How can fishing affect coral reefs

Coral reef ecosystems support important commercial, recreational, and subsistence fishery resources in the U. S and its territories. Fishing also plays a central social and cultural role in many island and coastal communities, where it is often a critical source of food and income. The impacts from unsustainable fishing on coral reef areas can lead to the depletion of key reef species in many locations.

Such losses often have a ripple effect, not just on the coral reef ecosystems themselves, but also on the local economies that depend on them. Additionally, certain types of fishing gear can inflict serious physical damage to coral reefs, seagrass beds, and other important marine habitats.

Coral reef fisheries, though often relatively small in scale, may have disproportionately large impacts on the ecosystem if conducted unsustainably. Rapid human population growth, increased demand, use of more efficient fishery technologies, and inadequate management and enforcement have led to the depletion of key reef species and habitat damage in many locations.

Unsustainable tourism and coastal development can cause lasting damage on a coral reef. Behaviors such as touching or standing on a coral reef or constructing buildings and roads to close to the shoreline without taking proper precautions can instantly damage a reef that is thousands of years old. Unsustainable boating practices, like anchoring on a reef, can also cause destruction.

However, our work suggests that conserving natural trophic interactions by protecting herbivorous fishes and reducing nutrient pollution may help stabilize coral microbiomes and shield corals against temperature-driven bacterial opportunism and mortality, at least in the near term To simulate the effects of overfishing, nutrient loading or the combination of these stressors, we conducted a 3-year field experiment. Four pairs of 9-m 2 plots were established. One member of each of these pairs was enriched with nitrogen and phosphorous, while the other remained at ambient nutrient levels Supplementary Fig.

Each 9-m 2 plot was delineated into nine 1-m 2 subplots with metal nails driven into the reef at the corners and centre of each plot. The locations of the plots were selected such that initial variation in rugosity and algal cover within each subplot was minimal. Within each plot, two randomly selected subplots were enclosed with herbivore exclosures, while two other random subplots were selected as exclosure controls.

Exclosure controls were fitted with open-topped exclosures. These controls allowed access by herbivorous fishes, but acted as controls for other potential artifacts of the cages.

All exclosures were made of plastic-coated wire mesh with 2. Smaller or juvenile herbivorous fishes are able to enter the exclosures, but these smaller herbivores generally contribute little to overall grazing rates on reefs and have minimal impacts on the algal communities In addition, access by smaller herbivores reflects patterns seen under intensive fishing, in which larger fish species are preferentially harvested while leaving smaller size classes of fish 60 , We scrubbed both exclosures and exclosure controls every 4—6 weeks to remove fouling organisms.

Nutrient pollution was simulated using slow-release fertilizer diffusers applied to each nutrient enrichment plot. Each diffuser was a cm diameter PVC tube, perforated with six 1.

The open ends of the PVC tube were wrapped in fine plastic mesh to keep fertilizer pellets inside, but allow diffusion of soluble nutrients.

PVC enrichment tubes were attached to each metal nail within the 9-m 2 enrichment plots for a total of 25 enrichment tubes per enrichment plot. Nutrients were replaced every 30—40 days to ensure continued delivery of N and P. Previous studies have shown Osmocote delivery using similar methods to be an effective way of enriching water column nutrients in benthic systems for example, ref.

Nitrogen and phosphorus levels were assessed in the water column above each enrichment and control plot as in ref. We also assessed nutrient enrichment efficiency by analysing tissue carbon:nitrogen C:N levels in the common alga Dictyota menstrualis.

The nutrient content of macroalgae such as D. We collected D. Nutrient data from both water and algal tissue for each replicate were averaged across summers for statistical analysis via analysis of variance ANOVA. Divers slowly swam the length of each transect counting individuals of the different herbivorous fishes in the genera Sparisoma , Scarus , Acanthurus and Kyphosus. Fishes were identified to species and their length was estimated to the nearest centimetre.

We used published length:weight relationships 64 to convert fish densities into herbivore biomass. We analysed these data with one-factor ANOVA examining potential differences in herbivore biomass over time.

These quadrats were divided into 49 points, and benthic organisms under each point were identified to species or genus. Algae that are challenging to identify taxonomically under field conditions for example, crustose coralline algae and filamentous algae were classified into algal functional groups. Benthic cover was quantified in June 1 week before treatments were initiated to provide a baseline from which to assess changes in algal abundance and community structure.

No significant differences among treatments in algal abundance could be detected at the beginning of the experiment see initial time points in Fig. Further, during the summer of each year — when algal cover was often at its highest, we also surveyed open areas of reef areas that did not have three-sided exclosure controls within the 9-m 2 plots to assess whether the exclosure controls had any effect on algal abundance or community composition.

We did not detect any differences in algal abundance or community composition between the open unmanipulated areas and exclosure controls Supplementary Data 1. In each picture a ruler or object of known size was placed next to the coral to provide scale. In total, we tracked the fate of individual corals spread across each of the treatments for over 3 years.

The most common corals were Porites porites These corals allowed us to evaluate the impact of the different treatments on coral growth or tissue loss across the time course of the experiment. Further, at each time point we scored each coral for: 1 algal competition as measured by direct contact with algal competitors and the identification of that algal competitor , 2 the presence of overlying sediment on the coral, 3 predation scars from parrotfishes and invertebrate corallivores only the former were observed at appreciable levels , and 4 signs of bleaching or disease.

The primary coral disease observed was DSS see ref. The nested, split-plot design of the experiment was incorporated into the model by nesting replicates of the exclosures and exclosure controls within ambient or nutrient-enriched plots. We analysed cover for important species or functional groups, as well as for overall upright algal cover, which is a proxy for the competitive environment of corals. Upright algal cover included all macroalgae and tall filamentous turf, but excluded crustose coralline algae and short filamentous turf as these two functional groups are relatively benign for corals Per cent coral mortality per treatment and coral tissue loss were analysed using similar mixed models to algal cover.

For growth measures, corals were nested within ambient or enriched plots, but we did not incorporate season as we only analysed change in tissue for corals at the end of the experiment. We calculated tissue loss statistics either excluding or including corals that suffered total colony mortality.

Including these corals in coral growth analyses resulted in non-normal distributions that could not be corrected via transformations. Therefore, we analysed coral growth both excluding the corals that died, which satisfied normality requirements for the analyses, and including the corals that died. To assess the impact of algal competition and parrotfish predation on coral mortality for Porites spp. Coral mucus microbial communities were studied in depth for three coral genera common to the study region and most abundant within the plots: Siderastrea Siderastrea siderea only , Porites and Agaricia.

We used 16S rRNA gene surveys to study the microbial changes in the coral mucus across the course of treatment, focusing especially on whether detectable microbial changes accompanied specific routes to mortality or tissue loss in corals. We focused on mucus communities because these are thought to provide a barrier against invasion by opportunistic pathogens, and can be sampled non-destructively from the same individuals over time. We deemed other methods, such as collecting live coral tissue, too harmful and invasive to the coral for our goal of monitoring the coral microbiome over the long term.

Coral-associated bacteria and archaea were collected using sterile syringe removal of the coral surface mucus layer on SCUBA. In this study, we did not attempt to evaluate changes in Symbiodinium abundance or taxonomy, as it has not yet been established that mucosal abundances of Symbiodinium types reflect abundances within tissues, and destructive tissue sampling would preclude time-series analysis of the same coral colonies. These calculations rely on a climatological baseline for the mean temperature of the warmest month known as the MMM.

Data used spanned —, and excluded study dates. Temperatures above the Temperatures were above In predictions of coral bleaching, accumulation of four DHWs is often associated with minor to moderate bleaching 5. We saw little bleaching within experimental plots, consistent with sub-bleaching levels of thermal stress.

In the laboratory, coral mucus samples were thawed, centrifuged and supernatant decanted. DNA was purified using an organic extraction as previously described After DNA extraction, microbial 16S amplicon libraries were generated using the primers F and R, both with added sequencing adaptors and with Golay barcodes added to the reverse primer. Amplification success was checked on a 1.

Sequence libraries were demultiplexed, and sequences with quality scores less than a mean of 35 were removed. Error-correcting barcodes were used to detect and recover sequences whose barcode sequence had exactly one sequencing error.

Barcode sequences with two or more errors were removed. This OTU-picking protocol clusters all reads, but assigns reference ids to OTUs in greengenes, which can be useful in comparisons across studies. OTUs represented in the overall analysis by only a single count singletons account for a large proportion of noisy reads.

Because our emphasis was overall community trends rather than exploration of the rare biosphere of corals , singleton sequences were removed. We took additional steps to account for aspects of the data set unique to host-associated samples. Because coral mucus can contain some amounts of sloughed tissue, we tested whether coral mitochondria were present in any mucus samples. Similarly, because Symbiodinium and other photosynthetic microbial eukaryotes frequently inhabit coral mucus, chloroplast sequences are frequently observed in microbial diversity surveys of corals.

A table of the coral mitochondrial sequences used is available as Supplementary Data 2 sheet b. We selected the highest most lenient e -value that removed mitochondria, but not related bacteria. This is a phylogenetic measure of community similarity that takes into account organismal abundance and phylogeny Significance was assessed by non-parametric t -tests, each with 1, Monte Carlo permutations permutation is important in this instance to account for the non-independence of distances.

The effect of this procedure is to ask whether different factors increase the dispersion of communities. Mantel tests test for correlation between two distance matrices. For example, a matrix of geographic distances for sample sites might be tested for correlation against a matrix of genetic distances.

The partial Mantel test is an extension that tests for correlation between two distance matrices after accounting for the effects of a third, confounding, distance matrix.

We used permutational Mantel tests to test whether between-sample variation in continuous environmental factors such as thermal stress, temperature or algal cover correlated with differences in the weighted UniFrac distance matrix 67 between coral microbial communities.

When data on hypothesized confounding factor was available, we used partial Mantel tests to test significance after accounting for the confounding parameter. For example, seasonal variation in algal cover might potentially confound the effects of temperature on microbial communities- partial Mantel tests were used to test for such effects.

Values for chao1 and equitability were calculated for each rarified table, and averaged into a single value and compared across categories. Specifically, the Shannon entropy using a base 2 logarithm was divided by the base 2 logarithm of the number of observed OTUs. The effects of treatment, temperature, coral genus and individual coral head on richness and evenness were analysed using linear models in R 3. Random forest analysis is a machine-learning method for supervised classification.

We used random forest analysis Supplementary Data 3 sheets i—j to i determine the conditions under which Synechococcales would be displaced as the most abundant order in the coral microbiome, and ii generate a model predicting whether specific coral colonies would lose tissue using ecological for example, long-term contact with algae from photo series and microbiological data the abundance of different bacterial orders. In both models, the random forest was constructed with 1, decision trees and validated by fold cross-validation.

This analysis resulted in an overall prediction accuracy, a per-category prediction accuracy and a ranking of feature importance, defined as the amount of accuracy gained or lost when a specific feature was not included in the model.

It is worth noting that we employed random forest analysis primarily to provide a summary of the overall predictive power of the data; the internal complexity of the algorithm does not allow inferences to be made about the relationships between variables, beyond their empirical influence on accuracy. For mechanistic insight we rely on other analyses presented here. In the first model, we sought to predict the dominant most abundant order of bacteria in samples given only non-microbial data about the coral colony from which the sample was taken at that point in time.

A total of features were used in the prediction including: herbivore exclusion, nutrient addition, coral species species were converted to binary variables , per cent cover of different macroalgae, turf algae and cyanobacterial mats from quadrat surveys, singly and grouped by functional type for example, total macroalgae , plot and subplot number, measures of instantaneous and weekly average temperatures and salinities, season quarter , and the simultaneous presence of parrotfish bites and nutrient enrichment.

Given these features, we asked the model to predict the dominant order of microorganism in each DNA sample, with our main purpose being to determine whether the conditions under which Synechococcus was dominant were predictable.

No microbiological data were included in the input feature table, so all predictions about the microbiology were based on externally observable features.

In the second model, we sought to test whether data about the competitive environment and microbial community composition of specific coral colonies could predict whether they would lose tissue over the 3-year study period. To predict potential functional consequences of observed changes in microbial taxonomy across treatments, functional profiles for each microbial sample were predicted using the PICRUSt tool This tool uses hidden state prediction, a form of evolutionary modelling closely related to ancestral state reconstruction, to put bounds on genomic copy numbers of each gene family in uncultivated environmental microorganisms, using their position in a reference bacterial phylogeny relative to all bacteria or archaea with sequenced genomes 40 , The resulting metagenomes have been used in recent analyses of human and environmental microbiomes, and have in at least one study been shown to correlate with overall metabolite profiles reviewed in ref.

The accuracy of the method depends on several factors 40 , 68 , but the availability of reasonably related reference genomes is generally the most important. This score is the average branch length between each member of the community OTU and its closest sequenced relative, weighted by abundance.

In these data, NSTI scores ranged from 0. We then tested whether functional changes in coral microbiomes correlated with either temperature increase, extremes of temperature or the abundance of upright algal cover within a subplot Supplementary Fig. The effects of upright algal cover or increasing temperature were tested by Spearman regression against KEGG functional category abundance.

To compare whether the functional categories predicted to change in this field study were broadly consistent with previous laboratory experiments, we compared PICRUSt data with a previous study that exposed corals to several stressors in aquaria and sequenced their metagenomes. We reanalysed data from a previously published experiment 19 that studied the effects of macroalgal contact on coral microbiomes of P.

In that study, macroalgae were placed in direct contact with P. Metadata for coral microbiome samples are available as Supplementary Data 2 sheets c,d , respectively. Short-read amplicon sequence data as fasta and qual files are deposited in the QIITA database, as study id: Other relevant data are available from the authors.

How to cite this article: Zaneveld, J. Overfishing and nutrient pollution interact with temperature to disrupt coral reefs down to microbial scales. Jackson, J. Bruno, J. Regional decline of coral cover in the Indo-Pacific: timing, extent, and subregional comparisons. PLoS One 2 , e Wilkinson, C. Thermal stress and coral cover as drivers of coral disease outbreaks.

PLoS Biol. Eakin, C. Caribbean corals in crisis: record thermal stress, bleaching, and mortality in PLoS One 5 , e Littler, M. Herbivory, nutrients, stochastic events, and relative dominances of benthic indicator groups on coral reefs: a review and recommendations. Sci 38 , — Article Google Scholar. Smith, J. The effects of top-down versus bottom-up control on benthic coral reef community structure. The results of these surveys will be used to assess fishing patterns and provide information to Tabuaeran leaders looking to achieve sustainable harvests.

Because the livelihoods of so many Tabuaerans depend on healthy fisheries, locals are eager to preserve fish numbers, McCauley said.

To engage the next generation of Tabuaerans, researchers taught science classes at local schools three times a week on topics such as reef ecology and genetics. The Stanford team also conducted town hall meetings at every village on the atoll. To broaden the scope of the project, team members have shared their results with Kiribati government officials, who face the twin challenges of geography and poverty.

With a population of about ,, the Republic of Kiribati is one of the least developed countries on Earth, consisting of more than 30 atolls spread across about 1.

Rob Dunbar W. Skip to content Skip to navigation. Stanford Woods Institute for the Environment. Search form Search.

Understanding the effects of fishing on coral reef ecosystems: An interdisciplinary approach. Funding Year:. Region s :.