Building heat-resilient Caribbean reefs: integrating thermal thresholds and coral colonies selection in restoration

Experimental design

For each species, samples from 10 colonies were collected across the three reefs except for A. cervicornis, which were collected from the rope structures of the Iberostar in-water nursery located at Coco Reef. Wild colonies were randomly selected, ensuring they were apparently healthy, that the samples collected represented no more than 10% of each colony, and that a minimum distance of five m was maintained between them (Baums et al., 2019). For A. cervicornis, only the apical tips of the ramets were used as experimental replicates, while for P. porites, branches were cut longitudinally and all resulting sections were used. For the remaining species, fragments were collected from the skirt or lower lateral portions of the colonies, where tissue was more accessible. Samples were kept in outdoor tanks (raceways) of the coral laboratory for the night before the start of the experiments. Raceways received water from the same open system as experimental tanks. Each colony was divided using a diamond band saw (Gryphon) into eight fragments (of around five cm²), one per temperature treatment. However, in this phase, different colonies were used as replicates for each temperature, with each temperature treatment having 10 replicates per species. Temperature treatments included one control temperature that corresponded to the maximum monthly mean (MMM) temperature of the area (28.5 °C) (NOAA Coral Reef Watch database) and seven temperatures that ranged from +3 to +9 °C over the MMM. However, as only four tanks could be run at a time, samples were divided into two sets, each of which was used for four temperature treatments over two consecutive days (Fig. 1). For all treatments, temperature ramps started at noon and continued for the following 3 h rising from 28.5 °C to the maximum temperature set for each tank. Maximum temperature was held for 3 h before being gradually reduced to the control value over the course of one hour. This temperature ramping scheme followed that of other experiments aimed at assessing coral bleaching thresholds and physiological responses (Voolstra et al., 2020; Evensen et al., 2021).

Heat stress evaluation

Heat stress response was assessed by three methods. The first method measured photosynthetic efficiency (F_v/F_m) through pulse amplitude modulated (PAM) chlorophyll fluorometry (Junior-PAM, Walz) in dark-acclimated samples (Kitajima & Butler, 1975; Maxwell & Johnson, 2000). Dark acclimation period corresponded to the 30 min following the end of the light cycle which also corresponded to the last 30 min of the temperature ramp (Fig. 1). PAM measurements were taken at this point. F_v/F_m quantifies physiological stress due to photoinhibition of symbionts at a specific moment in time, providing a snapshot of photosynthetic function under heat stress conditions. This metric is considered one of the most quantitative approaches to assess thermal stress in corals and is often used as a comparative standard in stress assays (Cunning et al., 2021; Klepac et al., 2024).

The other two methods evaluated color changes as a proxy of symbiont loss during bleaching: visual scoring of the fragments (modified from Edmunds, Gates & Gleason, 2003; Siebeck et al., 2006) and photographic analysis of chlorophyll content through pixel intensity (Winters et al., 2009). These metrics reflect cumulative bleaching effects throughout the experimental period, which may result from the loss of symbionts, their chlorophyll pigments, or both (Fitt et al., 2001). Bleaching assessments took place the morning after each experimental day, before the end of an experimental cycle. This timing was selected because visible signs of bleaching often develop gradually and become more apparent overnight, as physiological damage progresses after heat exposure (Fitt et al., 2001; Lesser, 2021). Fragments were visually evaluated and assigned a bleaching score from 1 (non-bleached) to 5 (totally bleached). Scores 2, 3, and 4 corresponded to “visually”, “moderately” and “severely” bleached categories, respectively (see Appendix S2: Fig. S2 for reference images of each species and corresponding bleaching score category). To ensure consistency and minimize observer bias, all visual assessments were conducted by the same team member. Fragments were subsequently photographed next to a grayscale reference (Kodak Color Separation Guide and Gray Scale Q-13) for later image calibration and analysis using MATLAB R2019a version 9.6.0.1072779 (The MathWorks Inc., Natick, MA, USA). Image analysis was based on the quantification of pixel intensities from red, green, and blue channels. Red channel pixel intensity was used as a proxy for loss of chlorophyll density as a result of bleaching (Winters et al., 2009), where an increase in the pixel intensity within the red channel would be proportional to an increase in tissue whitening associated with environmentally induced stress (Voolstra et al., 2020).

This multi-metric approach was chosen to provide complementary perspectives on thermal stress responses, capturing both real-time physiological effects and cumulative bleaching outcomes. The inclusion of a rapid visual score, though subjective, allowed for field-applicable insights and provided immediate feedback on the variability of bleaching expression across samples. Equal weight was assigned to the three stress evaluation methods, as each captures distinct and complementary aspects of the bleaching response, enhancing the robustness and applicability of genotype selection under restoration-relevant conditions.

T₅₀ determination

The objective of this first experimental phase was to find the T₅₀ temperature for each species. This temperature may vary depending on the stress assessment method used and therefore a T₅₀ temperature was obtained for each of the three methods in this study. To select the final T₅₀ temperature for the second intraspecific phase, the mean of the T₅₀ of the three methods was calculated and rounded according to the results obtained to ensure that the temperature chosen would elicit variable responses across methods. For both photosynthetic efficiency and pixel intensity, data from the ten colonies were first averaged by treatment. The temperature treatment corresponding to the median of these two data frames was selected as T₅₀. If the median value corresponded to the value between two temperatures, the intermediate temperature was selected. Regarding bleaching visual rank, the T₅₀ temperature corresponded to the treatment that elicited a more variable response among the colonies, resulting in a greater number of different bleaching scores or outcomes. There were some exceptions to this rule in which the temperature in between two treatments was selected. In cases where two adjacent temperatures yielded similar distributions of bleaching scores, or where a treatment showed >50% of fragments at the extremes (e.g., fully bleached or not bleached), intermediate temperatures were selected to better capture the range of variation. This decision rule was applied consistently across species and aimed to identify a temperature that maximized the ability to discriminate among individual responses without inducing uniformly severe or negligible bleaching. Given the applied objective of selecting thermotolerant colonies under short-term stress exposure, this approach prioritized ecological relevance and response variability over rigid parametric thresholds.

Experimental design

In this phase, five of the seven species from phase 1 were studied: D. labyrinthiformis, O. annularis, O. faveolata, M. cavernosa, and P. astreoides. For each species, a new set of samples from 19 to 20 colonies were collected across the three reefs. This time, colonies were not randomly collected but selected from Iberostar’s wild parent colonies repository, which is composed of colonies that have been previously labeled, mapped, photographed, and are periodically monitored. Sample collection and fragmentation were conducted as described in Phase 1. Following the same protocol, samples were kept in outdoor raceways, and experiments began a day after collection. Each colony’s tissue was divided into eight fragments, and assays were carried out in four tanks over two consecutive days. However, only two temperature treatments were used: control (MMM, 28.5 °C) and T₅₀ temperature (variable depending on the species) (Fig. 1). This way, four replicates per treatment and colony were employed. This single-temperature approach, based on the species-specific T₅₀ identified in Phase 1, was chosen to prioritize high replication across colonies and align with restoration-relevant applications, rather than conducting full dose–response assays as in the CBASS framework. Both temperature profiles and heat stress evaluation methods were performed following the same protocol as in the interspecific phase.

Ranking of colonies

Each colony was assigned a thermotolerance score from 1 (most thermotolerant) to 4 (least thermotolerant) based on the three heat stress evaluation methods. Due to the distinct nature of these methods, different scoring procedures were applied, summarized below and explained in detail in Appendix S2.

For photosynthetic efficiency (F_v/F_m), only heated samples were used for thermotolerance scoring, while control samples ensured no additional stress factors. F_v/F_m average values were calculated from heated replicates for each colony. Thermotolerance scores were assigned based on established stress ranges (Evensen et al., 2022): 0.5 ≥ value ≥ 0.7: Score 1; 0.4 ≥ value > 0.5: Score 2; 0.3 ≥ value > 0.4: Score 3; value < 0.3: Score 4. For pixel intensity, control values were used to standardize heated values due to variability in coloration across colonies. This way, average values were calculated for both control and heated treatments, and a ratio was calculated between the heated and control averages; higher ratios indicated greater bleaching. These ratios were divided into quartiles (Q) for thermotolerance scoring: value ≤ Q1: Score 1; Q1 < value ≤ Q2: Score 2; Q2 < value ≤ Q3: Score 3; Q3 < value ≤ Q4: Score 4. For bleaching visual rank, bleaching scores from the four heated replicates were summed for each colony, resulting in totals ranging from 4 (1+1+1+1) to 20 (5+5+5+5). Colonies received thermotolerance scores as follows: 4–8: Score 1 (most thermotolerant); 9–12: Score 2; 13–16: Score 3; 17–20: Score 4 (least thermotolerant). After assigning scores for each method, the scores were summed for each colony, and colonies were ranked from most thermotolerant (lowest total score) to least thermotolerant (highest total score).

It is worth noting that, unlike F_v/F_m and visual bleaching scores, which have recognized stress thresholds, pixel intensity lacks standardized reference values to interpret bleaching severity. For this reason, we calculated relative color retention (compared to control colonies) and used a quartile-based approach to assign thermal scores. While this method ensures a full distribution of scores (1 to 4), it may inherently introduce more variability compared to the other metrics.

Photosynthetic efficiency

Photosynthetic efficiency (F_v/F_m) demonstrated non-stressed optimal values (0.5–0.6) (Evensen et al., 2022) for samples at lower temperatures, and a decline in values became apparent across all species after reaching approximately 35 °C (Fig. 2A). At 36.5 °C (MMM+8), A. cervicornis, M. cavernosa, and Porites spp. still had high F_v/F_m mean values of around 0.4. However, at the highest temperature (37.5 °C; MMM+9), only A. cervicornis, and P. porites mantained high F_v/F_m mean values. When comparing the photosynthetic efficiency between species across all temperatures, significant differences were found (Kruskal–Wallis; χ² (6) = 88.971, p < 2.2e−16), especially driven by A. cervicornis and P. porites, which differed from all other species but not from each other (Appendix S1: Table S2), as both species maintained F_v/F_m values above 0.3 at high temperatures. Further analysis of species differences at each temperature revealed significant effects across all tested temperatures, with most pairwise differences involving A. cervicornis or P. porites (Appendix S1: Table S3). The temperatures at which the highest number of interspecific differences were observed were 35.5 °C (MMM+7) and 36.5 °C (MMM+8). Median values of F_v/F_m (Appendix S1: Table S4) revealed the highest temperature thresholds for M. cavernosa, followed by A. cervicornis, O. faveolata, and P. porites; whereas the lowest threshold corresponded to P. astreoides (Table 1).

Pixel intensity analysis

Red channel pixel intensity differed significanlty between species across temperatures, (Kruskal–Wallis; χ² (6) = 209.96, p < 2.2e−16), with the highest values consistently observed in A. cervicornis, which differed from all other species (Appendix S1: Table S5). As shown in Fig. 2B, A. cervicornis exhibited notably brighter coloration (higher pixel intensity values), in contrast to species such as M. cavernosa, Orbicella spp., and P. porites, which consistently displayed lower intensities across the temperature range. Significant differences were also found between D. labyrinthiformis and both O. annularis and M. cavernosa at all temperatures (Appendix S1: Table S5). Additional species-specific differences were observed at individual temperature levels, most frequently involving A. cervicornis and one or more other species (Appendix S1: Table S3). The temperature that elicited greater differences between species corresponded to 34.5 °C (MMM+6). Median values of pixel intensity (Appendix S1: Table S6) revealed similar thresholds across species around 34 °C (Table 1). The highest threshold value (34.5 °C) was shared by A. cervicornis, M. cavernosa, and both Orbicella spp., whereas the lowest ones corresponded to Porites spp.

Bleaching visual rank

For most species total bleaching was evident at 36.5 °C (MMM+8), except for D. labyrinthiformis which started at 34.5 °C, and P. porites which did not show total bleaching until 37.5 °C (MMM+9) (Fig. 2C). Similarly, temperatures that elicited the most variable bleaching responses across species were 35.5 °C and 36 °C (MMM+7 and +7.5, respectively) except for D. labyrinthiformis (34 °C, MMM+5.5) (Table 1).

Final T₅₀ determination across methods

T₅₀ temperatures varied between species and generally ranged between 34 and 36 °C (Table 1). The T ₅₀ values derived from different methods also showed varying degrees of consistency. For D. labyrinthiformis, thresholds were identical across methods (34 °C), while for A. cervicornis there was a 1 °C difference between one of the proxies (visual bleaching at 35.5 °C) and the other two (both 34.5 °C). Thresholds for M. cavernosa and both Orbicella species ranged within 0 to 1.5 °C across methods. In contrast, P. astreoides and P. porites showed greater variation, with method-derived thresholds ranging across 1.5 °C and up to 3.5 °C, respectively.

The T₅₀ selected for the intraspecific phase of Orbicella spp. and P. astreoides were adjusted from the arithmetic mean of T₅₀ values across methods based on the following rationale: (1) To ensure that variability in visual bleaching rank was captured; (2) At 35.5 °C, broad variation in photosynthetic efficiency responses was still observed for the three species; and (3) pixel intensity values at this temperature remained within a moderate range relative to the overall distribution, with average values of 100.32 ± 16.43 for O. annularis (median = 68), 101.39 ± 16.42 for O. faveolata (median = 86.41), and 112.41 ± 17.99 for P. astreoides (median = 88.36) (Appendix S1: Table S6). These values suggest bleaching was in progress while still allowing for variation across genotypes to be observed.

Principal component analysis demonstrated that photosynthetic efficiency and bleaching visual rank were highly correlated (Spearman’s rank; ρ = 0.8, p = 1.729e−13) whereas pixel intensity and bleaching visual rank were moderately correlated (Spearman’s rank; ρ = 0.6, p = 8.537e−07). No correlation was found between photosynthetic efficiency and pixel intensity (Spearman’s rank; ρ = −0.3, p = 0.0229). The first principal component (PC1) explained 69% of the variance, with bleaching visual rank and photosynthetic efficiency accounting for most of the performance success attributed to the PC1 (factor loadings of 0.67 and 0.59, respectively). The performance of the subjects (one species under a specific temperature treatment) was visually organized into three groups along the x-axis or PC1 (Fig. 3). The best performers on the left of the x-axis mostly corresponded to all species at lower temperatures. In contrast, the worst performers to the right of the x-axis corresponded to some species at the highest temperatures. The middle performers’ group showed some species and temperatures corresponding to the T ₅₀ thresholds selected for the Orbicella spp. and M. cavernosa (Table 1). The second principal component (PC2) corroborated A. cervicornis performance differences with the rest of the species across temperatures and methods, which is reflected as a division into two groups along the y-axis (Fig. 3). A PERMANOVA test confirmed that the performance categories identified visually in the PCA corresponded to statistically distinct multivariate stress responses (F = 13.49, R² = 0.34, p = 0.001).

Thermotolerance of colonies across methods

Orbicella spp. colonies showed a wide distribution of photosynthetic efficiency values: three colonies exhibited the highest thermotolerance (F_v/F_m > 0.5), four showed high thermotolerance (0.4–0.5), two were in the lowest category (F_v/F_m < 0.3), and the remaining colonies had values between 0.3 and 0.4 (Fig. 4A). Montastraea cavernosa, in contrast, showed a narrower distribution: four colonies had F_v/F_m > 0.5, two colonies were in the 0.3–0.4 range, and the majority of the colonies fell between 0.4 and 0.5. The temperatures selected for D. labyrinthiformis and P. astreoides did not lead to notable impairment of the photosynthetic system across colonies, as all individuals exhibitedF_v/F_m values above 0.4. Thermotolerance scores for these species were only distributed across the two highest categories, with colonies showing either high (0.4–0.5) or the highest (>0.5) photosynthetic efficiency (Fig. 4A). Pixel intensity ratios of heated samples relative to their respective controls revealed both the top and bottom 25% of colonies in terms of bleaching severity for all species (Fig. 4B). Porites astreoides exhibited the lowest variability in bleaching visual scores, with all but one colony falling within a single bleaching category (low; scores between 0.3–0.4) (Fig. 4C). In contrast, both Orbicella spp. showed high variability, with colonies more evenly distributed across all four bleaching categories, indicating a broader range of visual bleaching responses.

Comparative and contextual thermal resilience across species

Differences in the heat stress response were observed between species. Montastraea cavernosa and A. cervicornis exhibited the highest T₅₀ value across methods, followed by the Orbicella and Porites spp. On the other hand, D. labyrinthiformis showed the lowest T₅₀ (averaged T₅₀, Table 1). A similar response across methods was observed between Orbicella congenerics. These results are in agreement with other studies and field observations where M. cavernosa showed less susceptibility to heat stress in comparison to Orbicella spp. (Fitt & Warner, 1995; Swain et al., 2017), followed by D. labyrinthiformis and branching Porites spp. (Smith et al., 2013; Alemu & Clement, 2014).

However, this pattern contrasts with previous results for P. astreoides, which is generally considered one of the least affected species by laboratory-induced thermal stress and bleaching events in the region (Warner et al., 2006; Lima, Bursch & Dinsdale, 2022). Porites astreoides represents one of the species with the highest relative cover on Caribbean reefs (Green, Edmunds & Carpenter, 2008), and its abundance has been enhanced by its relatively low susceptibility to coral disease outbreaks, along with congeneric species (Aronson & Precht, 2001; Precht et al., 2016). However, it has also been affected during recent bleaching events (Manzello, Berkelmans & Hendee, 2007; Smith et al., 2019; Edmunds, Didden & Frank, 2021), and its recovery capacity has been limited compared to that of species such as O. faveolata (Grottoli et al., 2014). This limited resilience has been linked to an irreversible decline in carbohydrates and proteins following heat stress, as well as a reduced ability to shift its endosymbionts (Grottoli et al., 2014; Schoepf et al., 2015). The results of this study support predictions that despite their widespread presence on Caribbean reefs, the success of P. astreoides may be compromised in the long term assuming that bleaching events continue to intensify (Levas et al., 2018; Lima, Bursch & Dinsdale, 2022).

In the present study, P. porites exhibited one of the lowest average T₅₀ values (Table 1). However, this value was largely influenced by a low T₅₀ obtained through pixel intensity analysis, which could have underestimated bleaching severity since pixel values increased again at higher temperatures. Moreover, the visual bleaching threshold for this species was one of the highest observed, with no sign of symbiont physiological collapse even at elevated temperatures—levels at which such collapse was evident in more thermotolerant species of the study like M. cavernosa.

Many studies that highlight the greater vulnerability of branching Acropora species compared to massive and encrusting taxa have been conducted in the Red Sea and Indo-Pacific (Loya et al., 2001; Pratchett et al., 2013; Harrison et al., 2019). However, similar patterns have also been reported in the Caribbean (e.g., Langdon et al., 2018; Cramer et al., 2021; Palacio-Castro et al., 2021). In this study, A. cervicornis was among the most thermotolerant species, showing no collapse in the photosynthetic efficiency of its symbionts. This could be the result of strong natural selection among Caribbean genotypes following repeated disease and bleaching events (Aronson & Precht, 2001; Cramer et al., 2020). This is consistent with Manzello, Berkelmans & Hendee (2007), who reported low susceptibility of A. cervicornis in certain sites.

Recent studies in the Indo-Pacific have also suggested shifts in susceptibility patterns. Guest et al. (2016) observed higher tolerance in Acropora and Pocillopora during a bleaching event, whereas Porites species were more affected. Similar findings were reported by Szereday, Voolstra & Amri (2024). Burn et al. (2023) further observed that massive Porites exhibited high bleaching susceptibility in sites with greater overall bleaching severity across the Great Barrier Reef and Coral Sea, while Acropora species were also highly affected under these conditions. However, Porites and small Acropora colonies were identified among the “short-term winners” in Van Woesik et al. (2011), suggesting that bleaching outcomes can vary depending on colony size, morphology and environmental context. Conversely, some Caribbean studies have documented lower heat stress tolerance in A. cervicornis compared to O. faveolata (Langdon et al., 2018). Notably, this study involved a long-term experimental design, highlighting that species may respond differently depending on the type and duration of stress.

Therefore, conclusions about relative susceptibility or resilience should be drawn cautiously, considering the evaluation method and specific environmental context, as responses may also vary between reef sites (Manzello, Berkelmans & Hendee, 2007; McClanahan et al., 2020). Each species may exhibit distinct tolerance and recovery mechanisms (Swain et al., 2016; Thomas et al., 2019; Conti-Jerpe et al., 2020), but this can also differ among individuals and be shaped by environmental conditions (Thomas et al., 2018; Drury & Lirman, 2021; Ruggeri et al., 2024).

Interactions between host genetic and physiological factors, microbial and algal associated communities (Santoro et al., 2025), and other abiotic and biotic factors of a determined environment underscore the complexity of thermotolerance. No data on symbionts or the microbiome was collected for this study, so the potential influence of these factors on the thermotolerance of local individuals remains unknown. However, a recent study by O’Donnell et al. (2024), which analyzed coral populations from the same reefs of the present study, found that the symbiont community composition varied significantly among species of the genus Orbicella, while the symbiont communities of P. astreoides from different reefs were mostly dominated by Symbiodinium with little variation. These findings are consistent with previous literature showing Orbicella spp. hosting diverse Symbiodinium clade B genotypes, often including multiple B1 haplotypes with geographic structuring but relatively consistent distribution across species (Green et al., 2014). Additional studies in Florida have shown that Orbicella spp. can host both Breviolum and Cladocopium, occasionally shifting their symbiont communities under environmental stress (Dennison et al., 2021). In contrast, P. astreoides populations tend to maintain associations with a single dominant Symbiodinium lineage across sites (Kenkel et al., 2013). In this context, the importance of establishing thresholds at a local level and identifying specific thermotolerant colonies for each target species is highlighted (Van Woesik et al., 2011), especially given the strong interplay between host identity, symbiont composition, and thermal sensitivity.

Temperature thresholds

Although species differed in their heat stress responses, threshold temperatures generally fell within a similar range of approximately 34 to 36 °C (MMM + 5.5 to +7.5) across all corals tested under this experimental heat pulse design. This clustering of T₅₀ values is notable, especially considering that ED₅₀ values reported in other studies can vary widely within the same species. These differences are often due to factors like local environmental history, genetic variation, or differences in how the data are analyzed. In contrast, the relatively consistent T₅₀ range found here may reflect the ability of the standardized heat ramp protocol to identify moderate and ecologically meaningful stress points under controlled conditions.These values align with prior studies using comparable temperature profiles, such as Cunning et al. (2021), who demonstrated an ED₅₀ range of 35 to 37 °C for Acropora cervicornis based solely on photosynthetic efficiency, and Klepac et al. (2024), who found an ED₅₀ of 34.3 °C for the same species. Notably, our study identified a similar T₅₀ of 34.5 °C for A. cervicornis, reinforcing the relevance of the approach. All three studies applied heat stress protocols consistent with the CBASS framework, which has become increasingly used globally due to its standardized experimental design (Voolstra et al., 2020; Evensen et al., 2021; Evensen et al., 2022; Evensen et al., 2023).

In some cases, such as for A. cervicornis, Orbicella annularis, and Porites porites, the T₅₀ value preceded a marked decline in bleaching metrics (Fig. 2). This pattern likely resulted from the discrete set of temperatures tested and reflects the applied nature of T₅₀: it was selected to identify a moderate and variable response across genotypes, rather than to determine a precise physiological tipping point. While ED₅₀-based approaches aim to define species-wide tipping points through curve fitting, the T₅₀ approach serves a complementary purpose by emphasizing temperatures that capture intraspecific variation, thereby informing genotype-level selection for restoration.

Importantly, the thresholds identified in this study should not be interpreted as universal for each species. Rather, they represent context-dependent responses shaped by local environmental and biological factors (Fitt et al., 2001; Suggett & Smith, 2020), and may also shift over time due to acclimatization, phenotypic plasticity, or environmental change. Nonetheless, T₅₀ values remain a valuable comparative metric when interpreted within the boundaries of the experimental design. By helping to characterize species- and genotype-specific responses under controlled but ecologically relevant conditions, these thresholds can support improved predictions and decision-making in restoration planning across different geographic and environmental contexts.