled with the anomalously warm El Niño years of 1997-1998 could explain the muted warming signal for MLRF, as it likely biased the first 10 years of data towards higher values.
In every year of available data since 1996, maximum daily temperatures at H&C were ≥ 31 °C (Fig. 4, Table 1). Prior to 1995, maximum daily sea temperatures were < 31 °C in 11 of the 18 years of available data (61%) from 1975-1994. In almost half of the years from 1975-1992 (44%, 7 of 16 yrs), maximum temperatures were < 30.5 °C, but then exceeded this value every year beginning in 1993. Temperatures ≥ 32 °C were recorded for only one day in total from 1975-1996 (0.05 days year −1 ), but increased 2560% to a frequency of 1.33 days year −1 thereafter (  The cumulative number of hours > 32 °C over three years at the three inshore sites was 850-889 hours, whereas there were 904 hours over four years in 1977-1980 at Snake Creek channel. The inshore sites in this study experienced a comparable dosage of cumulative thermal stress over three years (2010-2012) as Snake Creek did over four years in the late 1970 s. Snake Creek was specifically chosen by Hudson 21 to illustrate its thermally unsuitable nature for corals and the coral transplanted there bleached from warm water stress. This site gets hot because it is shallow and subject to the outflow of Florida Bay waters, which heat up considerably during summer 21,22 . Table 3 shows the year when the duration of temperatures that were associated with major bleaching events in 1997, 1998, 2005, and 2014 are predicted to occur annually based on the regressions of Fig. 4 and Table S1. The time-temperature bleaching curve presented in Manzello et al. 15 for MLRF predicts that bleaching will occur when daily average temperatures are ≥ 30.5 °C for 16 days, or 1 day ≥ 31 °C. A minimum of 23 days ≥ 31 °C were associated with major bleaching at H&C in 1997, 1998, 2005, and 2014. When just the H&C data were used, bleaching is expected to occur annually within the next 20 years at this site (2021-2034). The most conservative estimate for H&C, using the values for the missing years that were estimated from nearby Snake Creek (see methods), predicts annual bleaching in 2045. When the values from the upper keys inshore site that is 1.4 km away from H&C were included to fill in missing years from 2010-2013, the regression with time changed little, predicting annual bleaching from 2027-2035. Annual bleaching was predicted to occur for MLRF no later than 2020. Annual bleaching is predicted to occur later than 2040 based on satellite sea surface temperatures (SST) and the most severe emission scenario (RCP 8.5) for reefs in the Florida Keys near H&C and MLRF 23 . All the predictions based on in-situ temperatures are much earlier than this, except for the most conservative estimate of 2045. This suggests that the in-water thermal environment in the Florida Keys may be more complex and warmer than what is captured from satellite-based SSTs. The presence of an inverse thermocline due to dense, salty, and hot waters exiting Florida Bay is one such mechanism whereby sea temperatures are higher near the corals than the surface and may be a factor in these differences 24 .
During the first widespread bleaching event in 1987, temperatures at Alligator Reef, an offshore reef site SW of H&C, were 30 °C or slightly greater for about a month 25 . These temperatures were measured at MLRF in 18 of 27 years from 1988-2014. The value that co-occurred with the 1990 mass bleaching (max. running 30-day mean = 30.2 °C) 14 occurred in 14 of the 27 years and 8 of the past 11 years (Fig. 3). These observations suggest that the bleaching thresholds for MLRF could be conservative and that stressful conditions may manifest as low as monthly temperatures of 30 °C. Heat stress may be more chronic than what is currently appreciated and there could be a shifting baseline whereby summertime bleaching has become the norm 26 . It's likely no coincidence that 2013 was the first year since 1996 that was considerably cooler than the climatology during summer (Fig. 3, max temperature = 29.5 °C) and is the only year in recent record when reports of bleaching were low 16 . The highest growth and calcification rates in Orbicella faveolata from 2004-2013 and Porites astreoides from 2001-2013 occurred during this cool period 17,27 .
Bleaching thresholds may increase as field and experimental evidence have shown that repeated warm water stress or routine exposure to high and variable temperatures can lead to elevated thermal tolerance due to adaptation and acclimatization [28][29][30][31] . As summarized in Chollett et al. 32 , the rate of seasonal warming is also important such that a rapid rate of temperature increase leads to a decreased potential for the holobiont to cope with thermal stress, aggravating the bleaching response and post-stress fate of the reef. The rate of seasonal warming at MLRF was high during all bleaching years (Fig. S1), which may have deterred any uptick in the bleaching threshold to-date. The potential for adaptation and acclimatization will likely not be able to compensate for unabated warming 31 . Furthermore, thermal stress does not just elicit coral bleaching and mortality, but also results in increases in coral disease, while causing declines in reproductive output, linear extension, and coral calcification 17,27,33,34 .
The interaction of heat stress and irradiance ultimately drives the bleaching response 35 , but all large-scale coral bleaching episodes over the past 30 years have been associated with warmer than average temperatures 11 . Local-scale variability during bleaching can be due to differential irradiance, as well as differential susceptibilities of coral taxa to heat stress 36,37 . In the Caribbean, however, the most important reef-building taxa, namely the acroporids and Orbicella annularis spp. complex, have experienced the greatest declines owing mainly to bleaching and disease 8 which have increased in relative abundance with the losses of the aforementioned species, have little potential to construct or maintain architecturally complex framework structures that are vital to ecosystem function 39 . The flattening of reefs, or loss in architectural complexity, has already occurred across many Caribbean reefs and many reefs are presently in a state of accretionary stasis, or net erosional, including those in the Florida Keys 6,40,41 . Thus, the differential susceptibilities to warm-water bleaching and persistence of certain taxa has had little impact in slowing the large scale degradation of Caribbean coral reefs. The shallow back-reef pools in Ofu, American Samoa have been of recent interest because > 50 coral species are able to tolerate short periods of high temperatures in this location 42   sites had similar maximum temperatures in 2010-2012, but the mean summer temperatures in Florida were generally > 1 °C warmer than Samoa (Table 4). At the inshore sites, the number of hours > 32 °C were much greater than Samoa, as the range was 133-404 hours per year at the reef sites from 2010-2012 (Table 2). Tennessee Reef, an offshore reef site where temperatures are more buffered than inshore from extremes, experienced 105 hours > 32 °C in 2010. The Florida Keys routinely experienced acute temperature stress similar to Ofu, but also greater chronic heat stress given the higher mean temperatures both inshore and offshore. The acceleration of both acute and chronic heat stress over the past ~30 years coincides with the degradation and loss of coral cover in the Florida Keys, suggesting that climate change and ocean warming are playing a crucial role in this degradation. These data suggest that the iconic coral reefs of the Florida Keys are routinely experiencing high levels of thermal stress on a near-annual basis. Given the magnitude of the recent in-situ thermal stress relative to that from the late 1970 s, the impact of heat stress in the Florida Keys may be underappreciated (Table 2). Chronic warm-water stress could help explain the continued decline of the Oribicella annularis species complex since the 1997-98 El Niño, the very low coral cover at most reefs other than the inshore patch reefs, and little recovery 44 . The synchronous decline of coral populations both within and outside of protected areas 45 coupled with the data presented here provide further evidence that the loss of coral cover in the Florida Keys is a result of ocean warming and climate change.

Materials and Methods
There were three different temperature records utilized in this analysis. First, hourly sea temperatures from the MLRF Coastal Marine Automated Network (C-MAN) station were analyzed from 1988-2014 (ref. 46) (Fig. 2). The climatology and bleaching thresholds for this site were determined as previously described 15 18,21 . A temperature sensor (Seabird SBE 56) was deployed at H&C in December 2013 at the same depth of the historical thermistor (4 m) and data from August and September of 2014 are presented. Data were collected every 5 minutes and converted to hourly and daily averages. Poisson linear regression, necessary for count data 47 , was used to determine if there was a trend with time in number of days ≥ 30.5, 31, 31.5, and 32 °C each year. The Durbin-Watson statistic was calculated to test for temporal autocorrelation and showed that these data did not exhibit autocorrelation. These regressions were used to estimate the year when temperatures associated with prior mass bleaching events are predicted to occur annually. Statistical analyses were performed using SigmaPlot 12.
The  21 alongside temperatures from a nearby site (Peanut) that is closer to shore and experienced warmer temperatures. The site closer to shore continued to cool and maximum temperatures never reached 30 °C during the period that data were unavailable at H&C. As a result, it is highly unlikely that temperatures were > 30 °C at H&C given that it was always cooler than this other site when data overlapped from 1977-1980. Thus, it was assumed temperatures were < 30 °C for the gaps in data in 1976 and 1977.
Hudson 21 quantified thermal stress as number of hours > 31 and 32 °C from 1976-1980. We calculated these metrics from six paired inshore and offshore sites in the upper, middle, and lower Florida Keys from 2010-2012 (the third temperature record) for comparison (Fig. 2). These sites were carefully selected so that they had similar depths (4-6 m). Temperature was measured every 30 minutes using HOBO Pro V2 thermistors (Onset Corp.) that were ~0.5 m above the substrate attached to rebar hammered into the reef. Hens and Chickens is 1.4 km (or 0.89 miles) away from the inshore site in the upper Florida Keys (Fig. 2).
for supporting this work and the Florida Keys National Marine Sanctuary for allowing this work to take place (Permits #FKNMS-2008-091 and #FKNMS-2011-049). In particular, Joanne Delaney of FKNMS provided valuable assistance with permitting. R. Carlton, I. Enochs, and G. Kolodziej helped with field work and I. Enochs made figure 2. The manuscript contents are solely the opinions of the author and do not constitute a statement of policy, decision, or position on behalf of NOAA or the US Government.