Environmental basis for experimental temperature conditions

We selected experimental temperatures based on environmental monitoring data from sites in Bocas del Toro, Panama (Fig. 2). The prior thermal exposure was chosen to meet the minimum temperature value and duration to be considered a local MHW, as defined by Oliver et al. (2018). We based this on the longest time series available, a 19-year record of temperature measurements from loggers deployed on reefs at 3 m (STRI’s Physical Monitoring Program; December 2000–May 2019). These temperatures were recorded every 30 minutes on 2 shallow reef sites, Isla Roldan Reef (−82.3376, 9.29164) and Isla Cayo Agua Reef (−82.110639, 9.242583), using HOBO Stow-Away TidbiT or HOBO Water Temperature Pro V2 (Onset Computer, Bourne, MA), with an accuracy of 0.25 °C. We took daily measurements at noon and calculated the 90th percentile temperature from the combined data from both sites. This was 30.5 °C and used as our prior MHW exposure treatment. We used the maximum temperature, 32 °C, as the extreme temperature in the subsequent hypoxia exposure (Fig. 1; Table 1). The ambient control temperature was 28 °C, which was slightly below the average bay temperature.

Determination of experimental hypoxic conditions

Our goal in the hypoxia treatments was to ensure that the oxygen levels were lower than the species’ P_crit across the experimental treatments. Therefore, we found the average P_crit under each temperature treatment (28 °C and 32 °C) across a range of sizes (n = 31). In late June and early July 2019, we snorkeled to collect animals from just below the surface to depths of 3 m in a well-exposed mixed rocky coral reef environment (Hospital Point, −82.2164, 9.3326). Animals were opportunistically chosen from the reef based on their accessibility and size. They were quickly dislodged from the rocky reef substrate by wedging a teaspoon in between the reef rock and animal and transported to the lab in large coolers filled with 70 L of seawater within 30 minutes of collection. They were placed in 150-L seawater tables with flowing seawater and allowed to adjust to laboratory aquaria conditions for at least 24 hours before use in respirometry trials.

Closed-chamber P_crit respirometry trials were conducted at 28 °C or 32 °C, with each animal participating in a single trial. For trials at 32 °C, the seawater was ramped from ambient to 32 °C over 1 hour in an isolated 10-L tank before the trial. Each animal was then individually placed in a closed glass respirometry chamber filled with about 0.19 L of ∼100% air-saturated 45-μm filtered seawater. Chambers were hermetically sealed glass jars equipped with a magnetic stirring bar, which was separated from the animal with a mesh screen. Chambers were placed in a 12-L water bath over a magnetic stir plate to keep temperatures constant. Oxygen in the chamber was recorded with fiber-optic sensor technology (Fibox 4 and Pst-3 sensors, PreSens, Regensburg, Germany), every 10 s until the oxygen level fell to 5% (1.0 kPa, or 0.3 mg L⁻¹). Temperature in the water bath was maintained with aquarium heaters and a chiller and recorded simultaneously with the Fibox 4 as well as a thermocouple thermometer (HH802U, Omega, Norwalk, CT; accuracy, 0.1 °C). Upon reaching the oxygen depletion target in the chamber, each animal was promptly moved into ambient oxygenated water. Wet weight was measured using a precision balance (PL6001E, Mettler Toledo, Columbus, OH; accuracy, 0.1 g); and volume was measured via displacement, using a graduated cylinder and beaker. The animals were then euthanized by rapid cooling in seawater and dissected to determine the presence of gonads and, thus, maturity. All urchins collected were mature.

We determined P_crit for each animal, using the broken stick regression approach (Yeager and Ultsch, 1989). In this approach, two segments of the oxygen consumption time series data are iteratively fit until the intersection with the smallest sum of the residual sum of squares between the two linear models is found. This is defined as the breakpoint, or P_crit, which we found using the respR package in R; we report the intercept (Harianto et al., 2019). Average P_crit values were generated for binned size groups (small and large) due to the broad range of sizes collected (6.51 to 26.17 g) in each of the two temperatures (28 °C and 32 °C). Additionally, we determined the resting standard metabolic rates (SMRs) from these P_crit trials. We calculated the rate of oxygen consumption inside the respirometry chambers from the initial saturation level of about 100% to about 65%, with all lower bounded values well above the respective individual’s P_crit. From this subset of the data, we calculated both absolute (μmol O₂ h⁻¹) and mass-specific (μmol O₂ h⁻¹ g⁻¹) oxygen consumption rates for each individual. The latter accounts for animal mass (g), corrected-chamber volume, trial temperature, and salinity (see supplemental data and code in Data Availability section, below).

All respirometry trials occurred between July 1 and July 20, 2019, on 11 different days. No animal was in the laboratory for more than five days prior to being used in a respirometry trial. Oxygen sensors (Pst-3 sensor spots) inside each respirometry chamber were calibrated once before all trials, using 100% and 0% point air saturation calibration standards. Additionally, no animals were fed before trials, and they were thoroughly rinsed with 0.45-μm filtered seawater to reduce attached bacteria, sediment, or microalgae that could interfere with the respirometry measurements. To account for microbial respiration, we included an additional chamber filled with the seawater used to rinse the animals undergoing measurements. These blank chambers were run simultaneously during each trial on each day. Microbial respiration was effectively contained to marginal fluxes (and the animals’ oxygen consumption was constant, i.e., linear decrease in oxygen concentration during trials; see supplemental data). Chamber volumes were 0.19 L, animal volumes ranged from 0.03 to 0.01 L (0.014 ± 0.005 L, mean ± SD), and trials lasted 4.41 h on average (±1.80 SD), with durations ranging from 2.16 to 7.90 h.

We determined the effect of temperature on P_crit and SMR, using a two-way ANOVA with temperature, size, and their interaction as factors. We removed interactions from models when interactions were not significant, checked for overdispersion and normality by building histograms, and plotted residuals against fitted values to ensure linearity. We also used Levene’s and Shapiro’s tests to ensure homogeneity of variance and normality of residuals. All assumptions were met for SMR, as well as P_crit after it was log transformed. Pairwise comparisons to check for differences in responses for each treatment were conducted using the estimated marginal means test with least significant difference test correction, using the emmeans package in R (Lenth, 2022).

Experimental marine heatwave exposure

We collected 376 E. lucunter animals for the main experiment. Trials were run between August 19 and November 19, 2019. Prior to each trial, new animals were collected from the same site and in the same way as described for respirometry trials. Animals were brought back from Hospital Point Reef to the BRS and allowed to rest after the collection for 24 h in outdoor 150-L seawater tables. They were then evenly distributed into 8 indoor 50-L tanks (20–30 animals per tank), 4 of which were held at the ambient control temperature of 28.5 °C (28.54 ± 0.29 °C) and 4 at 30.5 °C (30.53 ± 0.14 °C) to simulate a MHW. Water temperatures in all eight tanks were maintained using aquarium heaters for five days. During this period, each tank was bubbled with an air stone and constantly refreshed with flowing unfiltered seawater. Animals were not fed but could have scraped biofilm from the sides of the tanks or filtered particles from the water (Schoppe and Werding, 1996). They were provided with brick substrate for shelter and ambient lighting. No mortality occurred due to the 5-day thermal acclimation, but 6 animals showed evidence of skeletal trauma from collection and died during this time. Because all tanks were supplied with flowing seawater during these trials, salinity matched the bay’s, which was on average 35 ± 2 ppt. To ensure that the treatments were even with respect to animal body size, small and large size classes were chosen according to the mean weight (14.39 g), with animals weighing between 3.09 and 14.39 g in the small category and those between 14.40 and 38.96 g in the large category. Small and large animals were evenly distributed between tanks, treatments, and time categories.

Experimental hypoxia exposure

After the 5-day prior exposure treatment, cohorts of 3–4 animals were binned together in plastic mesh cages labeled by MHW treatment and equally distributed across 2 new 50-L hypoxia tanks, 1 each at 28 °C or 32 °C (∼60 individuals per tank). Cages served to keep track of individuals, that is, which treatment they were from, as well as keeping them from climbing out of the hypoxic water. Temperatures in the hypoxia tanks were set before moving the animals, while the oxygen level in each tank was gradually lowered over the course of 1 hour after adding the animals. This was done by bubbling nitrogen until the oxygen concentration reached 16.5% (3.4 kPa, or 1.01 ± 0.22 mg L⁻¹). This was, on average, 35% below the P_crit regardless of body size or temperature. Oxygen levels were maintained in the tanks with an automated oxygen regulator, using an Atlas Scientific galvanic dissolved oxygen (DO) probe (DO controller, Barreleye Designs, Northwood, NH). This functioned as a solenoid controller that bubbled nitrogen into tanks while monitoring temperature and DO every 30 s (DO precision, 0.1 mg L⁻¹). Flowing fresh seawater was constantly introduced into both tanks at a rate of 90 L h⁻¹, and a circulation pump was positioned in each tank to ensure high flow and consistent oxygen throughout each tank.

After the first hour, animal cohorts (two cages with four to six animals each) were removed from each of the hypoxic tanks. They were patted dry, weighed, and placed in 1 of 6 outdoor fully oxygenated 150-L recovery tanks with high-flowing seawater at ambient temperature. Animals were removed from the hypoxic tanks in this way every 3 hours for 24 hours until all individuals in the trial were added to the recovery tanks, at which point the trial ended. Due to limitations in the number of tanks and equipment malfunction, these trials were spread across 3 months; we were nevertheless able to obtain sample sizes of 6–22 animals at each time point in each of the 4 treatments (Table A1).

Survival was determined 24 hours after animals were transferred to their recovery tank. All animals were isolated from each other in the recovery tanks by using mesh dividers. Animals were considered dead when they were not attached to the tank and had no signs of spine or tube foot movement when gently prodded. Most deaths were extremely clear, with many of the dead animals displaying massive spine and tissue loss. Dead animals were removed from recovery tanks periodically throughout the 24-hour post-trial period to avoid impacting other individuals. Living animals were returned to the collection site.

To determine whether there was an effect of prior thermal exposure, temperature, size, exposure time, or their interaction on survival, we constructed generalized linear models (GLMs) with a binomial distribution and log-link function. The full model was analyzed with stepwise removal of non-significant interactions based on P-values (P > 0.1) (Crawley, 2012). The only interaction that was significant was between time and temperature; therefore, the other interactions were removed from the final model. Models were assessed for balance and non-linearity, unequal error variances, and outliers with a residual analysis (Zuur et al., 2010).

We determined what duration of hypoxia exposure caused 5% and 50% mortality (lethal time [LT]; LT₀₅ and LT₅₀, respectively) under each treatment combination. We did this by generating GLMs fitted with a binomial distribution for each treatment. The dose.p function in the MASS library in R was then used to estimate LT₀₅ and LT₅₀ for each model (Ripley et al., 2013). To check whether LT values differed between treatments, a 95% confidence interval was calculated using Student’s t-value. We also calculated LT values for the two temperature treatments (28 °C and 32 °C) without prior heatwave exposure as a factor.

Hypoxia and heatwave occurrences on the reef

To determine the history of past MHWs in the area, we used the 19-year temperature dataset to calculate how frequently and how long MHWs were in both Cayo Agua and Cayo Roldan. Based on the full years of temperature data in both sites, we determined the number of MHW occurrences, their duration, and the total number of MHW days in each year since 2001. These two sites are equidistant from Hospital Point, the collection site for the animals used in the above experiments. Based on other published monitoring data throughout the bay, we expected these sites to represent the far ends of a natural thermal gradient, with Cayo Roldan on the hot end and Cayo Agua on the cool end (Lucey et al., 2020a; Johnson et al., 2021). Therefore, we expected an average of conditions from these sites to be close to what Hospital Point experiences, although long-term monitoring records from Hospital Point would be the only way to verify this accurately.

To determine whether MHWs, hypoxia, and extreme temperatures co-occur, we analyzed a 9-month record from Figuerola et al. (2021), with high-resolution DO and temperature data from a logger deployed on the seafloor at 3 m. The logger was placed in a coral reef habitat in a protected back bay site, Obscura Reef (Fig. 2). Measurements were made every 10 min from March 2018 to November 2018 with a miniDOT logger (DO accuracy ± 5%, Precision Measurement Engineering, Vista, CA). From this March–November 2018 time series record, we documented the days when temperature on the reef was equal to or above our maximum 32 °C temperature threshold, as well as days when the minimum oxygen level was under the average P_crit of E. lucunter (∼4.8 kPa, or 1.5 mg L⁻¹). We determined how many consecutive hours these hypoxic and extreme high temperatures persisted, both alone and in combination. We also documented when MHWs occurred during this nine-month period. All statistical analyses were performed using the statistical software R version 4.2.0 (R Core Team, 2022).

Metabolic rates and critical oxygen limits

There was no effect of the interaction between temperature and body size on standard metabolic rate per unit mass (SMR). However, SMR was lower at 28 °C than at 32 °C for both large and small sea urchins (SMR = 0.69 ± 0.13 at 28 °C compared to 0.78 ± 0.11 at 32 °C, mean ± SD μmol O₂ h⁻¹ g⁻¹; P = 0.03, n = 31). The SMR of larger animals was, on average, lower than that of smaller animals (0.66 ± 0.11 compared to 0.81 ± 0.10, respectively; P = 0.0004, n = 31; Fig. 3a; Table 2). The greatest difference in SMR was at 28 °C, with the mass-specific metabolic rate of small animals 28% faster than the metabolic rate of larger animals. The absolute metabolic rates of these animals increased with body size by a factor of −0.72, for example, the slope of log-transformed SMR by log-transformed wet weight regression analysis (Fig. 3b; Carey and Sigwart, 2014). The scaling factor was slightly but not significantly higher at the warmer temperature (P = 0.31).

There was no effect of size, temperature, or their interaction on P_crit (Fig. 3c; Table 2). The overall average P_crit for the total 31 animals tested was 25.44 ± 6.32 SD% (5.25 ± 1.30 kPa, or 1.58 ± 0.39 mg L⁻¹). Hypoxia treatments in the subsequent experiment were ∼35% below the average P_crit at 16.50% (3.40 kPa, or 1.01 mg L⁻¹).

How temperature impacts hypoxia survival

The GLM analyses with time as a factor show that there was a significant interaction between temperature and time on survival during the hypoxia exposure (P < 0.001). Only 55% of the animals survived 24 hours of hypoxia at 32 °C, while >93% survived 24 hours of hypoxia at 28 °C (Fig. 4; Table 3). Prior thermal exposure, either to MHW or control conditions, had no significant effect on survival during hypoxia exposure (P = 0.560).

Because prior thermal exposure had no significant effect on survival, data were pooled by temperature to determine the time at which the two temperature populations (i.e., 28 °C or 32 °C under hypoxia) experienced either 5% or 50% mortality (LT₀₅ or LT₅₀, respectively). The LT₀₅ of animals exposed to hypoxia and ambient temperature was 19.83 ± 1.28 h (mean ± SE), while the time for the population to reach 50% mortality (LT₅₀) was predicted to be more than twice as long as the 24-h experimental exposure, at 46.15 ± 5.51 h. In contrast, the LT₀₅ and LT₅₀ for animals experiencing hypoxia and extreme temperature were much shorter, with LT₀₅ of 8.71 ± 0.79 h and LT₅₀ just under 24 h (23.67 ± 0.58 h).

Heatwave and hypoxia occurrences on the reef

At least one MHW has occurred in Bahia Almirante every year for the past 19 years. From 2001 to 2019 there have been 60 MHWs (Fig. 5); however, the rarity of these events varies substantially depending on reef location. For example, the number of MHW days on the Cayo Roldan Reef has been 10 times greater than the Cayo Agua Reef. The MHWs on the Cayo Roldan Reef are 5.3 times more frequent (e.g., 3.2 vs. 0.6 occurrences) and 5.8 times longer (e.g., 21.3 vs. 3.7 d) on average than on the Cayo Agua Reef. Site differences in mean temperature and variance reflect these MHW trends; the Cayo Roldan Reef is 0.61 °C warmer and has ∼10% more variation than the Cayo Agua Reef (Fig. 2; Table 1).

We find a positive correlation between the occurrence of MHWs and the manifestation of compound extreme conditions on shallow coral reefs. During the 9-month oxygen-temperature monitoring period on Obscura Reef, temperatures remained above the 90th temperature percentile for 5 or more days 8 distinct times (i.e., 8 MHWs). Compound extreme events consisting of temperatures above 32 °C and oxygen levels below 4.5 kPa occurred 4 times and always during a MHW event (Fig. 4). Hypoxic conditions only ever lasted 10 hours, while temperatures above 32 °C were sustained for over 24 hours. Compound extremes only ever occurred for a maximum of 3 hours.

Heating tolerance

All sea urchins exposed to five days of MHW conditions survived, clearly demonstrating that these MHWs do not kill Echinometra lucunter. Moreover, there was no effect of the 5-day MHW exposure on subsequent survival under extreme compound conditions of hypoxia and heating. This shows both a lack of acclimation, which would result in improved survival, and a lack of synergy, which would result in reduced survival, to MHW exposure. The lack of effect of MHW conditions most likely indicates that E. lucunter has high thermal tolerances. This could be because MHW temperatures today are not as stressful as they were in the past (i.e., shifting baselines in the definition of MHWs) or because these animals live near the surface, where they naturally experience more temperature extremes. Unfortunately, there are few long-term data on the survival of E. lucunter at warmer temperatures. Short-term trials have demonstrated significant thermal tolerance to very extreme shorter exposures in E. lucunter from the Caribbean (Sherman, 2015; Collin et al., 2018; Lucey et al., 2021). Two-hour exposures of E. lucunter compromised survival at 37 °C and above (Collin et al., 2018), while preliminary trials of 12-hour exposures of E. lucunter following the same protocols of Collin et al. (2018) and done during the same time period by the same researchers showed that all 9 individuals assayed at both 34 °C and 36 °C remained alive (i.e., none died) during the 12-hour exposure and for the 24-hour recovery period. However, some of those exposed to 36 °C had sagging spines and other signs of stress after the 24-hour recovery period (Francesco Rendina, STRI, and RC, STRI, unpubl. data, 2016). Not only is E. lucunter able to survive high temperatures, but it also has high thermal performance limits. Echinometra lucunter from the Cayman Islands had thermal performance tolerances ranging from 32 °C to 33.6 °C in the warmer summer conditions when temperatures were highest (∼30.4 °C) (Sherman, 2015). This strongly suggests that E. lucunter has the capacity under normoxia to survive in much more intense short-term warming events than our extreme temperature exposures and potentially in any MHW documented so far in sites it currently inhabits in Bocas del Toro.

Hypoxia tolerance

Just as thermal tolerance is sufficiently high to withstand current thermal extremes under normoxia, hypoxia tolerance is high at ambient temperatures. At oxygen levels 30% below average P_crit (3.8 kPa, or 1.0 mg L⁻¹), it took almost 20 hours to reach LT₀₅. This is twice as long as the longest recorded time on the reef for oxygen to remain so low, that is, 10 hours (Fig. 4). Therefore, hypoxia in the absence of extreme heating is unlikely to induce mass die-offs within a single day on the shallow reefs where these organisms occur.

Echinometra lucunter is not the only sea urchin species that can survive surprisingly long times under hypoxia. Its sister species Echinometra vanbrunti on the Pacific coast of Panama appears to endure cool hypoxic conditions driven by seasonal upwelling; however, the durations for its hypoxia tolerance beyond 2 hours remain untested (Lucey et al., 2021). Strongylocentrotus purpuratus and Mesocentrotus franciscanus can tolerate similar oxygen concentrations of 1.0 mg L⁻¹ for 36 hours at 14 °C to 15 °C (∼2.58 kPa) and have an LT₅₀ of between 48 and 90 hours (Low and Micheli, 2018). Previous work with other strongylocentrotid sea urchins demonstrated that they have effective anaerobic systems that can produce lactic acid, which can be detected in the coelomic fluid, and that alanine can be metabolized under hypoxia (Drozdov and Drozdov, 2015). This suggests that E. lucunter and other sea urchins may use this nonspecific adaptation to withstand extreme hypoxic events (Bookbinder and Shick, 1986; Drozdov and Drozdov, 2015).

Despite this evidence for low mortality under hypoxia at ambient temperatures on the order of a day or two, exposures of this duration may have negative effects on these animals. For example, in Strongylocentrotus nudus significant immune cell mortality was induced by 24-hour exposures to DO of 1.7 mg L⁻¹ and even to some extent 24-hour exposures to 4.7 mg L⁻¹ (Suh et al., 2014). These immune cells, also known as coelomocytes, help sea urchins respond to various types of environmental stressors, such as temperature and UV radiation; and their reduction due to hypoxia is suggested to be a widespread phenomenon among sea urchins. The most hypoxic exposures in this species also resulted in the suppression of gene expression across an array of genes, including RHN helicase and GATA-4/5/6, which are involved in cellular defense mechanisms that might help the individual cope with more moderate hypoxia (Suh et al., 2014). Additionally, S. purpuratus expressed significant sub-lethal effects on fitness-related traits, such as reduced calcification and grazing rates, and more poorly developed gonads after longer-term exposures at 5.5 mg L⁻¹ (Low and Micheli, 2018, 2020). This highlights how reduced oxygen levels could also negatively impact fitness (Vaquer-Sunyer and Duarte, 2008). If E. lucunter experiences similar sublethal effects, it may already be experiencing reduced performance with respect to these other functions despite having robust survival of current short-term extreme events.

Compound extreme events

Extreme heating and hypoxia act synergistically to significantly reduce survival, even though E. lucunter seems resistant to either factor alone. When oxygen conditions are below the critical oxygen limits of the animal (3.8 kPa), and temperatures are at the local maximum (32 °C), survival declines sharply after 8 hours, reaching LT₀₅ in 9 hours and LT₅₀ in 24 hours. Although we do not know the physiological mechanisms underpinning this synergistic response, we can confidently state that a mass die-off of sea urchins would occur if these extremes lasted for a 24-hour period. Two such die-offs of reef fauna have been documented locally in Bocas del Toro (Altieri et al., 2017; Johnson et al., 2021), although they were not where E. lucunter normally lives. The 2010 mortality event impacted reefs around STRI, while the 2017 event affected reefs near Cayo Roldan in the back bay (Fig. 2). Mortality was primarily assessed for corals, and during the 2017 event over 50% of the live coral was lost (Johnson et al., 2021); but many specimens of Echinometra viridis, a sister species of E. lucunter, were also found dead or dying (NL, pers. obs.). These devastating mass mortalities on coral reefs were associated with hypoxia and occurred during two of the longest MHWs on record, for example, a 38-day-long MHW in 2010 and a 48-day-long MHW in 2017. Unfortunately, the oxygen conditions were not well documented during these mortality events. Without high-resolution temporal data to determine how long low oxygen persisted during these MHWs, it is impossible to determine whether the particularly long exposures to high temperatures or the high temperatures in combination with hypoxia were primarily to blame for these mortality events.

While not associated with a mass mortality event, the high-resolution temperature and oxygen dataset we present in this study provides clues on how compound extremes arise on coral reefs. Hypoxic conditions often occurred on the days that temperature exceeded 32 °C. The extremes primarily occurred within the same 24-hour period but not at the same time. This is because temperature and oxygen have inverse diel cycles (Nelson and Altieri, 2019; Lucey et al., 2020a). Water temperature is warmest in the early afternoon after solar radiation has warmed the shallow waters of the bay. Conversely, oxygen is lowest during the night when community respiration draws down oxygen in the absence of photosynthesis. Increasing temperature during the day is associated with increasing hypoxia at night. This highlights the importance of diurnal fluctuations in coastal systems and the need to measure and understand the associated temporal complexities, such as the stressor sequences, and the degree of stressor overlap associated with multi-stressor extreme events at highly resolved spatiotemporal scales (Grégoire et al., 2021; Jackson et al., 2021).

Despite the normal pattern of hypoxia at night and extreme temperatures during the day, we observed short durations when these stressors overlapped or co-occurred in the field (i.e., compound extremes). Interestingly, compound extremes only ever occurred on days coinciding with the occurrence of MHWs. This is best visualized in Figure 6, where every occurrence of compound extremes corresponds with a MHW (i.e., blue-shaded boxes represent days of sustained temperatures above 30.5 °C). The longest duration of compound extremes we observed, 3 hours, is unlikely to cause mortality in E. lucunter. And while there seems to be a clear relationship between the occurrence of MHWs and compound events, the question remains: what exactly will increase the duration of compound extremes to the point at which mortality would be highly likely (e.g., length of MHWs, magnitude of daytime temperature)?

Physiology of extremes

During hot, hypoxic periods it is thought that deoxygenation reduces the supply of oxygen available to meet organisms’ metabolic demands, while warming simultaneously increases these demands (Pörtner and Knust, 2007). This mismatch between oxygen demand and supply could have serious, rapid impacts on the organisms’ ability to survive compound extreme events. This mechanism underlies a physiological breakdown at high temperatures. Examples of this mechanism in the context of hypoxia are scarce, particularly for marine sea urchins. One study provides evidence of temperature-mediated oxygen limitation in the tropical sea urchin Colobocentratus atratus (Wilbur and Moran, 2018). In this species, reduced performance due to increased temperature was improved by adding oxygen to high-temperature treatments. Like most other studies examining temperature effects on sea urchin oxygen consumption (Ulbricht and Pritchard, 1972), we also found that sea urchin metabolic rates increase with warming. In the Australian sea urchin Heliocidaris erythrogramma, +3 °C increases oxygen consumption by up to 35% (Harianto et al., 2021). Delorme and Sewell (2016) attributed increased metabolic rates of the New Zealand sea urchin Evechinus chloroticus to warming in the field during the summer, indicating that these metabolic responses can be sustained for relatively long periods. However, these higher metabolic rates from warming can come at a cost, such as reduced gonad size in the tropical Pacific sea urchin Echinometra sp. (Uthicke et al., 2014).

Metabolic rates are also expected to increase with animal size (Carey and Sigwart, 2014; Carey et al., 2016), and larger animals are expected to be more oxygen limited at higher temperatures than small ones. One leading theory for this is that they have lower surface area-to-volume ratios (Atkinson et al., 2006). This is likely true for E. lucunter because we found that the increase in resting metabolic rates with increasing temperature was significant only in larger animals (Fig. 3a). We also found that larger sea urchins have higher absolute, or whole-animal, metabolic rates than small animals across temperatures, with rates scaling with body mass by a factor of about 0.7 (i.e., b, slope of the regression; Fig. 3b). This closely agrees with another warming study using the sea urchin H. erythrogramma, which found that metabolic rate scaled with body mass to the 0.70–0.72 power (Carey et al., 2016). This highlights how the size of the animal can be a strong determinant for the magnitude of its response to temperature. Together, these results suggest that large adults with the greatest reproduction potential will be both the most susceptible to oxygen limitation and also demand the most oxygen during compound extreme events (Clark et al., 2013).

The critical oxygen limits of organisms, that is, P_crit, should also increase with warming; P_crit is temperature dependent in most marine ectotherms studied to date (Vaquer-Sunyer and Duarte, 2011; Deutsch et al., 2015) and should also decrease with size, with smaller animals expected to be more hypoxia tolerant than larger ones because of their lower O₂ demand. Surprisingly, the P_crit of E. lucunter did not increase significantly with temperature or body size over the range tested. It is possible that P_crit does depend on temperature, but that the intraspecific variation was too large and the range of temperatures examined here too small to statistically detect this difference. Future efforts focusing on temperature and hypoxia tolerance traits together are needed to establish more meaningful biologically relevant definitions of environmental extremes as well as how they scale with size and vary across taxa.

Conclusion

We find that compound events consisting of simultaneous hypoxia and extreme heating occur during MHWs on shallow Caribbean coral reefs. Surprisingly, there is no noticeable impact of prior exposure to these heatwaves on the survival of sea urchins when they are challenged with hypoxia or with hypoxia and extreme heating. Hypoxia does not impact short-term survival on its own either. However, when hypoxia is coupled with extreme heating, over half of the animals die within 24 hours. This illustrates how short durations of compound extreme events can drastically impact coral reef inhabitants. Because our short-term oxygen records show that compound extreme events occur during MHWs, and because long-term temperature records show that there is a 50% chance of reefs experiencing at least one MHW per year, the frequency of such compound events is likely to be either underestimated and/or increasing. This agrees well with the small but growing body of research that unanimously emphasizes the severe biological and ecological impacts of compound extreme ocean events (Ummenhofer et al., 2017; Gruber et al., 2021; Wolf et al., 2022). Increasing efforts to collect high-resolution oxygen and temperature data are necessary, as is knowledge of how multiple extreme stressors interact, how long extreme events last, and how their magnitude, intensity, and recurrences impact marine life. Understanding of the basic anatomy of these compound events, with their corresponding biological impacts, is urgently needed to assess the risk they pose to coastal ecosystems globally.