Skip to main content

Positive species interactions strengthen in a high-CO2 ocean

Cite this dataset

Ferreira, Camilo; Connell, Sean; Goldenberg, Silvan; Nagelkerken, Ivan (2021). Positive species interactions strengthen in a high-CO2 ocean [Dataset]. Dryad.


Negative interactions among species are a major force shaping natural communities and are predicted to strengthen as climate change intensifies. Similarly, positive interactions are anticipated to intensify, and could buffer the consequences of climate-driven disturbances. We used in situ experiments at volcanic CO2 vents within a temperate rocky reef to show that ocean acidification can drive community reorganization through indirect and direct positive pathways. A keystone species, the algal-farming damselfish Parma alboscapularis, enhanced primary productivity through its weeding of algae whose productivity was also boosted by elevated CO2. The accelerated primary productivity translated into increased densities of primary consumers (herbivorous invertebrates), which indirectly supported increased secondary consumers densities (predatory fish) (i.e. strengthening of bottom-up fuelling). However, this keystone species also reduced predatory fish densities through behavioural interference, releasing invertebrate prey from predation pressure and enabling a further boost in prey densities (i.e. weakening of top-down control). We uncover a novel mechanism where a keystone herbivore mediates bottom-up and top-down processes simultaneously to boost populations of a co-existing herbivore, resulting in altered food web interactions and predator populations under future ocean acidification.


Primary productivity

We used 36 plots (18 at vents and 18 at controls) distributed at the same depth range (6–8 m depth) to test farming and CO2 enrichment effects on algal production (calculated as mg of O2 produced per algal biomass, O2 mg.g-1). Eighteen of the 36 plots (9 at vents, 9 at controls) were covered by a cage to exclude feeding and farming by the keystone species, whilst 18 (9 at vents, 9 at controls) were open plots in which the keystone species was allowed to feed and weed. Individual cages and open plots each covered a substratum area of 225 cm2.

The exclusion cages were constructed from reinforced construction metal grid and were covered by wire mesh (12 × 12 mm mesh size), with a dimension of 15 × 15 × 15 cm. The mesh size did not prevent the movement of prey and predators in and out of the experimental plot. All cages and plots were placed in the centre of the farms and fixed to the substratum with 2-mm heavy-duty multi-filament rope. The cages were scrubbed every 12 days.

Turf algal biomass and productivity were sampled from the farmer exclusion experiment, and from random core samples (n = 20) at vent ( n = 5, at the centres; and n = 5, at the boundaries of the farms) and controls vent ( n = 5, at the centres; and n = 5, at the boundaries of the farms) along with the same depth range (6–8 m). The algal standing biomass and primary productivity were measured one month after the deployment of the plots and exclusion cages. Core samples were randomly collected by placing a plastic jar with a 4.25 cm diameter on top of the hard substratum and using a spatula to slice the algae off the substratum in such way that the spatula always covered the open end of the jar to prevent any loss benthic material.

One core (diameter 4.25 cm) of turf algal habitat was sampled from inside each open plot and closed cage (n = 36 in total). Also, core samples from the centres and boundaries of the farms (with and without farming effects – each with n = 5 at vent and control sites, respectively, for a total of 20 samples) were collected. These 20 samples were also used as procedural controls (i.e. were compared to the cages that excluded farming) to assess cage effects (e.g. alteration in water flow or presence of iron) on algal biomass and productivity.

Algal crop productivity was estimated based on oxygen production rates per unit of algal weight (mg O2.g-1) measured on the boat and subjected to the same light regime. Algal mats were placed in air-tight incubation chambers (73 ml) underwater and then taken to the boat. To avoid CO2 desaturation due to photosynthetic activity the chambers were refiled with water of similar pCO2 concentrations as that of the controls and vents, respectively, before the start of the productivity measurements. Baseline respiration was first determined following 30 min dark exposure, followed by net photosynthesis with one hour light exposure (O2 produced = final [O2] – initial [O2]), using an oxygen sensor (Fibox 4, PreSens, Germany). Chambers were slightly agitated every 10 min by turning then upside down four times. For algal standing biomass estimation, the algae from the same cores used for the productivity measurements were oven-dried at 60 °C.

Prey density

The effect of farming and CO2 enrichment on invertebrate prey, i.e. the small gastropod Eatoniella mortoni (height = 1.85 mm; width = 1.13 mm), was measured by comparing their densities in the presence and absence of damselfish at control and vent areas. For the year 2016, prey abundance was measured in the same core samples derived from the exclusion experiment which was used to measure primary productivity (see above) among the plots at vents and controls. In the year 2017, prey samples were also estimated using core samples and the same procedure was used as in the previous year. For 2017, a total of 20 samples core samples were collected, 10 with and 10 without farming effects (n = 5 at vent and control sites, respectively). All gastropods were collected and quantified inside each independent core.

Predator density

The density of the common triplefin (Forsterygion lapillum) was quantified by the same single observer inside and at the border of algal farms during the summer of 2016 and 2017. The density estimation of predators at the farm borders served as a quantification of predator densities in the “absence” of farmers, as the farmers rarely visited these borders and this was only performed to fend off other farmers rather than interact with the benthic predators (triplefins). The predator densities were assessed by taking one photo at the centre of (i.e. inside) and one photo at the border of each farm (n = 2 photo quadrats per farm) inside 0.5 × 0.5 quadrats. Each photo covered an area of approximately 0.25 m2 and was taken at a fixed distance of 50 cm above the substrate. A total of 20 photos were taken during each year at the vents and control (n = 10 farms sampled each year). This reduced number of farms sampled (5 at vent and 5 at control sites each year) was chosen to avoid sampling farms located at the border of the vents where animals might move in and out of the CO2 plumes.

Usage notes

Data was blinded analyzed. Samples labels were randomly assigned in the field. Sample label and local (vent or control site), as well as the photo-quadrat ID, were noted underwater in an underwater paper sheet. These notes were only revised and added to sample labels after all laboratory analyses (productivity measurements) and count been performed (prey and predator abundances).

We were unable to collect data on productivity for the year 2017 (missing values = 20) and to use the farming exclusion plot photos to access predator density (missing values = 36; year 2016b).  

Data files contain all data used in the manuscript preparation and are divided as:

  1. Main_data_set sheet contains data used to construct Figures 1 and 2, as well as to run the analyses shown in Tables S1-S3. Productivity (missing values = 20 samples; year 2017); Predator density (missing values = 36 samples; year 2016b). Column descriptions: (A) sample (sample ID); (B) date (data collection year); (C) treatment (Control vs Vent); (D) Cage.Ex (whether data were collected from a experiment, logical valus, Yes/No);  (E) Caged (if samples were from a experiment logical values of Open and Closed were assing to open plots and caged plots); (F) farmer (position of the sample in relation to the area within the fish farm, logical values of inside or border); (G) prey (invertebrate snail abundance); (H) d.prey (invertebrate snail density; individual/cm2); (I) predator (triplefin density; individual/m2); (J) algal.biomass (total turf biomass represented in grams); and (K) productivity (turf algae productivity expressed in mg O2.g). 
  2. Predator_visual_x_photo sheet was used to draw Figure S3 and perform the analysis which the result is shown in Table S6. Column descriptions: (A) sample (sample ID); (B) obs (visual census method; visual or photo); (C) date (date collection year); (D) treatment (Control vs Vent); and (E) density (triplefin density; individual/m2).
  3. Cage_effect sheet was used to draw Figure S1 and construct the analysis which result is shown in Table S5. It is important to note that Procedural Control samples (column = farming; 5 at vent and 5 at control sites) were only used to test the cage effect and were excluded from all other analyses. (A) sample (sample ID); (B) treatment (Control vs Vent); (C) farming (whether data was a Procedural control or Exclusion cage); and (D) p.mgO2.g (turf algae productivity expressed in mg O2.g).


Australian Research Council Future Fellowships, Award: FT120100183

Australian Research Council Future Fellowships, Award: FT0991953

ARC Discovery, Award: DP150104263

Ministry of Science, Technology and Innovation, Award: 13058134

Australian Research Council Future Fellowships, Award: FT120100183

ARC Discovery, Award: DP150104263