Search

There is increasing interest in using deep learning–based object recognition algorithms to automate the labeling of image data collected from marine surveys. However, underwater object detection is a particularly challenging problem due to changes in scattering and absorption of light, and spotty data collection efforts, which rarely capture the broad variability. Using deep learning–based object detection systems for long-term or multisite marine surveying is further complicated by shifting data distributions between training and testing stages. Using data from the 100 Island Challenge, we investigate how object detection performance is impacted by changes in site characteristics and imaging conditions. We demonstrate that the combined use of data augmentation and unsupervised domain adaptation techniques can mitigate performance drops in the presence of domain shift. The proposed methodologies are broadly applicable to observational data sets in marine and terrestrial environments where a single algorithm needs to adapt to and perform comparably across changing conditions.

The article focuses on the importance of resilience and adaptation in the face of climate change, demographic shifts, and political changes, as outlined in the Global Trends Report. The topics covered include the challenges facing coral reefs, the potential for resilience and adaptation in coral ecosystems, and the need for global collaboration to preserve and restore coral reefs for the future.

Predation is ubiquitous on coral reefs. Among the most charismatic group of reef predators are the top predatory fishes, including sharks and large-bodied bony fishes. Despite the threat presented by top predators, data describing their realized effects on reef community structure and functioning are challenging to produce. Many innovative studies have capitalized on natural experimental conditions to explore predator effects on reefs. Gradients in predator density have been created by spatial patterning of fisheries management. Evidence of prey release has been observed across some reefs, namely that potential prey increase in density when predator density is reduced. While such studies search for evidence of prey release among broad groups or guilds of potential prey, a subset of studies have sought evidence of release at finer population levels. We find that some groups of fishes are particularly vulnerable to the effects of predators and more able to capitalize demographically when predator density is reduced. For example, territorial damselfish appear to realize reliable population expansion with the reduction in predator density, likely because their aggressive, defensive behavior makes them distinctly vulnerable to predation. Relatedly, individual fishes that suffer from debilitating conditions, such as heavy parasite loads, appear to realize relatively stronger levels of prey release with reduced predator density. Studying the effects of predators on coral reefs remains a timely pursuit, and we argue that efforts to focus on the specifics of vulnerability to predation among potential prey and other context-specific dimensions of mortality hold promise to expand our knowledge.

Coral bleaching caused by heat stress (warm water) will arguably be the greatest driver of coral reef loss in the coming decades. Understanding how corals have adapted to distinct oceanographic regimes on multiple scales can provide insight into future tolerance and persistence, information critical to directed intervention or targeted protections. The northern Line Islands span a gradient in oceanographic regimes across latitudes, with seawater becoming warmer, fresher, more oligotrophic, and more saturated with aragonite away from the equator. The combination of this regional gradient and island‐scale (local) conditions was used as a natural experiment against which to test how massive Poritescorals respond to these background conditions and episodic heat stress. We found that coral condition, represented by a metric combining tissue thickness, lipids, and calcification, was similar at almost all islands, though there were differences in how corals allocate resources among these biological parameters. Corals at Teraina, the most densely inhabited island, showed evidence for reduced calcification, potentially associated with human impacts and/or outflow of freshwater from the island. In contrast, Poritescorals at Palmyra, a wildlife reserve, exhibited unexpectedly high tissue‐condition metrics for its latitude, suggesting an additional food source, possibly plankton‐rich lagoonal outflow. We did not find a strong relationship between human habitation and the response of Poritescorals to recent heat stress. However, differences in coral calcification rates and energy stores between observed values and those expected based solely on the regional oceanographic gradient, suggests local effects have indirect impacts on these corals.

Ongoing declines in the structure and function of the world’s coral reefs require novel approaches to sustain these ecosystems and the millions of people who depend on them. A presently unexplored approach that draws on theory and practice in human health and rural development is to systematically identify and learn from the ‘outliers’—places where ecosystems are substantially better (‘bright spots’) or worse (‘dark spots’) than expected, given the environmental conditions and socioeconomic drivers they are exposed to. Here we compile data from more than 2,500 reefs worldwide and develop a Bayesian hierarchical model to generate expectations of how standing stocks of reef fish biomass are related to 18 socioeconomic drivers and environmental conditions. We identify 15 bright spots and 35 dark spots among our global survey of coral reefs, defined as sites that have biomass levels more than two standard deviations from expectations. Importantly, bright spots are not simply comprised of remote areas with low fishing pressure; they include localities where human populations and use of ecosystem resources is high, potentially providing insights into how communities have successfully confronted strong drivers of change. Conversely, dark spots are not necessarily the sites with the lowest absolute biomass and even include some remote, uninhabited locations often considered near pristine. We surveyed local experts about social, institutional, and environmental conditions at these sites to reveal that bright spots are characterized by strong sociocultural institutions such as customary taboos and marine tenure, high levels of local engagement in management, high dependence on marine resources, and beneficial environmental conditions such as deep-water refuges. Alternatively, dark spots are characterized by intensive capture and storage technology and a recent history of environmental shocks. Our results suggest that investments in strengthening fisheries governance, particularly aspects such as participation and property rights, could facilitate innovative conservation actions that help communities defy expectations of global reef degradation.

Otolith chemical analyses enable researchers to follow the dispersal pathways of individual fish through time. Given that water masses have spatially or temporally variable chemical signatures (or correlates thereof) and that this variability can be modelled statistically, we have the potential to describe a fish's dispersal history by examining a temporal transect of elemental concentrations throughout the otolith generated from laser ablation inductively coupled plasma mass spectrometry. Statistical analyses tend to focus on temporal trajectories of individual elements or analyse multiple elements at single points in time. We have developed a customised statistical technique allowing detailed exploration of elemental signatures using maximum likelihood methods. The benefit of this approach is the ability to model chronological series of otolith measurements for all sampled fish and all elements simultaneously, while providing explicit treatment of variability in the data. We used data from a Caribbean fish population to compare traditional analysis techniques with this likelihood-based approach, showing their relative capacities to test among alternative hypotheses regarding the dispersal trajectories of individual fish. By incorporating information specific to the species' natural history and to the analytical techniques, we can explore more detailed models of fish movement than were possible using pre-existing approaches.

Ocean warming is causing declines of coral reefs globally, raising critical questions about the potential for corals to adapt. In the central equatorial Pacific, reefs persisting through recurrent El Niño heatwaves hold important clues. Using an 18‐year record of coral cover spanning three major bleaching events, we show that the impact of thermal stress on coral mortality within the Phoenix Islands Protected Area (PIPA) has lessened over time. Disproportionate survival of extreme thermal stress during the 2009–2010 and 2015–2016 heatwaves, relative to that in 2002–2003, suggests that selective mortality through successive heatwaves may help shape coral community responses to future warming. Identifying and facilitating the conditions under which coral survival and recovery can keep pace with rates of warming are essential first steps toward successful stewardship of coral reefs under 21st century climate change. Mass bleaching events caused by warming oceans and intensifying marine heatwaves have killed millions of corals globally. In the central equatorial Pacific, coral reefs experienced three extreme heatwaves within 15 years, providing valuable insights into the mechanisms that could facilitate coral survival under global warming. We examined an 18‐year record of coral cover in the Phoenix Islands Protected Area (PIPA) to track the community response to each event. In the 2002/2003 heatwave, coral communities across PIPA were decimated, with some reefs experiencing near complete mortality. Strong recovery followed and in 2009/2010, a heatwave of similar magnitude caused minimal mortality. The 2015/2016 heatwave was the strongest on record, exposing PIPA's coral communities to twice the thermal stress of 2002/2003. Yet coral mortality during this event was disproportionately low. An examination of 11 environmental metrics that may have modulated the impacts of extreme heat revealed no consistent explanation. Rather, our results support the hypothesis that the survival of thermally tolerant colonies in 2002/2003 and localized recruitment, facilitated recovered communities with greater thermal tolerance than those that inhabited the reefs two decades ago. Understanding and promoting the conditions that facilitate coral recovery and adaptation would enhance our ability to foster coral reef survival. An 18‐year record of coral cover spanning three strong El Niño heatwaves reveals a weakening impact of thermal stress over timeCoral communities recovered from the catastrophic 2002/2003 heatwave suffered lower mortality in 2015/2016 despite a doubling of thermal stressA decoupling of coral mortality and thermal stress suggests repopulation by thermally tolerant survivors may aid adaptation to ocean warming An 18‐year record of coral cover spanning three strong El Niño heatwaves reveals a weakening impact of thermal stress over time Coral communities recovered from the catastrophic 2002/2003 heatwave suffered lower mortality in 2015/2016 despite a doubling of thermal stress A decoupling of coral mortality and thermal stress suggests repopulation by thermally tolerant survivors may aid adaptation to ocean warming

Nonlinear time-series forecasting, or empirical dynamic modelling, has been used extensively in the past two decades as a tool for distinguishing between random temporal behaviour and nonlinear deterministic dynamics. Previous authors have extended nonlinear time-series forecasting to continuous spatial data. Here, we adjust spatial forecasting to handle discrete data and apply the technique to explore the ubiquity of nonlinear determinism in irregular spatial configurations of coral and algal taxa from Palmyra Atoll, a relatively pristine reef in the central Pacific Ocean. We find that the spatial distributions of coral and algal taxa show signs of nonlinear determinism in some locations and that these signals can change through time. We introduce the hypothesis that nonlinear spatial determinism may be a signal of systems in intermediate developmental (i.e. successional) stages, with spatial randomness characterizing early (i.e. recruitment dominated) and late-successional (i.e. ‘climax’ or attractor) phases. Common state-based metrics that sum community response to environmental forcing lack resolution to detect dynamics of (potential) recovery phases; incorporating signal of spatial patterning among sessile taxa holds unique promise to elucidate dynamical characters of complex ecological systems, thereby enhancing study and response efforts.