Composition of cetacean communities worldwide shapes their contribution to ocean nutrient cycling

Abstract

Defecation by large whales is known to fertilise oceans with nutrients, stimulating phytoplankton and ecosystem productivity. However, our current understanding of these processes is limited to a few species, nutrients and ecosystems. Here, we investigate the role of cetacean communities in the worldwide biological cycling of two major nutrients and six trace nutrients. We show that cetaceans release more nutrients in mesotrophic to eutrophic temperate waters than in oligotrophic tropical waters, mirroring patterns of ecosystem productivity. The released nutrient cocktails also vary geographically, driven by the composition of cetacean communities. The roles of small cetaceans, deep diving cetaceans and baleen whales differ quantitatively and functionally, with contributions of small cetaceans and deep divers exceeding those of large whales in some areas. The functional diversity of cetacean communities expands beyond their role as top predators to include their role as active nutrient vectors, which might be equally important to local ecosystem dynamics.

Fig. 1. Nutrient loads released by cetacean communities in 14 contrasted areas show quantitative and qualitative variations around the globe.

Results are from the bioenergetic model supplemented with an original dataset of abundance estimates, diet composition, prey composition and metabolic data. The model was set up with Monte-Carlo simulations combined with a bootstrap procedure with n = 1^e4. Dark grey shaded areas define locations surveyed for population abundance estimates used in the model. In each area (i.e. box), the sizes of the circles are proportional to the order of magnitude of mean absolute estimates in kg/yr/km²; the colour gradient of the circles indicates values of the fold-change ratio of nutrient release compared to the area of reference (French Polynesia, where absolute values are the lowest), i.e. how much more nutrients are released in a given area compared to this baseline one. Shapes in the circles identify primary (triangles) and secondary (squares) limiting nutrients for primary producers in specific areas as taken from Moore et al. (2013), Zhao & Quigg (2014), Drupp et al. (2011), Sonnekus et al. (2017). Vector map adapted from Felipe Menegaz/CC-BY SA 3.0/. Source data are provided as a Source Data file.

Fig. 2. Cetacean communities’ nutrient release correlates with indexes of ecosystem productivity.

Mean estimates of nutrient loads released by cetacean communities in 14 contrasted areas were normalized per nutrient and plotted against two indicators of ecosystem productivity: (a) the mean surface chlorophyll concentration (b) and the mean sea surface temperature estimated from satellite data on https://oceancolor.gsfc.nasa.gov. Mean nutrient release estimates result of a bioenergetic model supplemented with an original dataset of population abundances, diet composition, prey composition and metabolic data and set up with Monte-Carlo simulations combined with a bootstrap procedure with n = 1^e4. French Guyana area was removed for the left plot as the chlorophyll concentration estimate was driven by the water turbidity due to the Amazon River plume. Linear models were run for each nutrient and each slope was statistically significant (see Table 1). Source data are provided as a Source Data file.

Fig. 3. Cetaceans do not release equivalent amounts of nutrients in different habitats, depending on areas.

Differences between mean levels of nutrients released by cetacean communities in neritic and oceanic habitats, with levels of nutrient release per unit area and per year normalized per nutrient and per area. When values are negative (left panel), nutrient release is greater in neritic than oceanic habitats, and when values are positive (right panel), nutrient release is greater in oceanic than in neritic habitats. Habitat differences between nutrient release per unit area and per year estimates were assessed based on unilateral binary relations between estimates (see Methods), and significant differences between habitats are indicated with a black star (test results are provided in Supplementary Data 2). Mean nutrient release estimates result of a bioenergetic model supplemented with an original dataset of population abundances, diet composition, prey composition and metabolic data and set up with Monte-Carlo simulations combined with a bootstrap procedure with n = 1^e4. Green shades are violin plots, indicating the distribution of difference estimates. Source data are provided as a Source Data file.

Fig. 4. Principal component analysis (PCA) reveals distinction between the relative nutrient composition of wastes released by small cetaceans, deep divers and baleen whales.

Individual nutrient released per kilogram of food ingested daily was estimated using a bioenergetic model supplemented with an original dataset of diet composition, prey composition and metabolic data and set up with Monte-Carlo simulations combined with a bootstrap procedure with n = 1^e4, normalized per nutrient and computed for 38 cetacean species belonging to small cetaceans (deep blue ellipse and square points), deep divers (light blue ellipse and triangle points) or baleen whales (red ellipse and circle points). Each point represents a species. The contribution of variables (lowest 2.5% quantile, mean and highest 97.5% quantile, for all nutrients) to the first two principal components are plotted as arrows on the biplot, colour indicates the nutrient. Only variables with cos2 > 0.5 were plotted. Source data are provided as a Source Data file.

Fig. 5. Relative nutrient composition of wastes produced by small cetaceans, deep divers and baleen whales.

Individual nutrient released per kilogram of food ingested daily was normalized per nutrient and computed for 38 cetacean species, as estimated using a bioenergetic model supplemented with an original dataset of diet composition, prey composition and metabolic data and set up with Monte-Carlo simulations combined with a bootstrap procedure with n = 1^e4. For all nutrients except copper (Cu; with our calculated p-value p = 2.2 1^e-2 for comparison with small cetaceans) there is no significant difference between the relative composition of each taxon. Boxplots display the median with a solid black line in each box, lower and upper hinges correspond to the 25th and 75th percentile, respectively; upper and lower whiskers extend respectively from the hinges to the largest and lowest values no further than 1.5 times the inter-quartile range, and data beyond the end of whiskers are not plotted. Source data are provided as a Source Data file.

Fig. 6. Different cetacean taxa show different contributions to nutrient cycling worldwide.

Respective contribution (in %) of baleen whales (red), deep divers (light blue) and small cetaceans (deep blue) to loads of nutrients released by whole cetacean communities in 14 contrasted areas. Results are from the bioenergetic model supplemented with an original dataset of abundance estimates, diet composition, prey composition and metabolic data. The model was set up with Monte-Carlo simulations combined with a bootstrap procedure with n = 1^e4. Vector map adapted from Felipe Menegaz/CC-BY SA 3.0/. Source data are provided as a Source Data file.

Fig. 7. Sensitivity of the bioenergetic model output to uncertainty in model inputs.

Ranges of estimates of Sobol indices (first-order, in blue, and second-order, accounting for interactions between parameters, in red) calculated for each species of cetacean included in the model, in each habitat of each area. BM is the body mass, β is as species-specific metabolic multiplier, E is the mean energy content in the diet, x is the mean nutrient content in the diet, AE is the mean assimilation of energy rate, r is the nutrient release rate, A the population abundance and t the number of days of presence. The bioenergetic model was set up with Monte-Carlo simulations combined with a bootstrap procedure with n = 1^e4. Boxplots display the median with a solid black line in each box, lower and upper hinges correspond to the 25th and 75th percentile, respectively; upper and lower whiskers extend respectively from the hinges to the largest and lowest values no further than 1.5 times the inter-quartile range, and data beyond the end of whiskers are not plotted. Source data are provided as a Source Data file.