Palmyra Atoll (5.89 °N, 162.08 °W), U.S. Minor Outlying Islands, is located in the Northern Line Islands, central Pacific. Palmyra was designated as a National Wildlife Refuge in 2001 and this protection was further expanded in 2009 as part of the Pacific Remote Islands Marine National Monument. The Atoll was temporarily occupied by the U.S. military during World War II but is currently uninhabited aside from a small field research station. Thus, its reefs are considered quasi-pristine (Sandin et al., 2008) and relatively undisturbed from localized human impacts such as fishing or pollution, yet are still susceptible to global climate change. Palmyra’s benthic communities are dominated by reef-builders such as hard corals and CCA, with remaining surfaces covered by turf algae, macroalgae, soft corals, and other invertebrates (Braun et al., 2009; Williams et al., 2013; Khen et al., 2022).

In September 2009, permanent monitoring plots were established in the two major reef habitats on Palmyra: the wave-exposed fore reef (FR) at 10 m depth and the wave-sheltered reef terrace (RT) at 5 m depth, with four sites per habitat and ten replicate plots (90 cm x 60 cm) per site ( Supplementary Figure 1 ), for a total surveyed area of 21.6 m² at each habitat. Replicate plots were 5 m apart along a 50 m transect perpendicular to shore, marked by stainless steel eye bolts in opposing corners that were secured to the benthos with marine epoxy. At least once a year from 2009 to 2019, usually in the late summer or early fall, plots were photographed by SCUBA divers with a Canon G-series camera attached to a PVC frame that maintained a fixed distance from the substrate. All images were digitized (i.e., manually traced) in Adobe Photoshop (Creative Cloud) to quantify abundance of algal taxa in terms of planar areas or percent cover at the functional group level for CCA and turf, family-level for peyssonnelioids, and genus or species-level for other macroalgae. Algae were identified visually by morphology, and taxa were grouped as either calcareous (CCA, Halimeda spp., Galaxaura rugosa, and Peyssonneliaceae sp.) or fleshy (Avrainvillea sp., Lobophora sp., Dictyosphaeria spp., Caulerpa serrulata, and turf) based on the presence or absence of biogenic calcium carbonate structures. Palmyra’s thermal history was obtained from a revised percentile-based method of estimating Degree Heating Weeks (DHW; Liu et al., 2006) developed by Mollica et al. (2019), which more accurately captures the degree of accumulated thermal stress experienced by central equatorial Pacific reefs than traditional DHW (Fox et al., 2021).

All analyses were conducted in R software version 3.6.3 (R Core Team, 2018). First, using only annual time points taken during the late summer or fall (excluding irregular time points to minimize the effect of seasonal variation), we constructed a non-metric multidimensional scaling (nMDS, via metaMDS in vegan for R; Oksanen et al., 2019) ordination plot visualizing the trajectory of algal community composition through time at each habitat. This nMDS was based on Bray-Curtis dissimilarity measures for square-root-transformed algal percent cover data (Anderson, 2001). We applied a square-root transformation to balance the effect of disproportionately-abundant taxa. We tested the effects of habitat, year, and/or their interaction by conducting a three-way permutational multivariate analysis of variance (PERMANOVA with 9999 permutations via adonis in vegan; Anderson, 2001; Oksanen et al., 2019) on the same Bray-Curtis distance matrix. We did not include site as a nested factor because not all algal taxa were present at each site within a habitat. To identify which algal taxa were the main contributors to differences among habitats, we ran a SIMPER or “similarity percentages” analysis (via simper in vegan; Clarke, 1993; Oksanen et al., 2019).

To test whether percent cover of fleshy or calcareous algae varied by habitat and/or over time (only for consistent annual time points), we ran two-way analyses of variance (ANOVAs) with Type-II sum of squares. Assumptions of normality and homogeneity of variance were checked through visual inspection of the residuals. We did not incorporate repeated measures and instead treated years independently because different algal populations were sampled each year rather than the same individuals. Post-hoc letter groupings were assigned via Tukey’s multiple comparisons using multcomp (Hothorn et al., 2008).

Next we explored possible effects of temperature on a single taxon of interest, Halimeda, through an analysis of covariance (ANCOVA) with Type-II sum of squares. Habitat was considered a fixed factor and temperature (in terms of percentile-based DHW values during the week of sampling) was considered a continuous factor. We also examined the relationship between accumulated thermal stress and Halimeda cover using Pearson’s correlation. To further investigate patterns in abundance for this genus, we plotted its percent cover within each quadrat, by site, over time. Lines were smoothed by locally-weighted regression (i.e., LOESS in ggplot2; Wickham, 2016). Finally, we calculated the difference in mean percent cover of Halimeda by site (with quadrats as replicates) between consecutive years. Two-tailed t-tests were used to determine which sites experienced significant changes not overlapping zero (e.g., an increase or decrease in percent cover one year later).

The benthic algal community on Palmyra’s fore reef was calcifier-dominated compared to the fleshy-dominated reef terrace ( Figure 1 ; Supplementary Figure 1 ). Certain taxa were only present in either habitat: G. rugosa on the reef terrace and Avrainvillea sp. on the fore reef. Across both habitats, the most abundant algal taxa or groups on Palmyra included CCA (exhibiting a percent cover range of 0 to 87.6% of the benthos within a single quadrat, average = 20.2 ± 17.4% SD), turf (percent cover = 0 to 88.3%, average = 16.7 ± 17.6%), and Halimeda (percent cover = 0 to 92.3%, average = 8.4 ± 12.4%). The least abundant algal genera were Avrainvillea (percent cover = 0 to 1.7%, average = 0 ± 0.1%), Dictyosphaeria (percent cover = 0 to 27.1%, average = 0.4 ± 1.7%), and Caulerpa (percent cover = 0 to 46.6%, average = 0.6 ± 3.3%). Distinct yearly trajectories of algal community composition were seen in each habitat ( Figure 2 ).

Benthic algal community composition on Palmyra varied significantly by habitat (p <0.001) and year (p <0.001), with an interaction indicating that habitats changed differently across years (p <0.001; Supplementary Table 1 ). There was more year-to-year variation in algal community composition on the fore reef compared to the reef terrace, particularly after the second thermal anomaly in 2015. However, habitat was a better predictor for algal community composition than year, explaining 11.6% of the variation (R² = 0.116; Supplementary Table 1 ) compared to 4.3%. A SIMPER analysis revealed that the taxa contributing most to habitat differences were CCA, turf algae, and Halimeda ( Supplementary Table 2 ). Calcareous algae (particularly CCA, Halimeda spp., and Peyssonneliaceae sp.) were more abundant on the fore reef whereas fleshy algae (turf, Lobophora sp., C. serrulata, and Dictyosphaeria spp.) were more abundant on the reef terrace.

Overall, CCA were more abundant on the fore reef than the reef terrace, covering 25.3 ± 0.8% (mean ± SE) and 15.4 ± 0.8% of the total benthos, respectively ( Figure 3B ). In contrast, turf algae were more abundant on the reef terrace than the fore reef at 21.3 ± 1.0% and 11.5 ± 0.5% cover, respectively ( Figure 3H ). Between fall 2014 and fall 2015 on the reef terrace, there was a decline in CCA from 20.0 ± 3.2% to 12.7 ± 2.9% and a concomitant rise in turf algae from 19.6 ± 3.5% to 28.6 ± 3.7%; the increase in turf at the time of the second thermal anomaly was seen to a lesser extent on the fore reef. However, by fall 2017, turf and CCA cover were restored to pre-disturbance levels in both habitats. Other algal groups were far less abundant than turf and CCA. Benthic cover of C. serrulata, found almost exclusively on the reef terrace, was highest in the fall of 2010 and 2019 at 3.6 ± 1.5%, but dropped to undetectable levels in fall 2012, 2014, and 2018 ( Figure 3A ). Similarly, also on the reef terrace, Dictyosphaeria spp. (D. cavernosa and D. versluysii) comprised up to 1.5% total cover but were nearly negligible in the fall of 2014, 2015, 2018, and 2019 ( Figure 3C ). The reef terrace had 4.3 ± 0.6% cover of G. rugosa in fall 2019 but was typically around 2.5% ( Figure 3D ). There was consistently higher cover of Lobophora sp. on the reef terrace (5.0 ± 0.5%) compared to the fore reef (1.9 ± 0.2%; Figure 3F ). Cover of Peyssonneliaceae sp., found mainly at the fore reef, was lowest in the fall of 2017 at 1.2 ± 0.3% yet reached up to 10-15% of the benthos every fall between 2011 and 2014 ( Figure 3G ). Avrainvillea sp. was not plotted because it occupied less than 0.01% of the benthos. Halimeda spp. (primarily H. opuntia with minor coverage by H. taenicola and H. fragilis) were more abundant on the fore reef, at 14.2 ± 0.8% cover throughout the time series compared to 4.4 ± 0.3% on the reef terrace ( Figure 3E ).

Throughout the time series, the fore reef had higher cover of calcareous algae than fleshy algae, at 46.5 ± 0.8% (mean ± SE) and 13.4 ± 0.5%, respectively ( Figure 4A ), whereas the reef terrace had similar cover of calcareous and fleshy algae, at 22.1 ± 0.8% and 28.0 ± 0.9%, respectively ( Figure 4B ). Percent cover of fleshy algae varied by habitat (p <0.001) and year (p <0.001) with no significant interaction ( Supplementary Table 3 ). Percent cover of calcareous algae also varied by habitat (p <0.001) and year (p = 0.004), with habitats changing differently over time (p = 0.011). On the reef terrace, the cover of calcareous algae remained consistent through time whereas on the fore reef, calcareous algae were replaced by fleshy algae at the time of the second thermal anomaly in 2015 but re-stabilized by fall 2017. A similar yet less pronounced response was observed on the reef terrace.

Several months after the first thermal anomaly, Halimeda cover dropped from 18.8 ± 3.2% (mean ± SE) in fall 2009 to 5.8 ± 0.8% in spring 2010 on the fore reef and 5.4 ± 1.0% to 2.2 ± 0.4% on the reef terrace ( Figure 3E ). By late summer 2010, Halimeda cover had decreased significantly at four out of eight sites ( Supplementary Table 5 ) but increased in subsequent years. Between fall 2014 and fall 2015, Halimeda cover decreased again at all sites; its cover during the second thermal anomaly was among its lowest throughout the time series, at 4.5 ± 0.9% on the fore reef and 0.7 ± 0.2% on the reef terrace. Regardless of the amount of Halimeda within each quadrat or site, its abundance followed a similar trajectory with sharp declines by 2015, and growth or no change thereafter ( Supplementary Figure 2 ). Between fall 2016 and fall 2017, Halimeda cover increased significantly at six out of eight sites by up to 20% ( Supplementary Table 5 ). Thus, in all cases where significant differences were detected, the sites that changed did so in the same direction. There were significant effects of percentile-based DHW (p = 0.023) and habitat (p <0.001) on Halimeda cover ( Supplementary Table 4 ). A negative relationship between Halimeda cover and accumulated thermal stress was seen ( Figure 5 ), with a linear correlation on the reef terrace (Pearson’s r = -0.65, p = 0.03) but not on the fore reef (Pearson’s r = -0.03, p = 0.92).

Ecological succession and community structure can be shaped by physical forces such as light and sediment transport (Glynn, 1976), irradiance and water motion (Done, 1982), and wave energy (Dollar, 1982). On Palmyra, local environmental factors likely contributed to the spatial variability in benthic algal communities by habitat. The shallower, wave-sheltered reef terrace, which receives more light, solar irradiance (Hamilton et al., 2014), and an influx of nutrients and sediments from the nearby lagoon (Rogers et al., 2017), had a higher relative abundance of turf and other fleshy algae throughout the study ( Figure 1 ; Supplementary Figure 1 ). The fore reef, which is subject to more wave action and water motion (Williams et al., 2013; Hamilton et al., 2014; Gove et al., 2015), had a higher relative abundance of calcareous algae. Calcified crusts such as CCA and peyssonnelioid taxa are resistant to high wave energy, which may explain their dominance at this habitat, as has been seen elsewhere in the tropical Pacific (Page-Albins et al., 2012). Coralline algae can also shed their epithallial cells to prevent fouling by fleshy organisms and reinforce their foundation in wave-exposed habitats (Keats et al., 1997b). Articulated algal morphologies such as Halimeda are more vulnerable to dislodgement by waves (Steneck and Dethier, 1994), but nutrients supplied from upwelling and internal tides on the fore reef (Williams et al., 2018) may have promoted their growth (Smith et al., 2004). While temperature could be expected to differ by habitat, our observations were limited to 10 m depth and upwelling-induced cooling on Palmyra has only been found to occur below 15 m (Fox et al., 2023).

Although we did not quantify herbivore abundance in this study, given that Palmyra has very high fish biomass (Williams et al., 2011; Edwards et al., 2014) and that grazing pressure drives algal succession (Carpenter, 1986; Hixon and Brostoff, 1996), biological factors such as grazing may have further contributed to differences in algal community structure. In our photoquadrat time series, algal turfs often appeared cropped (pers. obs.), indicative of grazing. Herbivores can help control fleshy algal cover (Littler et al., 2006; Burkepile and Hay, 2009) and their presence is associated with higher cover of corals and CCA (Smith et al., 2010). With herbivores now being used as a restoration tool to reverse coral-algal phase shifts on degraded reefs (Mumby, 2014; Ladd and Shantz, 2020), Palmyra exemplifies the role of herbivory in maintaining a “healthy” calcifier-dominated reef. Palmyra’s reef system is dominated by top predators and larger-bodied grazers (e.g., parrotfish and surgeonfish) as opposed to small planktivores or echinoids (Sandin et al., 2008). Hamilton et al. (2014) found that Palmyra’s reef terrace had a higher density of herbivorous fish and higher grazing intensity (in terms of bite rates) than the fore reef. Most herbivorous fish on Palmyra feed preferentially on algal turfs (Hamilton et al., 2014), which are more abundant on the reef terrace (although parrotfish bite scars are also seen frequently on CCA on the fore reef; see Charendoff et al., 2023), suggesting that habitat-specific differences in algal and herbivore assemblages are interrelated.

Our study also provides observational evidence that the calcareous macroalgal genus, Halimeda, may be sensitive to warming. At both habitats on Palmyra, benthic cover of Halimeda was among its lowest in 2015 ( Figure 5 ), when percentile-based DHWs reached a value of 7.76 (or a monthly mean sea surface temperature of 29.8 °C; National Oceanic and Atmospheric Administration’s Coral Reef Watch). Perhaps if temperatures on Palmyra had reached a more extreme upper limit, this would have had a more measurable impact on Halimeda cover across the atoll. It has previously been proposed that Halimeda growth and calcification could benefit from seawater temperatures ranging from 24 to 32 °C, but that temperatures above 34 °C will have consequences that may become lethal at 36 °C (Wei et al., 2020). Other experimental studies have shown that exposure to elevated temperatures can either inhibit (Sinutok et al., 2011) or enhance (Campbell et al., 2016) photosynthetic efficiency, calcification, and growth in Halimeda spp., indicating that results may be context-dependent or species-specific (Schubert et al., 2023). Given their role in both primary and calcium carbonate production on reefs (Rees et al., 2007), and as a preferred food source to many reef fishes (Mantyka and Bellwood, 2007; Hamilton et al., 2014), refining the thermal sensitivity limits of Halimeda by species (while also taking into account accumulated thermal stress) and identifying the mechanisms behind this observed phenomenon will be ecologically relevant in the face of global climate change.