In October 2012, we sampled communities of organisms on floating plastic debris in the North Pacific Subtropical Gyre, during the Plastics @ SEA North Pacific Expedition aboard the SSV Robert C. Seamans (operated by Sea Education Association, Woods Hole, MA, USA). We sampled 31 plastic debris pieces from 14 locations that covered a distance of 2,300 km, beginning from 622 km off the coast of California to 1,158 km off the coast of Hawai’i (Fig. 1). Plastic debris pieces were haphazardly sampled from the ocean surface, based on daytime deck sightings, ship maneuverability, availability of personnel to process samples and our ability to bring the debris on board for analysis (i.e., pieces >1.5 m diameter could not be brought on board and were excluded from this study). Sampled debris pieces were quickly brought on board using dip nets and placed into tubs to minimize the loss of mobile taxa.

Once each piece of debris was brought aboard, it was visually inspected (using the naked eye) for rafting taxa ≥5 mm long, which were removed using hands and scraping tools (e.g., spatulas, knives, paint scrapers). Mobile taxa and sessile taxa with discrete individuals (e.g., anemones, Lepas barnacles) were counted, while the relative areal cover of all sessile fauna was estimated visually and then converted to absolute measures of surface area, using total surface area measurements for each piece of debris (described below). We considered Lepas barnacles of <1 cm test length to be juveniles and we recorded their areal coverage only, while barnacles larger than this size were considered adults and were counted discretely. When >100 adult Lepas barnacles were present on a piece of debris, we estimated the number of adult barnacles by first removing the barnacles from the debris and then breaking the removed barnacles up into smaller groups of similar volume and barnacle size distribution, which was quite uniform in many cases. We then counted the number of adult barnacles in a single group and multiplied this amount by the total number of groups that comprised the full sample of barnacles that we removed from the debris, making every effort to keep counts consistent and accurate. We photographed and catalogued each novel morphospecies encountered, identified by distinct morphological characteristics and with the assistance of species identification guides. After the cruise, voucher samples of taxa (preserved in 70% ethanol) were identified to the lowest possible taxonomic level (genus or species, in most cases), with the assistance of expert zoologists associated with the Florida Museum of Natural History. Voucher samples are now housed at the Florida Museum of Natural History.

The total surface area of each debris piece was calculated by the same person (precluding observer effects), using meter tape measurements and geometric equations that corresponded to the shape of the debris piece or components thereof (Tables S1 and S2). Because each debris piece was only partially submerged (Table S1) and submerged habitat is required by marine organisms, we noted the orientation at the water surface and the locations of water lines and biofilms for each piece of debris to determine the ‘submerged surface area’. We used this as our primary predictor variable for habitat area. In addition, we examined a second predictor variable that provided a measure of habitat area that was independent of barnacle effects, by subtracting barnacle cover from submerged surface area, yielding a measure of ‘open surface area’, available to non-barnacle sessile colonizers. Once all living material was removed, each plastic debris piece was soaked in bleach, stored in the ship’s cargo hold, and either recycled or donated upon completion of the cruise.

We used linear regression models run in the program R 3.1.141 to analyze main effects and interactions of debris surface area and Lepas barnacle abundance on the number of both sessile and mobile taxa. We log-transformed (using a + 1 correction, when necessary) both our predictor variables, to improve coverage and visualization, and our response variables, to meet model assumptions of normality and homoscedasticity. We used the variance inflation factor (in the car package42) to assess the effect of co-linearity between our two predictor variables, for each of which we calculated standardized partial regression coefficients (i.e., beta weights) to quantify their relative importance in multiple regression models. We used the corrected Akaike Information Criterion (AICc) to compare the relative fit of linear models with (1) only an intercept (null model), (2) debris surface area as a predictor, (3) barnacle abundance as a predictor, (4) both debris surface area and barnacle abundance as predictors, and (5) the interaction between debris surface area and barnacle abundance. We considered models to be unequal in their fit to the data if they differed in AICc (ΔAICc) by 2 or more units43. Adjustments for sampling effort (e.g., rarefaction) were unnecessary in our dataset, because our sampling effort scaled with debris size (i.e., we surveyed the entire surface of each debris piece).