Global Change Biology Global Change Biology (2014) 20, 2712–2724, doi: 10.1111/gcb.12644
Macroalgal blooms alter community structure and primary productivity in marine ecosystems DEVIN A. LYONS1, CHRISTOS ARVANITIDIS2, ANDREW J. BLIGHT3, EVA CHATZINIKOLAOU2, TAMAR GUY-HAIM4, JONNE KOTTA5, HELEN ORAV-KOTTA5, 6 , G I L R I L O V 4 , P A U L J . S O M E R F I E L D 6 and T A S M A N P . C R O W E 1 A N A M . Q U E I R OS 1 University College Dublin, Dublin, Ireland, 2Hellenic Centre for Marine Research, Herakilon, Greece, 3University of St. Andrews, St. Andrews, UK, 4Israel Oceanographic and Limnological Research, Haifa, Israel, 5Estonian Marine Institute, University of Tartu, Tartu, Estonia, 6Plymouth Marine Laboratory, Plymouth, UK
Abstract Eutrophication, coupled with loss of herbivory due to habitat degradation and overharvesting, has increased the frequency and severity of macroalgal blooms worldwide. Macroalgal blooms interfere with human activities in coastal areas, and sometimes necessitate costly algal removal programmes. They also have many detrimental effects on marine and estuarine ecosystems, including induction of hypoxia, release of toxic hydrogen sulphide into the sediments and atmosphere, and the loss of ecologically and economically important species. However, macroalgal blooms can also increase habitat complexity, provide organisms with food and shelter, and reduce other problems associated with eutrophication. These contrasting effects make their overall ecological impacts unclear. We conducted a systematic review and meta-analysis to estimate the overall effects of macroalgal blooms on several key measures of ecosystem structure and functioning in marine ecosystems. We also evaluated some of the ecological and methodological factors that might explain the highly variable effects observed in different studies. Averaged across all studies, macroalgal blooms had negative effects on the abundance and species richness of marine organisms, but blooms by different algal taxa had different consequences, ranging from strong negative to strong positive effects. Blooms’ effects on species richness also depended on the habitat where they occurred, with the strongest negative effects seen in sandy or muddy subtidal habitats and in the rocky intertidal. Invertebrate communities also appeared to be particularly sensitive to blooms, suffering reductions in their abundance, species richness, and diversity. The total net primary productivity, gross primary productivity, and respiration of benthic ecosystems were higher during macroalgal blooms, but blooms had negative effects on the productivity and respiration of other organisms. These results suggest that, in addition to their direct social and economic costs, macroalgal blooms have ecological effects that may alter their capacity to deliver important ecosystem services. Keywords: biodiversity, ecosystem functioning, green tide, harmful algal bloom, macroalgal bloom, macroalgal mat, species richness Received 17 October 2013; revised version received 14 April 2014 and accepted 2 May 2014
Introduction Marine and estuarine ecosystems are under pressure from a wide variety of anthropogenic stressors, which have degraded their environmental quality (Lotze et al., 2006; Halpern et al., 2008). This ongoing degradation is commonly believed to have contributed to an increase in the frequency, magnitude, and extent of outbreaks by marine diseases, gelatinous zooplankton, and harmful algae (Harvell et al., 1999; Anderson, 2009; Brotz et al., 2012; Duarte et al., 2013). Understanding how ecosystems respond to these outbreaks is an important step in fully accounting for the social, economic, and ecological costs and benefits of the factors that drive Correspondence: Devin A. Lyons, tel. +780 680-1085, fax +780 492-9234, e-mail: [email protected]
2712
outbreaks, as well as the management actions that might be used to deal with them. In this study, we attempt to facilitate this process by evaluating the ecological effects of one type of species outbreak: macroalgal blooms. Macroalgal blooms are increasing worldwide, and are feared to have a number of undesirable effects on marine and estuarine ecosystems (Teichberg et al., 2012). Expanding human population, fertilizer use, livestock waste production, and fossil fuel combustion have overcome nitrogen and phosphorus limitation of coastal and estuarine waters, leading to accelerated nutrient uptake, faster growth, and more frequent blooms by macroalgae around the world (Nixon, 1995; Conley, 1999; Howarth, 2008; Teichberg et al., 2010). These changes have been aggravated by the loss of herbivores in some ecosystems. For example, harvesting of © 2014 John Wiley & Sons Ltd
E C O L O G I C A L E F F E C T S O F M A C R O A L G A L B L O O M S 2713 large predatory fish in the Baltic Sea appears to have increased macroalgal abundance via a trophic cascade (Eriksson et al., 2009; Sieben et al., 2011). With reduced predatory pressure of large fish, the abundance of smaller predators has increased, driving down populations of invertebrate grazers, and allowing macroalgae to proliferate. While both eutrophication and loss of herbivory increase macroalgal biomass, they also interact synergistically to enhance each other’s effect (Burkepile & Hay, 2006). Opportunistic macroalgae proliferate as a result, forming dense canopies and creating mats of algae that float in the water column or settle on the substrate. Often macroalgal blooms are confined to relatively small areas within a single estuary or embayment, but they can become very large. Massive blooms of Ulva spp. have occurred annually in the Yellow Sea since 2007, creating floating mats that covered up to 3489 square kilometres of its surface (Liu et al., 2013). Regardless of whether they are relatively small or very large, macroalgal blooms can have a number of detrimental social, economic, and ecological consequences. Macroalgal blooms inhibit recreation, diminish aesthetic enjoyment of the coastal zone, and interfere with tourism, fishing and mariculture (Charlier & Lonhienne, 1996; Dion & Bozec, 1996; de Leo et al., 2002). Toxic gases are emitted from mats of rotting macroalgae, posing a potential risk to human health (Chrisafis, 2009; Samuel, 2011). Proliferating macroalgae often alter the biological composition and ecological processes of the ecosystems they affect (Fletcher, 1996; Raffaelli et al., 1998). These undesirable effects have caused considerable concern and prompted costly algal removal programmes in some affected areas (Morand & Merceron, 2005). The negative ecological effects of macroalgal blooms occur via their alteration of the physical and chemical environment and their interactions with other species. When mats of blooming macroalgae settle on the seafloor, they can alter environmental conditions by inducing anoxia and the release of hydrogen sulphide from the sediments, resulting in reductions of species richness and community abundance (Gamenick et al., 1996; Wetzel et al., 2002). The physiological stress these changes create can also elicit sublethal responses from benthic fauna. Mobile burrowers typically respond by repositioning themselves towards the sediment–water interface (Wright et al., 2010). This may alter their susceptibility to predation and can also impact important sedimentary ecosystem processes that these species mediate (Raffaelli et al., 1998; Gribben et al., 2009; Queir os et al., 2011). The competitive effects of macroalgal blooms have been implicated in the worldwide decline in seagrass beds (Valiela et al., 1997; McGlathery, 2001), and linked to local declines in nonblooming algae © 2014 John Wiley & Sons Ltd, Global Change Biology, 20, 2712–2724
(Kautsky et al., 1986; Worm et al., 1999). By driving declines in such foundation species, macroalgal blooms may also have negative indirect effects on other members of the biological community. Some macroalgal bloom species produce chemicals that inhibit the growth, development, and feeding of other species (van Alstyne & Houser, 2003; Nelson et al., 2003; Nan et al., 2008), suggesting that macroalgal toxins may contribute to blooms’ effects. Much of the literature on macroalgal blooms focuses on their negative effects on individual species, communities, and ecological processes, but macroalgal blooms also have several effects that may be considered beneficial. Macroalgal blooms increase transfer of nutrients from the water column to the sediments and other macroalgae, thereby reducing nutrient levels in eutrophic waters (Thybo-Christesen et al., 1993; Hardison et al., 2010). Accumulations of macroalgae can also increase habitat complexity, enhance dispersal of other species, and provide animals with food and shelter (Wilson et al., 1990; Holmquist, 1994, 1997). As a result, macroalgal blooms may actually enhance, rather than reduce, biodiversity and secondary production in some ecosystems (Holmquist, 1997; Bolam & Fernandes, 2002; Dolbeth et al., 2003). Several qualitative reviews of macroalgal blooms’ ecological impacts have been published in the past (e.g. Fletcher, 1996; Raffaelli et al., 1998). However, elucidating the net impact of blooms’ varied positive and negative effects remains an important challenge. Here we provide a quantitative synthesis that complements previous reviews by estimating the overall impact of macroalgal blooms on seven key measures of ecosystem structure and functioning, using meta-analysis. We were interested in investigating community- and ecosystem-level effects of blooms, and focused on how they affected species richness, community diversity (Shannon Index), community evenness (Pielou Index), community abundance (biomass, density, cover), gross primary productivity, net primary productivity, and community respiration in temperate and subtropical ecosystems. In addition, we investigated the effects of the algae responsible for the bloom, the habitat the bloom occurs in, the type of community the bloom affects, and the methodology used to study the bloom as potential drivers of variation in the responses observed in different studies.
Materials and methods We carried out this synthesis using the process of ‘systematic review’, following guidelines recommended by the Centre For Evidence-Based Conservation (2010). Our full, peer-reviewed protocol for conducting this review is available in a separate
2714 D . A . L Y O N S et al. publication (Lyons et al., 2012). Below we provide a brief description of our methods, and note a number of amendments made to the protocol during the review process.
Search strategy and screening of publications We searched online databases to find publications (i.e. research articles) that could be used to evaluate the effects of macroalgal blooms on the structure and functioning of marine and estuarine ecosystems. Specifically, we searched Web Of Science and Scopus databases on 28 June 2012, using a complex set of search terms that included keywords related to macroalgal blooms, macroalgal mats, blooming algal taxa, and a broad range of ecosystem structure and functioning measures (Data S1). Following our search, we assessed the publications in a three-stage screening process designed to identify relevant studies according to a set of predefined inclusion criteria (described below). First, one reviewer assessed the titles to remove any publications dealing with completely unrelated topics. Second, eight different reviewers assessed the abstracts of the remaining publications, with each reviewer evaluating an independent subset of the articles. Finally, each reviewer assessed the full text of the articles that they had accepted. To ensure consistent application of the inclusion criteria, all reviewers assessed two samples of the abstracts before the publications were distributed. We assessed consistency among reviewers with Fleiss’ Kappa statistic (Fleiss, 1971), using a Kappa of at least 0.5 to indicate an acceptable degree of consistency. The assessment of the first sample was not acceptable (Kappa = 0.27), so we discussed inconsistent assessments and clarified how the inclusion criteria were to be applied. This resulted in a sufficient degree of consistency in the second assessment (Kappa = 0.57), but we discussed the inconsistent assessments to further clarify the inclusion criteria and improve performance.
Inclusion criteria To be retained at each stage, a publication had to include at least one study (i.e. individual experiment or observational study) that met a number of criteria. The study had to: (i) examine a temperate or subtropical ecosystem affected by marine or estuarine macroalgal blooms or macroalgal mats; (ii) involve a comparison between an affected state (with a macroalgal bloom or mat) and a control state (without macroalgal bloom or mat); (iii) report on a measure of ecosystem structure or functioning; (iv) report the results of a manipulative experiment or observational study; and (v) include sufficient information so that an effect size (Hedges’ g) could be calculated. We did not define blooms or mats according to particular thresholds in their size or severity, in part because publications frequently neglected to report on the density or spatial extent of macroalgae in their study or used incomparable metrics. Rather, we considered something a bloom or mat if the authors of a publication referred to them as such, or if there was an indication that an accumulation of algae was
either unusually large or caused by anthropogenic factors such as eutrophication. For the first two steps in the screening process, we retained studies that potentially reported on any measurement of ecosystem structure or functioning. When we screened the full text of publications, we elected to include only studies examining species richness, community diversity (Shannon index), community evenness (Pielou Index), and communitylevel measures of gross primary productivity, net primary productivity, respiration, and biomass, density, or cover (‘community abundance’ hereafter). We selected these outcomes because we felt that they were among the most important community- or ecosystem-level indicators for the structure and functioning of marine ecosystems. In addition, sufficient data to conduct a meta-analysis for these outcomes were available. We also searched for data on communitylevel secondary productivity but failed to find appropriate studies. We only included experimental studies if algal abundance was directly manipulated. If macroalgal abundance was not directly manipulated, but an experimental manipulation of some other factor (e.g. nutrients, grazer abundance) resulted in a macroalgal bloom in some replicates we did not include the study. To calculate effect sizes (Hedge’s g), means, sample sizes, and either standard deviations, standard errors, or confidence intervals were necessary. When incomplete information was available, we attempted to obtain the missing information from the authors of the publication prior to excluding it. If we discovered that more than one publication reported the results of the same study (i.e. made the same comparison using the same data), we chose the publication that presented the data for the comparison most clearly, or we randomly selected one of the publications.
Data extraction and meta-analyses When possible, we extracted means, standard errors, standard deviations and sample sizes directly from tables and the text of the articles. In other cases, we used ImageJ (Schneider et al., 2012), DataThief (Tummers, 2006), or Engauge Digitizer (Mitchell, 2010) software to extract them from figures. The extracted data were used to calculate an effect size, Hedges’ g, for each study. Hedges’ g is the bias-corrected mean difference in the response variable between the ‘treatment’ (i.e. algal bloom) and control conditions, standardized by the withingroup standard deviation. Some publications reported results for more than one outcome variable. The data for our seven outcome variables were analysed in separate meta-analyses (see below), allowing us to treat different outcomes from the same publication as independent studies. Other publications contained multiple results for the same outcome variable. How these data were treated depended on how they were generated and structured. A detailed account of how we dealt with these data, and how we estimated the effect size and variance for studies with different designs, is available in Data S1. We synthesized the data for the effects of macroalgal blooms using a random effects model to estimate the overall effect of macroalgal blooms on each of our outcome variables.
© 2014 John Wiley & Sons Ltd, Global Change Biology, 20, 2712–2724
E C O L O G I C A L E F F E C T S O F M A C R O A L G A L B L O O M S 2715 These analyses were conducted in R (R Development Core Team, 2012), using the ‘metafor’ package (Viechtbauer, 2010), and the DerSimonian–Laird estimator (Dersimonian & Laird, 1986). We used Cochran’s Q-test to assess study heterogeneity. We used funnel plots and a rank correlation test for funnel plot asymmetry to assess whether there was any evidence of publication bias or other ‘small study effects’.
Subgroup analyses Study-level covariates may contribute to variation in the effects estimated in different studies. We carried out subgroup analyses using mixed effects meta-analysis to investigate the potential influence of algal identity (i.e. the type of alga responsible for the bloom), habitat, community type (e.g. macroalgae, invertebrates, bacteria), measurement methods, study type (experiment vs. observational) and study setting (lab vs. field) on the estimated effects of macroalgal blooms. The importance of each modifier variable was examined in a separate analysis, using the potential modifier as a fixed effect in the mixed model. The significance of modifier variables was assessed using a Q-test. In addition to determining whether macroalgal blooms’ effects differed among subgroups, these analyses allowed us to evaluate the significance of macroalgal blooms’ effects for the individual subgroups within each analysis. We assessed the significance of individual subgroups using the mean effect and 95% confidence interval estimated by the model. The number of studies we found for some subgroups was low. As a result, the power of some of our subgroup analyses may be low, and cause us to underestimate the importance of the modifying variable. Moreover, the effect sizes estimated from a small number of studies should be viewed with caution. We carried out the subgroup analyses for all outcome variables, as long the modifier variable was relevant to the outcome and sufficient data were available (i.e. one or more studies in each of two or more subgroups). The ‘measurement methods’ subgroup analysis for community abundance compared the results of studies that used either biomass, density, or cover as measures of abundance. For net primary productivity and respiration, it compared studies that either measured total productivity/respiration or the productivity/ respiration of the rest of the affected community (i.e. excluding the productivity or respiration of the algal bloom). We expected that the blooms would increase total productivity and respiration when their contribution was taken into account, and have a negative effect when only the rest of the community was considered. We did not conduct a similar subgroup analysis for gross primary productivity because all studies measured total productivity, including that of the bloom.
Results
Description of the data sets Our bibliographic search returned 906 publications. We retained 638 of these after screening the titles, 298 after © 2014 John Wiley & Sons Ltd, Global Change Biology, 20, 2712–2724
assessing the abstracts, and 52 after screening the full text (Data S1). The majority of publications contained multiple studies of a single outcome or studies of more than one outcome, allowing us to estimate effect sizes for a total 153 studies. These studies were conducted on coasts of all continents except Antarctica, but more than 70% were from European waters (Fig. 1). The number of studies for each outcome variable ranged from 6 to 76 (Fig. 2). The datasets for all outcomes were biased towards studies of blooms by ephemeral green algae (e.g. Ulva, Cladophora), studies conducted in soft sediment (sand or mud) and seagrass bed habitats, and studies of invertebrate communities. The number of studies in the modifier variable subgroups for each outcome is presented in Figs 3, 4 and Figures S3.1–S3.4 in Data S1.
Overall effects Averaged across all studies, macroalgal blooms had significant negative effects on the abundance and species richness of organisms in marine and estuarine communities, and a significant positive effect on gross primary productivity (Fig. 2; Table 1). On average, blooms led to increases in net primary productivity and respiration, and a decrease in community diversity, though these effects were not statistically significant. Macroalgal blooms appeared to have very little effect on community evenness. Funnel plots and rank correlation tests suggested that publication bias did not affect our effect size estimates (Data S1). All seven of the outcome variables displayed significant among-study heterogeneity (Table 1; Figures S3.5–S3.11 in Data S1), so we investigated potential sources of this heterogeneity in a series of subgroup analyses.
Subgroup analyses: impacts of blooms by different taxa The identity of the seaweed(s) responsible for a bloom had a significant effect in the subgroup analyses of blooms’ effects on community abundance and species richness, but did not have a significant influence on the other five outcomes (Table 2). For community abundance, Ulvaria and Cladophora blooms had significant negative effects, while Vaucheria had a significant positive effect (Fig. 3a). The effects of Ulva and mixed algal blooms on community abundance were similar to the overall mean effect, but these effects were nonsignificant, as were those of the other algal groups (Fig. 3a). For species richness, Cladophora had a strong negative effect on species richness, and algal mixtures had a moderate negative effect (Fig. 3b). Single studies of Vaucheria and Laurencia
(a) Abundance
(b) Richness
(c) Diversity
(d)Evenness
(e) GPP
(f) NPP
(g) Respiration
−150
100 50
−75
0
75
150
Longitude
0
50
100
−50
0
Latitude
−50
0
50
100
−50
0
50
100
2716 D . A . L Y O N S et al.
−50
−20 −15 −10 −5
0
5
10 15 20
Hedge's g −150
−75
0
75
150
Longitude Fig. 1 Locations of the studies examining the effects of macroalgal blooms on: (a) community abundance, (b) species richness, (c) diversity (Shannon index), (d) evenness (Pielou’s index), (e) gross primary productivity (GPP), (f) net primary productivity (NPP), and (g) respiration. The size of each circle is indicative of the magnitude of the observed effects. Positive effects are plotted in blue, negative effects in orange.
each found significant positive effects. The effects of blooms by other algal taxa were not statistically significant.
Abundance (76) Richness (34) Diversity (13) Evenness (7)
Subgroup analyses: impacts of blooms in different habitats
GPP (6) NPP (8) Respiration (9) –2
0
2
4
Hedges' g Fig. 2 The mean effects (Hedges’ g and 95% CI) of macroalgal blooms on community abundance, species richness, diversity (Shannon index), evenness (Pielou’s index), gross primary productivity (GPP), net primary productivity (NPP), and respiration. Each effect was estimated using a random effects metaanalysis. The number next to the effect size label is the number of studies found for that outcome. The vertical dashed line indicates no effect.
Habitat was a significant modifier variable in the subgroup analyses of macroalgal blooms’ effects on species richness and net primary productivity, but did not have a significant influence on the other five outcomes (Table 2). For species richness, the strongest effects of macroalgal blooms were observed in subtidal sand/mud habitats and in the rocky intertidal (Fig. 3c). Effects in both these habitats were negative, as were the nonsignificant effects in oyster reefs and rocky subtidal habitats. In contrast, the mean effects in subtidal seagrass beds and intertidal sand/mud habitats were slightly positive, but nonsignificant. For net primary productivity, the clearest difference was © 2014 John Wiley & Sons Ltd, Global Change Biology, 20, 2712–2724
E C O L O G I C A L E F F E C T S O F M A C R O A L G A L B L O O M S 2717 Vaucheria (1) Ulvaria (1) Ulva (38) Mix (21) Laurencia (2) Gracilaria (3) Furcellaria(1) Fucus (2) Digenia (2) Cladophora (4)
(a)
Vaucheria (1) Mix (10) Laurencia (1) Gracilaria (2) Fucus (2) Cladophora (3)
−10
−5
0
5
−5
Hedges' g (Abundance)
Subtidal seagrass (6)
(b)
Ulva (15)
(c)
0
5
10
15
Hedges' g (Richness)
Subtidal seagrass (2)
(d)
Subtidal sand/mud (9) Rocky subtidal (1)
Subtidal sand/mud (3)
Rocky intertidal (1)
Intertidal seagrass (1)
Oyster reef (2) Intertidal sand/mud (2)
Intertidal sand/mud (15) −8
−4
0
−8
Subtidal seagrass (27)
(e)
Subtidal seagrass (1)
Subtidal sand/mud (22) Rocky subtidal (1)
−4
0
4
Hedges' g (NPP)
Hedges' g (Richness)
(f)
Subtidal sand/mud (8)
Rocky intertidal (1) Pelagic (1)
Pelagic (2) Oyster reef (2)
Intertidal sand/mud (3)
Intertidal sand/mud (21) −8
−4
0
Hedges' g (Abundance)
−4
−2
0
2
4
Hedges' g (Diversity)
Fig. 3 The influence of different blooming algal species (a–b) and affected habitats (c–f) on the effects (Hedges’ g and 95% CI) of macroalgal blooms. The effects within each panel were estimated in a mixed effects meta-analysis. The number next to each effect size label is the number of studies found for that outcome. The vertical dashed line indicates no effect and the vertical dotted line indicates the mean effect estimated in the overall meta-analysis.
between the strong negative effects observed in intertidal sand/mud habitats and the strong positive effect observed in subtidal sand/mud habitats (Fig. 3d). However, this comparison was confounded because measurements made in the intertidal excluded the productivity of the bloom, whereas measurements made in the subtidal included it (see below). Although we found no significant differences among habitats in the effects of blooms on either community abundance or diversity, the habitat subgroup analyses for these outcomes revealed two noteworthy results. First, macroalgal blooms reduced community abundance in all habitats, but these effects were only statistically significant in rocky intertidal and subtidal sand/mud habitats (Fig. 3e). Second, we found that macroalgal blooms had a significant negative effect on community diversity in subtidal sand/mud habitats (Fig. 3f), despite not having a significant overall effect (Fig. 2). © 2014 John Wiley & Sons Ltd, Global Change Biology, 20, 2712–2724
Subgroup analyses: impacts of bloom on different types of community Community type did not have significant effects in the subgroup analyses for community abundance, species richness, or diversity of marine and estuarine ecosystems (P > 0.189, Table S3.1). However, the subgroup analyses revealed that macroalgal blooms reduced the abundance, richness, and diversity of species in invertebrate communities (Fig. 4a–c). The effects on other community types varied, but none were statistically significant. A subgroup analysis for community evenness was not possible because all 13 studies examined invertebrate communities.
Subgroup analyses: influence of measurement methods on effect size estimates Measurement methods had a significant influence on the effects observed in studies of net primary productivity
2718 D . A . L Y O N S et al. Seagrasses (3)
(a)
Mix (4)
(b)
Mix (3) Macroalgae (2)
Microalgae (16) Macroalgae (4)
Invertebrates (25)
Invertebrates (45)
Fish (2)
Fish (4) Birds (1)
Bacteria (1) −2
0
2
−2
0
2
Hedges' g (Richness)
Hedges' g (Abundance)
(c)
(d)
Mix (1)
Including macroalgae (5)
Invertebrates (11) Excluding macroalgae (2)
Bacteria (1) −2
0
2
4
−6
Hedges' g (Diversity)
−4
−2
0
2
4
Hedges' g (NPP)
(e)
(f) Density (30)
Including macroalgae Observational (7)
Cover (2) Excluding macroalgae Experimental (2)
Biomass (44) −2
0
2
Hedges' g (Respiration)
−3
−2
−1
0
1
Hedges' g (Abundance)
Fig. 4 The influence of community type and different measurement methods on the effects (Hedges’ g and 95% CI) of macroalgal blooms. Community type subplots compare the effects of macroalgal blooms on the abundance (a), species richness (b), and diversity (c) of assemblages composed of different. ‘Mix’ refers to assemblages that included both macroalgae and invertebrates. For net primary productivity (d) and respiration (e) the plots compare measurement methods that either included the contribution of the macroalgal bloom to the response, or measured only the response of the community under the macroalgal mat (macroalgae excluded). For respiration, this plot also compares the results of experimental and observational studies because all the studies including the respiration of the bloom in their estimates were observational, and all the studies excluding it were experimental. For community abundance (f), the plot compares studies that used the density, cover, or biomass to measure abundance. The effects within each panel were estimated in a mixed effects meta-analysis. The number next to each effect size label is the number of studies found for that outcome. The vertical dashed line indicates no effect and the vertical dotted line indicates the mean effect estimated in the overall meta-analysis.
and respiration, but we found no significant difference among studies using biomass, density, or cover as measures of community abundance (Table 2; Fig. 4d–f). For net primary productivity, studies that included the productivity of the bloom in their measurement found a strong and significant positive effect whereas studies that did not found a strong negative effect (Fig. 4d). The effects observed in individual studies within each of these two subgroups were very consistent, and were not significantly heterogeneous (Table 2; Figure S3.10). Similarly, studies that included the respiration of the bloom in their measurements found that macroalgal blooms had significant positive effects, and those that
did not found that blooms had nonsignificant negative effects (Fig. 4e; Figure S3.11). This comparison was confounded with study type. All observational studies included the respiration of the bloom in their measurements, while all experimental studies measured only the respiration of the communities living under macroalgal mats.
Subgroup analyses: influence of study type and study setting We found no significant differences between the effects observed in experimental and observational studies of © 2014 John Wiley & Sons Ltd, Global Change Biology, 20, 2712–2724
E C O L O G I C A L E F F E C T S O F M A C R O A L G A L B L O O M S 2719 Table 1 Results of significance tests for the random effects meta-analyses examining the overall effects of macroalgal blooms on measures of ecosystem structure and functioning and associated tests for study heterogeneity. The P-values of statistically significant meta-analyses are in bold. The n refers to the number of studies in each analysis Primary analysis
Study heterogeneity
Outcome variable(n)
z
P
s2
Cochran’s Q
P
Community abundance(76) Species richness(34) Community diversity(13) Community evenness(7) Gross primary productivity(6) Net primary productivity(8) Respiration(9)
2.81 2.59 1.91 0.15 3.58 0.30 1.45
0.005 0.010 0.056 0.884
Macroalgal blooms alter community structure and primary productivity in marine ecosystems DEVIN A. LYONS1, CHRISTOS ARVANITIDIS2, ANDREW J. BLIGHT3, EVA CHATZINIKOLAOU2, TAMAR GUY-HAIM4, JONNE KOTTA5, HELEN ORAV-KOTTA5, 6 , G I L R I L O V 4 , P A U L J . S O M E R F I E L D 6 and T A S M A N P . C R O W E 1 A N A M . Q U E I R OS 1 University College Dublin, Dublin, Ireland, 2Hellenic Centre for Marine Research, Herakilon, Greece, 3University of St. Andrews, St. Andrews, UK, 4Israel Oceanographic and Limnological Research, Haifa, Israel, 5Estonian Marine Institute, University of Tartu, Tartu, Estonia, 6Plymouth Marine Laboratory, Plymouth, UK
Abstract Eutrophication, coupled with loss of herbivory due to habitat degradation and overharvesting, has increased the frequency and severity of macroalgal blooms worldwide. Macroalgal blooms interfere with human activities in coastal areas, and sometimes necessitate costly algal removal programmes. They also have many detrimental effects on marine and estuarine ecosystems, including induction of hypoxia, release of toxic hydrogen sulphide into the sediments and atmosphere, and the loss of ecologically and economically important species. However, macroalgal blooms can also increase habitat complexity, provide organisms with food and shelter, and reduce other problems associated with eutrophication. These contrasting effects make their overall ecological impacts unclear. We conducted a systematic review and meta-analysis to estimate the overall effects of macroalgal blooms on several key measures of ecosystem structure and functioning in marine ecosystems. We also evaluated some of the ecological and methodological factors that might explain the highly variable effects observed in different studies. Averaged across all studies, macroalgal blooms had negative effects on the abundance and species richness of marine organisms, but blooms by different algal taxa had different consequences, ranging from strong negative to strong positive effects. Blooms’ effects on species richness also depended on the habitat where they occurred, with the strongest negative effects seen in sandy or muddy subtidal habitats and in the rocky intertidal. Invertebrate communities also appeared to be particularly sensitive to blooms, suffering reductions in their abundance, species richness, and diversity. The total net primary productivity, gross primary productivity, and respiration of benthic ecosystems were higher during macroalgal blooms, but blooms had negative effects on the productivity and respiration of other organisms. These results suggest that, in addition to their direct social and economic costs, macroalgal blooms have ecological effects that may alter their capacity to deliver important ecosystem services. Keywords: biodiversity, ecosystem functioning, green tide, harmful algal bloom, macroalgal bloom, macroalgal mat, species richness Received 17 October 2013; revised version received 14 April 2014 and accepted 2 May 2014
Introduction Marine and estuarine ecosystems are under pressure from a wide variety of anthropogenic stressors, which have degraded their environmental quality (Lotze et al., 2006; Halpern et al., 2008). This ongoing degradation is commonly believed to have contributed to an increase in the frequency, magnitude, and extent of outbreaks by marine diseases, gelatinous zooplankton, and harmful algae (Harvell et al., 1999; Anderson, 2009; Brotz et al., 2012; Duarte et al., 2013). Understanding how ecosystems respond to these outbreaks is an important step in fully accounting for the social, economic, and ecological costs and benefits of the factors that drive Correspondence: Devin A. Lyons, tel. +780 680-1085, fax +780 492-9234, e-mail: [email protected]
2712
outbreaks, as well as the management actions that might be used to deal with them. In this study, we attempt to facilitate this process by evaluating the ecological effects of one type of species outbreak: macroalgal blooms. Macroalgal blooms are increasing worldwide, and are feared to have a number of undesirable effects on marine and estuarine ecosystems (Teichberg et al., 2012). Expanding human population, fertilizer use, livestock waste production, and fossil fuel combustion have overcome nitrogen and phosphorus limitation of coastal and estuarine waters, leading to accelerated nutrient uptake, faster growth, and more frequent blooms by macroalgae around the world (Nixon, 1995; Conley, 1999; Howarth, 2008; Teichberg et al., 2010). These changes have been aggravated by the loss of herbivores in some ecosystems. For example, harvesting of © 2014 John Wiley & Sons Ltd
E C O L O G I C A L E F F E C T S O F M A C R O A L G A L B L O O M S 2713 large predatory fish in the Baltic Sea appears to have increased macroalgal abundance via a trophic cascade (Eriksson et al., 2009; Sieben et al., 2011). With reduced predatory pressure of large fish, the abundance of smaller predators has increased, driving down populations of invertebrate grazers, and allowing macroalgae to proliferate. While both eutrophication and loss of herbivory increase macroalgal biomass, they also interact synergistically to enhance each other’s effect (Burkepile & Hay, 2006). Opportunistic macroalgae proliferate as a result, forming dense canopies and creating mats of algae that float in the water column or settle on the substrate. Often macroalgal blooms are confined to relatively small areas within a single estuary or embayment, but they can become very large. Massive blooms of Ulva spp. have occurred annually in the Yellow Sea since 2007, creating floating mats that covered up to 3489 square kilometres of its surface (Liu et al., 2013). Regardless of whether they are relatively small or very large, macroalgal blooms can have a number of detrimental social, economic, and ecological consequences. Macroalgal blooms inhibit recreation, diminish aesthetic enjoyment of the coastal zone, and interfere with tourism, fishing and mariculture (Charlier & Lonhienne, 1996; Dion & Bozec, 1996; de Leo et al., 2002). Toxic gases are emitted from mats of rotting macroalgae, posing a potential risk to human health (Chrisafis, 2009; Samuel, 2011). Proliferating macroalgae often alter the biological composition and ecological processes of the ecosystems they affect (Fletcher, 1996; Raffaelli et al., 1998). These undesirable effects have caused considerable concern and prompted costly algal removal programmes in some affected areas (Morand & Merceron, 2005). The negative ecological effects of macroalgal blooms occur via their alteration of the physical and chemical environment and their interactions with other species. When mats of blooming macroalgae settle on the seafloor, they can alter environmental conditions by inducing anoxia and the release of hydrogen sulphide from the sediments, resulting in reductions of species richness and community abundance (Gamenick et al., 1996; Wetzel et al., 2002). The physiological stress these changes create can also elicit sublethal responses from benthic fauna. Mobile burrowers typically respond by repositioning themselves towards the sediment–water interface (Wright et al., 2010). This may alter their susceptibility to predation and can also impact important sedimentary ecosystem processes that these species mediate (Raffaelli et al., 1998; Gribben et al., 2009; Queir os et al., 2011). The competitive effects of macroalgal blooms have been implicated in the worldwide decline in seagrass beds (Valiela et al., 1997; McGlathery, 2001), and linked to local declines in nonblooming algae © 2014 John Wiley & Sons Ltd, Global Change Biology, 20, 2712–2724
(Kautsky et al., 1986; Worm et al., 1999). By driving declines in such foundation species, macroalgal blooms may also have negative indirect effects on other members of the biological community. Some macroalgal bloom species produce chemicals that inhibit the growth, development, and feeding of other species (van Alstyne & Houser, 2003; Nelson et al., 2003; Nan et al., 2008), suggesting that macroalgal toxins may contribute to blooms’ effects. Much of the literature on macroalgal blooms focuses on their negative effects on individual species, communities, and ecological processes, but macroalgal blooms also have several effects that may be considered beneficial. Macroalgal blooms increase transfer of nutrients from the water column to the sediments and other macroalgae, thereby reducing nutrient levels in eutrophic waters (Thybo-Christesen et al., 1993; Hardison et al., 2010). Accumulations of macroalgae can also increase habitat complexity, enhance dispersal of other species, and provide animals with food and shelter (Wilson et al., 1990; Holmquist, 1994, 1997). As a result, macroalgal blooms may actually enhance, rather than reduce, biodiversity and secondary production in some ecosystems (Holmquist, 1997; Bolam & Fernandes, 2002; Dolbeth et al., 2003). Several qualitative reviews of macroalgal blooms’ ecological impacts have been published in the past (e.g. Fletcher, 1996; Raffaelli et al., 1998). However, elucidating the net impact of blooms’ varied positive and negative effects remains an important challenge. Here we provide a quantitative synthesis that complements previous reviews by estimating the overall impact of macroalgal blooms on seven key measures of ecosystem structure and functioning, using meta-analysis. We were interested in investigating community- and ecosystem-level effects of blooms, and focused on how they affected species richness, community diversity (Shannon Index), community evenness (Pielou Index), community abundance (biomass, density, cover), gross primary productivity, net primary productivity, and community respiration in temperate and subtropical ecosystems. In addition, we investigated the effects of the algae responsible for the bloom, the habitat the bloom occurs in, the type of community the bloom affects, and the methodology used to study the bloom as potential drivers of variation in the responses observed in different studies.
Materials and methods We carried out this synthesis using the process of ‘systematic review’, following guidelines recommended by the Centre For Evidence-Based Conservation (2010). Our full, peer-reviewed protocol for conducting this review is available in a separate
2714 D . A . L Y O N S et al. publication (Lyons et al., 2012). Below we provide a brief description of our methods, and note a number of amendments made to the protocol during the review process.
Search strategy and screening of publications We searched online databases to find publications (i.e. research articles) that could be used to evaluate the effects of macroalgal blooms on the structure and functioning of marine and estuarine ecosystems. Specifically, we searched Web Of Science and Scopus databases on 28 June 2012, using a complex set of search terms that included keywords related to macroalgal blooms, macroalgal mats, blooming algal taxa, and a broad range of ecosystem structure and functioning measures (Data S1). Following our search, we assessed the publications in a three-stage screening process designed to identify relevant studies according to a set of predefined inclusion criteria (described below). First, one reviewer assessed the titles to remove any publications dealing with completely unrelated topics. Second, eight different reviewers assessed the abstracts of the remaining publications, with each reviewer evaluating an independent subset of the articles. Finally, each reviewer assessed the full text of the articles that they had accepted. To ensure consistent application of the inclusion criteria, all reviewers assessed two samples of the abstracts before the publications were distributed. We assessed consistency among reviewers with Fleiss’ Kappa statistic (Fleiss, 1971), using a Kappa of at least 0.5 to indicate an acceptable degree of consistency. The assessment of the first sample was not acceptable (Kappa = 0.27), so we discussed inconsistent assessments and clarified how the inclusion criteria were to be applied. This resulted in a sufficient degree of consistency in the second assessment (Kappa = 0.57), but we discussed the inconsistent assessments to further clarify the inclusion criteria and improve performance.
Inclusion criteria To be retained at each stage, a publication had to include at least one study (i.e. individual experiment or observational study) that met a number of criteria. The study had to: (i) examine a temperate or subtropical ecosystem affected by marine or estuarine macroalgal blooms or macroalgal mats; (ii) involve a comparison between an affected state (with a macroalgal bloom or mat) and a control state (without macroalgal bloom or mat); (iii) report on a measure of ecosystem structure or functioning; (iv) report the results of a manipulative experiment or observational study; and (v) include sufficient information so that an effect size (Hedges’ g) could be calculated. We did not define blooms or mats according to particular thresholds in their size or severity, in part because publications frequently neglected to report on the density or spatial extent of macroalgae in their study or used incomparable metrics. Rather, we considered something a bloom or mat if the authors of a publication referred to them as such, or if there was an indication that an accumulation of algae was
either unusually large or caused by anthropogenic factors such as eutrophication. For the first two steps in the screening process, we retained studies that potentially reported on any measurement of ecosystem structure or functioning. When we screened the full text of publications, we elected to include only studies examining species richness, community diversity (Shannon index), community evenness (Pielou Index), and communitylevel measures of gross primary productivity, net primary productivity, respiration, and biomass, density, or cover (‘community abundance’ hereafter). We selected these outcomes because we felt that they were among the most important community- or ecosystem-level indicators for the structure and functioning of marine ecosystems. In addition, sufficient data to conduct a meta-analysis for these outcomes were available. We also searched for data on communitylevel secondary productivity but failed to find appropriate studies. We only included experimental studies if algal abundance was directly manipulated. If macroalgal abundance was not directly manipulated, but an experimental manipulation of some other factor (e.g. nutrients, grazer abundance) resulted in a macroalgal bloom in some replicates we did not include the study. To calculate effect sizes (Hedge’s g), means, sample sizes, and either standard deviations, standard errors, or confidence intervals were necessary. When incomplete information was available, we attempted to obtain the missing information from the authors of the publication prior to excluding it. If we discovered that more than one publication reported the results of the same study (i.e. made the same comparison using the same data), we chose the publication that presented the data for the comparison most clearly, or we randomly selected one of the publications.
Data extraction and meta-analyses When possible, we extracted means, standard errors, standard deviations and sample sizes directly from tables and the text of the articles. In other cases, we used ImageJ (Schneider et al., 2012), DataThief (Tummers, 2006), or Engauge Digitizer (Mitchell, 2010) software to extract them from figures. The extracted data were used to calculate an effect size, Hedges’ g, for each study. Hedges’ g is the bias-corrected mean difference in the response variable between the ‘treatment’ (i.e. algal bloom) and control conditions, standardized by the withingroup standard deviation. Some publications reported results for more than one outcome variable. The data for our seven outcome variables were analysed in separate meta-analyses (see below), allowing us to treat different outcomes from the same publication as independent studies. Other publications contained multiple results for the same outcome variable. How these data were treated depended on how they were generated and structured. A detailed account of how we dealt with these data, and how we estimated the effect size and variance for studies with different designs, is available in Data S1. We synthesized the data for the effects of macroalgal blooms using a random effects model to estimate the overall effect of macroalgal blooms on each of our outcome variables.
© 2014 John Wiley & Sons Ltd, Global Change Biology, 20, 2712–2724
E C O L O G I C A L E F F E C T S O F M A C R O A L G A L B L O O M S 2715 These analyses were conducted in R (R Development Core Team, 2012), using the ‘metafor’ package (Viechtbauer, 2010), and the DerSimonian–Laird estimator (Dersimonian & Laird, 1986). We used Cochran’s Q-test to assess study heterogeneity. We used funnel plots and a rank correlation test for funnel plot asymmetry to assess whether there was any evidence of publication bias or other ‘small study effects’.
Subgroup analyses Study-level covariates may contribute to variation in the effects estimated in different studies. We carried out subgroup analyses using mixed effects meta-analysis to investigate the potential influence of algal identity (i.e. the type of alga responsible for the bloom), habitat, community type (e.g. macroalgae, invertebrates, bacteria), measurement methods, study type (experiment vs. observational) and study setting (lab vs. field) on the estimated effects of macroalgal blooms. The importance of each modifier variable was examined in a separate analysis, using the potential modifier as a fixed effect in the mixed model. The significance of modifier variables was assessed using a Q-test. In addition to determining whether macroalgal blooms’ effects differed among subgroups, these analyses allowed us to evaluate the significance of macroalgal blooms’ effects for the individual subgroups within each analysis. We assessed the significance of individual subgroups using the mean effect and 95% confidence interval estimated by the model. The number of studies we found for some subgroups was low. As a result, the power of some of our subgroup analyses may be low, and cause us to underestimate the importance of the modifying variable. Moreover, the effect sizes estimated from a small number of studies should be viewed with caution. We carried out the subgroup analyses for all outcome variables, as long the modifier variable was relevant to the outcome and sufficient data were available (i.e. one or more studies in each of two or more subgroups). The ‘measurement methods’ subgroup analysis for community abundance compared the results of studies that used either biomass, density, or cover as measures of abundance. For net primary productivity and respiration, it compared studies that either measured total productivity/respiration or the productivity/ respiration of the rest of the affected community (i.e. excluding the productivity or respiration of the algal bloom). We expected that the blooms would increase total productivity and respiration when their contribution was taken into account, and have a negative effect when only the rest of the community was considered. We did not conduct a similar subgroup analysis for gross primary productivity because all studies measured total productivity, including that of the bloom.
Results
Description of the data sets Our bibliographic search returned 906 publications. We retained 638 of these after screening the titles, 298 after © 2014 John Wiley & Sons Ltd, Global Change Biology, 20, 2712–2724
assessing the abstracts, and 52 after screening the full text (Data S1). The majority of publications contained multiple studies of a single outcome or studies of more than one outcome, allowing us to estimate effect sizes for a total 153 studies. These studies were conducted on coasts of all continents except Antarctica, but more than 70% were from European waters (Fig. 1). The number of studies for each outcome variable ranged from 6 to 76 (Fig. 2). The datasets for all outcomes were biased towards studies of blooms by ephemeral green algae (e.g. Ulva, Cladophora), studies conducted in soft sediment (sand or mud) and seagrass bed habitats, and studies of invertebrate communities. The number of studies in the modifier variable subgroups for each outcome is presented in Figs 3, 4 and Figures S3.1–S3.4 in Data S1.
Overall effects Averaged across all studies, macroalgal blooms had significant negative effects on the abundance and species richness of organisms in marine and estuarine communities, and a significant positive effect on gross primary productivity (Fig. 2; Table 1). On average, blooms led to increases in net primary productivity and respiration, and a decrease in community diversity, though these effects were not statistically significant. Macroalgal blooms appeared to have very little effect on community evenness. Funnel plots and rank correlation tests suggested that publication bias did not affect our effect size estimates (Data S1). All seven of the outcome variables displayed significant among-study heterogeneity (Table 1; Figures S3.5–S3.11 in Data S1), so we investigated potential sources of this heterogeneity in a series of subgroup analyses.
Subgroup analyses: impacts of blooms by different taxa The identity of the seaweed(s) responsible for a bloom had a significant effect in the subgroup analyses of blooms’ effects on community abundance and species richness, but did not have a significant influence on the other five outcomes (Table 2). For community abundance, Ulvaria and Cladophora blooms had significant negative effects, while Vaucheria had a significant positive effect (Fig. 3a). The effects of Ulva and mixed algal blooms on community abundance were similar to the overall mean effect, but these effects were nonsignificant, as were those of the other algal groups (Fig. 3a). For species richness, Cladophora had a strong negative effect on species richness, and algal mixtures had a moderate negative effect (Fig. 3b). Single studies of Vaucheria and Laurencia
(a) Abundance
(b) Richness
(c) Diversity
(d)Evenness
(e) GPP
(f) NPP
(g) Respiration
−150
100 50
−75
0
75
150
Longitude
0
50
100
−50
0
Latitude
−50
0
50
100
−50
0
50
100
2716 D . A . L Y O N S et al.
−50
−20 −15 −10 −5
0
5
10 15 20
Hedge's g −150
−75
0
75
150
Longitude Fig. 1 Locations of the studies examining the effects of macroalgal blooms on: (a) community abundance, (b) species richness, (c) diversity (Shannon index), (d) evenness (Pielou’s index), (e) gross primary productivity (GPP), (f) net primary productivity (NPP), and (g) respiration. The size of each circle is indicative of the magnitude of the observed effects. Positive effects are plotted in blue, negative effects in orange.
each found significant positive effects. The effects of blooms by other algal taxa were not statistically significant.
Abundance (76) Richness (34) Diversity (13) Evenness (7)
Subgroup analyses: impacts of blooms in different habitats
GPP (6) NPP (8) Respiration (9) –2
0
2
4
Hedges' g Fig. 2 The mean effects (Hedges’ g and 95% CI) of macroalgal blooms on community abundance, species richness, diversity (Shannon index), evenness (Pielou’s index), gross primary productivity (GPP), net primary productivity (NPP), and respiration. Each effect was estimated using a random effects metaanalysis. The number next to the effect size label is the number of studies found for that outcome. The vertical dashed line indicates no effect.
Habitat was a significant modifier variable in the subgroup analyses of macroalgal blooms’ effects on species richness and net primary productivity, but did not have a significant influence on the other five outcomes (Table 2). For species richness, the strongest effects of macroalgal blooms were observed in subtidal sand/mud habitats and in the rocky intertidal (Fig. 3c). Effects in both these habitats were negative, as were the nonsignificant effects in oyster reefs and rocky subtidal habitats. In contrast, the mean effects in subtidal seagrass beds and intertidal sand/mud habitats were slightly positive, but nonsignificant. For net primary productivity, the clearest difference was © 2014 John Wiley & Sons Ltd, Global Change Biology, 20, 2712–2724
E C O L O G I C A L E F F E C T S O F M A C R O A L G A L B L O O M S 2717 Vaucheria (1) Ulvaria (1) Ulva (38) Mix (21) Laurencia (2) Gracilaria (3) Furcellaria(1) Fucus (2) Digenia (2) Cladophora (4)
(a)
Vaucheria (1) Mix (10) Laurencia (1) Gracilaria (2) Fucus (2) Cladophora (3)
−10
−5
0
5
−5
Hedges' g (Abundance)
Subtidal seagrass (6)
(b)
Ulva (15)
(c)
0
5
10
15
Hedges' g (Richness)
Subtidal seagrass (2)
(d)
Subtidal sand/mud (9) Rocky subtidal (1)
Subtidal sand/mud (3)
Rocky intertidal (1)
Intertidal seagrass (1)
Oyster reef (2) Intertidal sand/mud (2)
Intertidal sand/mud (15) −8
−4
0
−8
Subtidal seagrass (27)
(e)
Subtidal seagrass (1)
Subtidal sand/mud (22) Rocky subtidal (1)
−4
0
4
Hedges' g (NPP)
Hedges' g (Richness)
(f)
Subtidal sand/mud (8)
Rocky intertidal (1) Pelagic (1)
Pelagic (2) Oyster reef (2)
Intertidal sand/mud (3)
Intertidal sand/mud (21) −8
−4
0
Hedges' g (Abundance)
−4
−2
0
2
4
Hedges' g (Diversity)
Fig. 3 The influence of different blooming algal species (a–b) and affected habitats (c–f) on the effects (Hedges’ g and 95% CI) of macroalgal blooms. The effects within each panel were estimated in a mixed effects meta-analysis. The number next to each effect size label is the number of studies found for that outcome. The vertical dashed line indicates no effect and the vertical dotted line indicates the mean effect estimated in the overall meta-analysis.
between the strong negative effects observed in intertidal sand/mud habitats and the strong positive effect observed in subtidal sand/mud habitats (Fig. 3d). However, this comparison was confounded because measurements made in the intertidal excluded the productivity of the bloom, whereas measurements made in the subtidal included it (see below). Although we found no significant differences among habitats in the effects of blooms on either community abundance or diversity, the habitat subgroup analyses for these outcomes revealed two noteworthy results. First, macroalgal blooms reduced community abundance in all habitats, but these effects were only statistically significant in rocky intertidal and subtidal sand/mud habitats (Fig. 3e). Second, we found that macroalgal blooms had a significant negative effect on community diversity in subtidal sand/mud habitats (Fig. 3f), despite not having a significant overall effect (Fig. 2). © 2014 John Wiley & Sons Ltd, Global Change Biology, 20, 2712–2724
Subgroup analyses: impacts of bloom on different types of community Community type did not have significant effects in the subgroup analyses for community abundance, species richness, or diversity of marine and estuarine ecosystems (P > 0.189, Table S3.1). However, the subgroup analyses revealed that macroalgal blooms reduced the abundance, richness, and diversity of species in invertebrate communities (Fig. 4a–c). The effects on other community types varied, but none were statistically significant. A subgroup analysis for community evenness was not possible because all 13 studies examined invertebrate communities.
Subgroup analyses: influence of measurement methods on effect size estimates Measurement methods had a significant influence on the effects observed in studies of net primary productivity
2718 D . A . L Y O N S et al. Seagrasses (3)
(a)
Mix (4)
(b)
Mix (3) Macroalgae (2)
Microalgae (16) Macroalgae (4)
Invertebrates (25)
Invertebrates (45)
Fish (2)
Fish (4) Birds (1)
Bacteria (1) −2
0
2
−2
0
2
Hedges' g (Richness)
Hedges' g (Abundance)
(c)
(d)
Mix (1)
Including macroalgae (5)
Invertebrates (11) Excluding macroalgae (2)
Bacteria (1) −2
0
2
4
−6
Hedges' g (Diversity)
−4
−2
0
2
4
Hedges' g (NPP)
(e)
(f) Density (30)
Including macroalgae Observational (7)
Cover (2) Excluding macroalgae Experimental (2)
Biomass (44) −2
0
2
Hedges' g (Respiration)
−3
−2
−1
0
1
Hedges' g (Abundance)
Fig. 4 The influence of community type and different measurement methods on the effects (Hedges’ g and 95% CI) of macroalgal blooms. Community type subplots compare the effects of macroalgal blooms on the abundance (a), species richness (b), and diversity (c) of assemblages composed of different. ‘Mix’ refers to assemblages that included both macroalgae and invertebrates. For net primary productivity (d) and respiration (e) the plots compare measurement methods that either included the contribution of the macroalgal bloom to the response, or measured only the response of the community under the macroalgal mat (macroalgae excluded). For respiration, this plot also compares the results of experimental and observational studies because all the studies including the respiration of the bloom in their estimates were observational, and all the studies excluding it were experimental. For community abundance (f), the plot compares studies that used the density, cover, or biomass to measure abundance. The effects within each panel were estimated in a mixed effects meta-analysis. The number next to each effect size label is the number of studies found for that outcome. The vertical dashed line indicates no effect and the vertical dotted line indicates the mean effect estimated in the overall meta-analysis.
and respiration, but we found no significant difference among studies using biomass, density, or cover as measures of community abundance (Table 2; Fig. 4d–f). For net primary productivity, studies that included the productivity of the bloom in their measurement found a strong and significant positive effect whereas studies that did not found a strong negative effect (Fig. 4d). The effects observed in individual studies within each of these two subgroups were very consistent, and were not significantly heterogeneous (Table 2; Figure S3.10). Similarly, studies that included the respiration of the bloom in their measurements found that macroalgal blooms had significant positive effects, and those that
did not found that blooms had nonsignificant negative effects (Fig. 4e; Figure S3.11). This comparison was confounded with study type. All observational studies included the respiration of the bloom in their measurements, while all experimental studies measured only the respiration of the communities living under macroalgal mats.
Subgroup analyses: influence of study type and study setting We found no significant differences between the effects observed in experimental and observational studies of © 2014 John Wiley & Sons Ltd, Global Change Biology, 20, 2712–2724
E C O L O G I C A L E F F E C T S O F M A C R O A L G A L B L O O M S 2719 Table 1 Results of significance tests for the random effects meta-analyses examining the overall effects of macroalgal blooms on measures of ecosystem structure and functioning and associated tests for study heterogeneity. The P-values of statistically significant meta-analyses are in bold. The n refers to the number of studies in each analysis Primary analysis
Study heterogeneity
Outcome variable(n)
z
P
s2
Cochran’s Q
P
Community abundance(76) Species richness(34) Community diversity(13) Community evenness(7) Gross primary productivity(6) Net primary productivity(8) Respiration(9)
2.81 2.59 1.91 0.15 3.58 0.30 1.45
0.005 0.010 0.056 0.884
