Skip to main content
Dryad logo

Global meta-analysis of how marine upwelling affects herbivory

Citation

Sellers, Andrew; Leung, Brian; Torchin, Mark (2021), Global meta-analysis of how marine upwelling affects herbivory, Dryad, Dataset, https://doi.org/10.5061/dryad.9ghx3ffcv

Abstract

Aim: Nutrient subsidies support high primary productivity, increasing herbivore abundance and influencing their top-down control of producers. Wind-driven upwelling events deliver cold nutrient-rich water to coastlines, supporting highly productive marine environments. Results from studies comparing ecological processes across upwelling regimes are mixed: some reveal weaker herbivory in upwelling regions, while others report a positive relationship between upwelling and herbivory. In this synthesis we examine the influence of upwelling on top-down control of producers across the globe.

Location: Global; marine ecosystems.

Time period: 1978-2017.

Major taxa studied: Marine herbivores and algae.

Methods: We used data from herbivory studies focusing specifically on the influence of upwelling activity (upwelling studies), and a broader collection of herbivore exclusion studies dating back four decades. For the upwelling studies we compared herbivore effects between experiments replicated across sites for which upwelling conditions were described by the authors. Meanwhile, for the broader collection of experiments we used externally sourced oceanographic data to characterize upwelling activity, and examined how herbivory changed along a gradient of upwelling activity.

Results: Our results consistently reveal that upwelling weakens herbivore effects on producers. Herbivory was, on average, four times weaker in upwelling sites relative to sites under weak upwelling or downwelling regimes in studies that specifically examined upwelling. The analysis of the broader herbivory literature revealed a similar weakening influence of upwelling on herbivory; however, the effect size was smaller and varied across producer functional groups.

Main conclusions: Nutrient subsidies from upwelling events reduce top-down control by herbivores in coastal ecosystems, however, the negative relationship between upwelling intensity and herbivory is likely the result of a combination of co-occurring processes. First, increased primary production overwhelms consumption by herbivores. Second, cold water reduces herbivore metabolism and activity. Finally, surface currents associated with upwelling activity transport herbivore larvae offshore, decoupling secondary production from herbivory.

Methods

Marine herbivory literature search

               We compiled studies that measured the response of marine producers to the removal or exclusion of herbivores in intertidal and subtidal environments by searching ISI's Web of Science using the following terms: (graz* OR herbiv*) AND (exclud* OR exclus* OR fenc* OR cage* OR remov*) AND (macrophyte* OR alga* OR seagrass* OR eelgrass* OR seaweed*). We also included studies cited by other meta-analyses of marine herbivory (Burkepile & Hay, 2006; Poore et al., 2012). To meet our criteria, authors must have reduced herbivore densities in exclusion treatments via manual removal, by installing cages or fences, or through chemical means such as copper-based paints and pesticides. At the end of each experiment, authors measured producer percent cover, biomass, density, or growth inside exclusion and unmanipulated (control) treatments. Lastly, all studies reported the mean producer abundance inside the exclusion and control plots, and their respective number of replicates and measure of variance.

 

Response variable and moderators

               We calculated herbivore effects as:  where e is the mean producer abundance in the exclusion treatment, and   c is the mean abundance in the control treatment. Thus, yi measures the proportional change resulting from the experimental removal of herbivores, such that yi > 0 when herbivore removal results in an increase in producer abundance relative to the control, and yi < 0 when producer abundance is lower in the exclusion relative to the control. We obtained means and standard deviations from each study. When data were reported as time-series, we used the data from the end of the experiment. If a procedural control was used to test for methodological artifacts, we recorded that treatment’s mean, deviation, and sample size.

               For each experiment, we recorded information regarding habitat type, herbivore type, and the method used to exclude them. If enough taxonomic information was provided, we also classified the producer according to functional groups proposed by Steneck & Dethier (1994). Light availability is a determinant of primary productivity, so we estimated the mean day length (MDL) in hours for the duration of each experiment using the ‘geosphere’ package from the R statistical software, and used it as a covariate in the analysis. Water clarity may also influence light availability, so we obtained data for diffuse attenuation coefficients of the photosynthetically available radiation (KdPAR). The KdPAR provides an indicator of turbidity, and is available through the National Oceanographic and Atmospheric Administration’s (NOAA) portal for remotely sensed oceanographic data (https://coastwatch.pfeg.noaa.gov/erddap/index.html).

               We quantified the intensity and variation of upwelling at each experimental site using the Bakun Upwelling Index data (BUI) obtained from NOAA. The BUI reflects the water flux (cubic meters per second per 100 m of coastline) away from the coast (upwelling; positive values) or towards it (downwelling; negative values). NOAA generates upwelling indices worldwide at 0.5° intervals and a temporal resolution of 6 hours; we obtained the data using the 'xtractomatic' package for R. This index has been used to characterize upwelling activity in previous studies, however, it is unreliable for locations in latitudes below 25°, complex coastlines, and small islands. We therefore excluded experiments that matched those criteria, as well as studies in estuaries to avoid confounding effects from terrestrial processes. To characterize upwelling regimes, including within year variation, we calculated the mean (BUIM) and standard deviation (BUISD) of the six-hourly upwelling indices across two years following the initiation of each experiment.

Usage Notes

We used these data to examine relationships between upwelling activity and the effects of marine herbivores. We used three analytical approaches to assess the robustness of our analysis and consistency of our inferences. First, we focused on the upwelling literature, comparing herbivore effect-sizes between experiments replicated across contrasting upwelling regimes. For the first approach we used the 'upwellStudies.csv' file. Second, we expanded our search beyond the upwelling literature and examined the influence of upwelling intensity on herbivore effects reported by the broader set of studies. For this, we relied on a broader set of published herbivore-exclusion experiments and examined how the strength of herbivory varies along a gradient of upwelling intensity defined by the Bakun Upwelling Index. While this approach allowed us to broaden the scope of our analysis and reduce potential publication bias, it also led us to include other factors that may influence herbivory and were not well represented in the data. Based on a preliminary exploration of the data, we excluded levels of factors for which there was insufficient data to compare across upwelling regimes. This data can be found in the 'allSudies.csv' file. Third, we applied a nearest-neighbor algorithm, using the ‘optmatch’ package for R, to pair individual experiments according to contrasting upwelling regimes (positive and negative BUIM), and equivalent producer functional group, grazer type, habitat type, and region. For each pair, one experiment was associated with positive mean BUIM and the other with negative values, but both experiments within each pair were associated with the same region, habitat type, producer functional group, and grazer type. You can access that data in the 'pairedData.csv' file.

 

 

 

Funding

McGill University