Where is coral reef restoration occurring?

(A) Offshore and urban reefs exposed to minimal climate stress are likely suitable for restoration

Four groups of reefs formed from k-means clustering: reef refugia, stressed reefs, offshore reefs, and urban reefs (Figure 10). These reef types similarly emerged along the first and second axes of the PCoA analysis, which captured 10.24% and 5.13% of variation in reef conditions, respectively. Specifically, the first axis defines a gradient in local stressors and habitat quality, whereby reefs exposed to higher levels of local, human-induced stress score high and those that are exposed to waves and less-frequent high thermal stress events score low along the first axis (Figure 10). The second axis is most associated with climate stress, whereby reefs that are exposed to greater climate stress and occur at shallower depths score higher along this axis.  However, the variation explained by this PCoA was limited and, thus, additional variation among reefs may also depend on variables external to this analysis. 

Each reef type was thus characterized by different levels of climate and local stress, as well as distinct habitat conditions. Clustering and PCoA loadings show that reef refugia are defined by fewer, less-frequent high thermal stress events, as well as more competitive coral species. However, it should be noted that, while reef refugia experience less frequent high thermal stress events, maximum temperatures and the duration over which these maximum temperatures occur (DHW) tend to be high in these areas (Table 5). Stressed reefs are associated with high amounts of sediment, more stress-tolerant coral species, and high maximum temperatures. These reefs are also closer to developed coastal areas. Offshore reefs are characterized by higher wave exposure, greater depths, and lower maximum temperatures. Urban reefs are defined by more coastal development and human population growth in surrounding areas. Unlike stressed reefs, urban reefs are not associated with high climate stress. Reefs where restoration is currently being implemented are characterized by high coastal development, sedimentation, and maximum temperatures, overlapping substantially with stressed and urban reef types. Although the PCoA was applied to all data and variables, we report the top 50 % of scores to simplify the presentation of our results (Figure 10). 

Those sites identified as offshore reefs could be most suitable for restoration, as these reefs are seemingly less likely to experience future climatic and local stress and occur across a variety of reef habitats that facilitate the flushing of sediments, which inhibit coral growth, through wave exposure (Fabricius 2011) and limit bleaching effects at greater depths, as high thermal stress tends to be greater at shallower depths (Baird et al. 2018). Such factors would likely increase the chance of replanted coral survival, which could lead to faster recovery and the establishment of self-sustaining coral populations. While exposed to higher levels of local, human pressure than offshore reefs, urban reefs could also be suitable for restoration activity, as these reefs are likely to experience less climate stress and turbidity than stressed reefs. Restoration could be highly beneficial for urban reefs, as these reefs tend to have lower coral cover (Table 5) but occur in key coastal areas where reef structure is necessary to protect coastlines and harbor fish to feed human populations. Thus, restoration could be used to recover reef structure in key urban reefs. 

Although we predict that offshore reefs would be most suitable for restoration, almost no restoration is occurring in these areas (Table 5). Instead, restoration is seemingly targeting stressed and urban reef types. This may be expected, as offshore reefs tend to have slightly higher abundances of competitive coral species (Figure 10). Competitive corals, such as Acropora palamata (Figure 3), are large, fast-growing species that contribute greatly to the framework structure of coral reefs (Lemoine and Valentine 2012). Thus, restoration practitioners may not be targeting these reefs, as they already appear healthier than those highly stressed reefs. Rather, restoration often targets more degraded reefs and may succeed at urban reefs, which are also likely suitable for restoration. In fact, urban reefs were the group with the largest percentage of sites currently targeted by restoration (Table 5). Restoration may be preferentially occurring in such areas with high human development because these are the regions that have sufficient funds to implement restoration in nearby reefs. Remote reefs are also far more costly and time-consuming to restore than reefs that are easily accessible. 

When compared to urban reefs, almost an equal percentage of stressed reefs are currently being targeted by restoration (Table 5). Stressed reefs are likely particularly poor locations for restoration, as the high average amount of sediment, industrial development, and temperature at these sites (Table 5) suggests that stressed reefs are highly disturbed and near coastal areas with high runoff or large, polluted catchments. In these locations, not only will reefs suffer from bleaching due to high temperatures, but also limited growth, as turbid waters smother corals and inhibit growth (Morgan et al. 2020). In fact, turbid waters have been shown to augment the growth of large algal mats that directly compete with corals (Perry and Larcombe 2003). Stressed reefs are mainly dominated by stress-tolerant corals, which can better withstand these conditions, but, when only these corals survive, reefs will lose substantial diversity and associated ecosystem-level functions (Burman et al. 2012). The survival of restored corals in these locations is likely to be far lower than in areas that endure fewer disturbances.

(B) Current coral reef restoration is occurring at degraded reefs exposed to high stress

The CART analysis of restoration occurrence across reef globally showed that restoration practitioners are mainly targeting reefs exposed to high climatic and local stress (relative error =  0.11, cross-validated error =  0.12, SE = 0.03; data not shown). The first partition between reefs explained 27.37% of variation and showed that reefs with high net primary production (NPP>=4333 mg C/m2/day) are currently being selected for restoration, whereas many reefs with low net primary production (NPP<4333 mg C/m2/day) are not currently being selected for restoration. Among reefs with lower net primary production, those exposed to higher climate stress (max DHW>=7.77 C°-weeks) are more commonly being selected for restoration than reefs exposed to lower climate stress. This subset of reefs is also defined by further thresholds in historical climate stress, future climate stress, coastal development, and human population growth rate, whereby restoration occurs mostly at reefs above threshold values for these metrics (historical climate>=-0.01; future climate>=-0.15; coastal development>= 0.02 ppl; human population growth rate>=0.01 ppl/10km2). These results suggest that practitioners are not currently considering the effects of high climatic and local stress on recovery when selecting sites to restore. Frequent and intense disturbances at restoration sites can limit the survival and growth of replanted corals, impacting restoration success. 

What determines coral reef restoration success?

(C) Fishing pressure, coastal development, coral fragment source, and restored coral morphology determine restoration success

The CART analysis revealed predictors of replanted coral survival, or restoration success (Figure 11). Specifically, reefs were partitioned into restoration sites with high or low survival based on environmental stressors, reef habitat quality, and restoration methods (relative error =  0.08, cross-validated error =  0.14, SE = 0.06). Globally, 25.67% of variation in restoration success was explained by a threshold fishing pressure (market gravity) value of 0.85 ppl/hr2 separating high-survival (% survival = 73) restoration sites (fishing pressure>=0.85 ppl/hr2) from low-survival (% survival = 32.2) restoration sites (fishing pressure<0.85 ppl/hr2). High-survival sites were further partitioned based on coastal development, whereby less developed areas (coastal development<250,000 ppl) had on average higher success (% survival = 78.4) than more developed areas (coastal development>=250,000 ppl), which had on average less success (% survival = 65). The restoration sites in more developed areas were further distinguished based on the sources of coral fragments used in each project. Specifically, restoration projects that nurse and transplant coral fragments during restoration had on average lower percent survival (% survival = 66.1) than projects that replant “corals of opportunity,” which are corals that have been detached from the reef through natural processes or unknown events (% survival = 79.7). Together, the coastal development and coral fragment source thresholds explained 47.49% of variation in restoration success (Figure 11). The restoration sites in less developed areas were also further delineated based on the number of months over which restoration outcomes (i.e. success) were monitored. Restoration projects for which monitoring was conducted over a longer period of time (monitoring months>=10) showed on average lower success (% survival = 34), whereas restoration projects for which monitoring was conducted over a shorter period of time (monitoring months<10) recorded on average greater success (% survival = 71.8). These restoration sites were also partitioned based on sources of coral fragments. The projects that transplanted coral fragments had on average high percent survival (% survival = 82.9), while projects that used methods like sexual reproduction and the translocation of large corals experienced lower survival (% survival = 63.2). Within these sites, the projects that replanted corals with branching morphologies had less success (% survival = 54) than sites that replanted corals with diverse morphologies (% survival = 69). 

These results are somewhat unexpected. In particular, high fishing pressure (market gravity) is known to degrade reefs by removing key large-bodied and herbivorous fish from reef ecosystems, but my findings indicate that restoration is more successful at highly-fished sites. While some fish can damage replanted corals during restoration through predation (Mumby 2009), this finding could also be an artifact of the data, as restoration is generally more prevalent in highly populated areas, which also tend to be more heavily fished. Since restoration is more common in these areas, we may have more outcome (e.g. percent survival) data for heavily-fished reefs, which could result in this observed pattern. In fact, the subsequent partitioning of reefs based on coastal development contradicts any relationship between high local, human-induced stress and restoration success, as restoration is on average more successful in less developed areas in this case. The monitoring month threshold could similarly be an artifact of the data, since restoration monitoring is more commonly short-term, which does not capture the negative effects of disturbances like high thermal stress events that occur over longer time intervals (Boström-Einarsson et al. 2020). This could result in inflated metrics of survival or success at sites monitored less than 10 months. Another restoration design choice seems to be a strong determinant of success: coral fragment source. The corals that are replanted can be sourced different ways, but the most common method is strategically taking coral fragments from a living reef, growing these fragments in a nursery, and then replanting them at a restoration site (Herlan and Lirman 2008). Restoration projects that employed this particular method were separated from the others at different nodes in the regression tree and showed opposing trends, whereby restoration success was lower for sites that used this method in less developed areas and higher for sites that used this method at more developed locations that monitored restoration for less than 10 months. These results could again depend on the data available, since this method is typically much more common than others, like replanting corals of opportunity. The most refined groupings based on coral morphology align with our prediction that restoration will be most successful when a diversity of coral morphologies and functional groups are restored, rather than just a single species, as is most common in restoration currently. Additionally, coral success was lower when only branching corals (like Acropora sp.) were restored. While branching coral are most commonly selected for restoration due to their fast growth rates and high structural complexity, these corals also tend to have more surface area and thinner tissues that can't withstand high environmental stress (Hoogenboom et al. 2017). Thus, these species may not support the long-term recovery of coral reefs, particularly in stressful environments, as they will likely die-off when exposed to disturbance.

Where will coral reef restoration be successful?

(D) Marginally higher restoration success is predicted at urban reefs

The random forest applied to predict restored coral survival across the remaining reefs globally showed the highest percent survival of replanted corals at urban reefs (mean = 72.06%) and the lowest at offshore reefs (mean = 58.96%; Figure 12; Table 5). As percent survival was primarily available for urban and stressed reefs, where restoration is current taking place, these sites had higher predicted success. Although the conditions at urban reefs may be suitable for successful restoration, the higher success predicted for these sites could also be due to the concentration of restoration at urban and stressed reefs. Little to no data on restoration outcomes were available for the other reef types and, thus, more data on restoration at these reefs is required to accurately predict restoration success.

Conclusions

By developing a conservation-action strategy based on reef type, my findings identify urban and offshore reefs where restoration would be suitable based on low environmental stress and habitat quality. While current restoration programs are targeting urban reefs near human population centers, they are also targeting heavily degraded, stressed reefs exposed to high environmental disturbance and may be unresponsive to facilitated recovery. My results also show that restoration benefits from methods that replant a diversity of coral morphologies, rather than just a single morphology. Finally, from this information, I was able to predict restoration success at non-restoration sites or reefs. While success was predicted to be higher at urban reefs, which likely have suitable conditions for restoration, data limitations largely prevented accurate predictions across reef types. Moreover, as coral reef restoration is a relatively novel field, many projects fail to record metrics of success over time periods and spatial scales large enough to capture the effects of disturbances and spatial heterogeneity that occur outside the scope of monitoring. Restoration design is also commonly one-sided, focusing on one particular area, e.g. the Florida Keys, or one particular method, e.g. fragment transplantation, which can influence trends and patterns seen in the data, as well as how well these findings can be extrapolated. 

Nevertheless, in addition to adapting restoration design to account for environmental stress and habitat quality in site selection, my results suggest that programs could also benefit from considering the traits/environmental tolerances of coral species selected for restoration. Results from this work could be used to inform a globally adopted framework for restoration action that establishes key target regions and reefs where restoration can strengthen coral reef resilience under climate change, as well as the most effective methods for restoration success. Additionally, my findings can be directly applied to help guide current and future restoration projects worldwide. In combination with local conservation measures and global action against climate change, effective restoration practices could be key to maintaining key coastal ecosystem processes and functions.