1 3

Mar Biol (2014) 161:2395–2405DOI 10.1007/s00227-014-2514-7

ORIGINAL PAPER

Single and joint effects of regional‑ and local‑scale variables on tropical seagrass fish assemblages

Elisa Alonso Aller · Martin Gullström · Floriaan K J Eveleens Maarse · Michaela Gren · Lina Mtwana Nordlund · Narriman Jiddawi · Johan S. Eklöf

Received: 13 March 2014 / Accepted: 30 July 2014 / Published online: 14 August 2014 © Springer-Verlag Berlin Heidelberg 2014

local scale, seagrass biomass had a positive (but weaker) influence on fish density. However, the positive effect of seagrass biomass decreased with increasing level of human development. In summary, our results highlight the importance of assessing how multiple local and regional variables, alone and together, influence fish communi-ties, in order to improve management of seagrass ecosys-tems and their services.

Introduction

Seagrass beds are among the most productive and diverse ecosystems on Earth (Duarte and Chiscano 1999) and are of great importance to human wellbeing (Cullen-Unsworth et al. 2014). They support abundant and diverse fish assem-blages (Gell and Whittington 2002) and typically have a higher density and diversity of fish than non-vegetated areas (Pollard 1984; Edgar et al. 1994; Eklöf et al. 2006). In tropical areas, seagrass beds are commonly found in close proximity to other shallow-water habitats, such as mangrove forests, coral reefs, mud and sand flats (Ogden 1988; Olds et al. 2012). These habitats are connected by an array of physical, chemical and biological processes, form-ing the spatially diverse ‘tropical seascape’ (Ogden 1988; Pittman et al. 2011). Numerous fish species are known to migrate between seagrass beds and other shallow-water habitats, either on a daily basis (Unsworth et al. 2007) or through ontogenetic migrations (Dorenbosch et al. 2006a; Berkström et al. 2012). Consequently, seagrass beds are an essential habitat for many fish species that spend most of their life in other habitats (Parrish 1989; Nagelkerken et al. 2000; Gullström et al. 2012).

A number of variables are known to influence sea-grass-associated fish assemblages. On a local patch scale,

Abstract Seagrass beds are highly important for tropi-cal ecosystems by supporting abundant and diverse fish assemblages that form the basis for artisanal fisheries. Although a number of local- and regional-scale vari-ables are known to influence the abundance, diversity and assemblage structure of seagrass-associated fish assem-blages, few studies have evaluated the relative and joint (interacting) influences of variables, especially those act-ing at different scales. Here, we examined the relative importance of local- and regional-scale factors structur-ing seagrass-associated fish assemblages, using a field survey in six seagrass (Thalassodendron ciliatum) areas around Unguja Island (Zanzibar, Tanzania). Fish density and assemblage structure were mostly affected by two regional-scale variables; distance to coral reefs, which positively affected fish density, and level of human devel-opment, which negatively affected fish density. On the

Communicated by D. Goulet.

Electronic supplementary material The online version of this article (doi:10.1007/s00227-014-2514-7) contains supplementary material, which is available to authorized users.

E. Alonso Aller (*) · M. Gullström · F. K. J. Eveleens Maarse · M. Gren · L. M. Nordlund · J. S. Eklöf Department of Ecology, Environment and Plant Sciences, Stockholm University, 106 91 Stockholm, Swedene-mail: elisa.alonso-aller@su.se

L. M. Nordlund Western Indian Ocean – Community, Awareness, Research, and Environment (WIO CARE), P.O. Box 4199, Zanzibar, Tanzania

N. Jiddawi Institute of Marine Sciences, University of Dar Es Salaam, P.O. Box 668, Zanzibar, Tanzania

2396 Mar Biol (2014) 161:2395–2405

1 3

seagrass structural complexity (often estimated as shoot density, canopy height or shoot biomass) is known to posi-tively affect fish density and biomass (Adams 1976; Gull-ström et al. 2008), because of reduced predation risk (Orth et al. 1984; Heck and Orth 2006) and/or increased food availability with increasing complexity (Bell and Westoby 1986). However, not only local-scale but also regional- or ‘seascape’-scale variables influence seagrass fish. For example, the composition of and distance to nearby habitats in the seascape may influence seagrass fish assemblages (Parrish 1989; Gullström et al. 2008; Olds et al. 2012), both in seagrass beds (Baelde 1990; Dorenbosch et al. 2007; Unsworth et al. 2008) and coral reefs (Nagelkerken et al. 2000; Mumby et al. 2004).

Most studies examining which variables structure fish assemblages have used a single-scale approach, focus-ing either on local or regional variables (Bell and Westoby 1986; Jelbart et al. 2007). Meanwhile, a few studies sug-gest that seagrass fish assemblages are in fact shaped by multiple variables acting at different scales (e.g., Berkström et al. 2012). For example, Pittman et al. (2004) found that landscape composition was as important as within-patch seagrass structure for shaping seagrass fish assemblages. Similarly, Gullström et al. (2008) showed that seagrass fish assemblages were influenced by both the location of a seagrass habitat within the seascape, and by attributes of local seagrass structure. Importantly, while we know that multiple variables may simultaneously affect fish commu-nities, we have very limited understanding about the extent to which different variables interact to jointly affect fish communities. For example, it is theoretically possible that the effect of seagrass complexity (e.g., biomass) on a local scale may vary with, or depend upon, variables operating at larger spatial scales (e.g., the structure of the surrounding seascape).

From a socioeconomic perspective, seagrass beds pro-vide a range of ecosystem services to local coastal commu-nities, most importantly seagrass fisheries (Unsworth and Cullen 2010; Cullen-Unsworth et al. 2014). In the Western Indian Ocean (WIO) region, for example, seagrass beds are one of the most important fishing grounds and there-fore play an important role for local communities (de la Torre-Castro and Rönnbäck 2004; Nordlund et al. 2010). However, because of poorly implemented or missing management plans, seagrass habitats are currently under increasing pressure from anthropogenic disturbances on a global scale (Waycott et al. 2009; Nordlund et al. 2013). Many stocks of commercially important fish species using seagrass beds are heavily fished and show signs of overex-ploitation (UNEP 2001; Jiddawi and Öhman 2002; Jacquet and Zeller 2007). Moreover, other anthropogenic stressors such as effluent disposal, high input of inorganic nutrients from land, physically destructive fishing methods, damage

from boating activities and chemical pollution may also affect seagrass ecosystems and associated plant and animal communities (Orth et al. 2006; Ralph et al. 2006). More research is therefore needed to better understand the rela-tive importance of both natural and anthropogenic factors for seagrass fish assemblages, and how such knowledge can be used to reduce pressure on seagrass ecosystems (Unsworth and Cullen 2010).

The aim of this study was to examine the relative and joint influence of local- and regional-scale variables on seagrass-associated fish assemblages. We hypothesized that fish density, diversity and assemblage structure would be affected by variables both at a local scale—(1) vari-ables affecting seagrass structural complexity (% seagrass bottom cover, shoot density and shoot biomass) and (2) water depth—and at a regional scale—(3) level of human development in the surrounding area (as a proxy for fishing pressure and other disturbances), (4) wave exposure and (5) seascape configuration (distance to neighboring man-groves and coral reefs)—and that (6) the local effect of sea-grass structural complexity may interact with (or depend upon) regional-scale variables. The effects were also inde-pendently assessed for densities of juvenile and adult fish, since patterns could be life-stage specific, and for densities of fish species with different trophic guilds.

Materials and methods

Study area

The study was conducted between January and March 2012 in six sites around Unguja Island (Zanzibar, Tan-zania); Changuu Island (also known as Prison Island), Mbweni, Chumbe Island, Fumba, Chwaka Bay and Nungwi (remote site off the village; Fig. 1). The sites were selected so that they together would span as wide of a range of conditions in terms of levels of wave expo-sure and human development (two regional-scale variables typically affecting fish communities) as possible. At each site, sampling was conducted in seagrass beds dominated by the seagrass Thalassodendron ciliatum Forskål, a com-mon habitat-forming Indian Ocean species that harbor abundant and diverse fish communities (Gullström et al. 2002). Other seagrass species such as Thalassia hemp-richii, Cymodocea rotundata, C. serrulata, Syringodium isoetifolium, Halophila ovalis and Enhalus acoroides were occasionally present but in low densities (<40 % bottom cover, data not shown).

In each of the six sites, sampling was conducted in ten points located within a 100 × 100 m area and spaced >10 m apart (n = 10 within each site, N = 60 in total). Six of these points were randomly placed to reflect site

2397Mar Biol (2014) 161:2395–2405

1 3

conditions, while the remaining four were selectively placed so that the ten samples together would encompass as wide of a range in seagrass cover as possible. This mix of random and stratified sampling allowed us to assess how seagrass structural complexity influenced fish communities at the local scale (within each site), and along gradients of regional-scale variables (across sites).

Fish surveys

Fish sampling took place during high neap tide (peak ±3 h; except for Prison Island, where sampling was per-formed during spring tide due to logistical problems). Fish assemblages were surveyed using underwater visual census (Hill and Wilkinson 2004) by means of snorke-ling along a 25 × 2 m belt transect placed parallel to the shoreline starting at each of the sampling points. After placing a 25-m transect line on the bottom, the snorkeler waited for 10 min to reduce potential disturbance on fish. The transect was then swam through twice; the first time counting the most mobile, large fish, and the second time counting less mobile, small fish. Swimming speed was

kept constant (~0.1 m/s) to standardize the search time, and care was taken to avoid counting the same fish or group of fish more than once. Individuals found within or pass-ing through a transect were counted and identified to the lowest taxonomic level possible (usually species), and body lengths were visually estimated to the nearest 10 cm. Small schooling pelagic fish that are generally not closely asso-ciated with seagrass habitat (Atherinidae, Clupeidae and Engraulidae) were excluded from the analyses to avoid distortion of the dataset. This included Plotosus lineatus; a species with strong shoaling behavior and sporadic occur-rence (Gullström et al. 2008). For data analyses, fish were classified as juveniles when smaller than one-third of the species’ maximum length (following Nagelkerken and van der Velde 2002). Maximum length data were obtained from FishBase (Froese and Pauly 2013). Fish were then assigned to one of the following trophic guilds: algal herbivores, sea-grass herbivores, invertebrate feeders, invertebrate and fish feeders, and omnivores (following Gullström et al. 2008; Berkström et al. 2012; Froese and Pauly 2013; see Appen-dix 2). All fish density data were calculated as individuals m−2. As for measurements of fish diversity, fish species richness and Shannon–Weaver diversity index (H′) were calculated for each sampling point using the vegan package for R (Oksanen et al. 2013). Fish assemblage structure was defined by the taxonomic composition at each sampling point and the relative contributions of fish taxa in terms of density.

Local-scale variables

Depth

Water depth was measured at the beginning of each tran-sect by sounding. Depths ranged from 1.5 to 4.5 m. Prior to analyses, the actual depth was corrected to the depth rela-tive to the mean low water level using standard tide tables.

Seagrass habitat characteristics

Within each transect, the percentage cover of seagrass and other live substrates was estimated using 0.25 m2 quad-rates placed at 5 m intervals, resulting in six subsamples which were averaged to form a single replicate. Based on the averaged cover values, one of the quadrates was cho-sen as representative of the transect, where shoots from T. ciliatum and other seagrass species present were collected from a 0.0625 m2 quadrate. Samples were frozen until analyses could be carried out. In the laboratory, samples were carefully rinsed in freshwater and all epiphytes were scraped off. Seagrass leaves and their epiphytes were dried separately at 60° C for a minimum of 48 h (Duarte and Kirkman 2001) and weighed. The biomass was determined

N6

1

2

3

4

5

kilometers10

Africa

39° 30’ E

39° 30’ E

6° S6° S

a

b

Fig. 1 Map of (a) Africa (where the black square marks the location of Unguja Island, Zanzibar), and (b) Unguja Island with the six sam-pling sites: (1) Prison Island, (2) Mbweni, (3) Chumbe, (4) Fumba, (5) Chwaka Bay and (6) Nungwi

2398 Mar Biol (2014) 161:2395–2405

1 3

as dry weight (g DW m−2). Epiphyte biomass was stand-ardized in relation to the seagrass shoot biomass (as g DW epiphytes/g DW seagrass). The number of shoots of T. cili-atum was expressed as shoots m−2.

Regional-scale variables

Landscape configuration

Distances from each sampling point to the nearest coral reef and to the nearest mangrove forest were estimated using Google Earth (version 6.1) from satellite images taken from 2006 to 2010. These variables were treated as continuous variables in the statistical analyses, even though variance across samples within sites was very low.

Level of development

As a rough measure of the level of human development (which correlates with disturbances such as fishing pres-sure, boating activities and eutrophication), we estimated the ‘level of development’ as the number of houses within a 3 km radius from each sampling point (using Google Earth v. 6.1). The 3 km radius was chosen based on the distance that local fishermen usually travel for fishing in the study area (UNEP 2001; Lokrantz et al. 2010). Assuming that fishing pressure is greater in areas with higher number of houses within the radius (i.e., higher population), we used the level of development as a rough proxy for fishing pres-sure (Stewart et al. 2010).

Wave exposure

As a measure of wave exposure, the effective fetch (in km) was calculated for each sampling point (after Håkanson and Jansson 1983). The distance from each sampling point to land was measured for 14 different directions (in 6° inter-vals to either side from the perpendicular to the shore), and effective fetch was calculated as:

where yi is the angle of a given direction and xi is the dis-tance to land in that direction.

Statistical analyses

Prior to analyses, all possible variables (predictor variables) were tested for multicollinearity using pairwise Pearson’s correlation tests (r > 0.6). Based on the results, three vari-ables (distance to mangroves, epiphyte biomass and sea-grass shoot density) were excluded prior to the statistical analyses.

Effective fetch =

∑xi cos yi

∑cos yi

To explore the relative importance of five continuous predictor variables—seagrass shoot biomass, depth, dis-tance to coral reefs, wave exposure and level of devel-opment—for the univariate response variables (total fish density: adult and juvenile fish densities; herbivores den-sity, which included both algal and seagrass herbivores; algal herbivores, seagrass herbivores, invertebrate feed-ers, invertebrate and fish feeders, and omnivores densi-ties; fish diversity and species richness), we used general linear mixed effect models. The factor site (six levels) was treated as a random offset, while all predictor vari-ables were considered fixed. This approach allowed us to account for the natural and random variability among sites and generalize the effects of the fixed predictor variables (Beck 1997). Cube root (3√) transformation was applied to fish variables (total, adult, juvenile, herbivore and sea-grass herbivore fish densities) prior to analyses to obtain normal distribution. Predictor variables were either square root (√) or logarithmically (log10) transformed, if neces-sary. Data were standardized (by mean subtraction and dividing this difference by standard deviation) previous to analyses, which makes coefficients comparable (Grace and Bollen 2005).

Statistical analyses were conducted using R v. 2.15.1 (R Core Team 2013). Linear mixed effect models were fitted with the lme() function from the nlme package for R (Pin-heiro et al. 2013). Model selection was done using model averaging based on Akaike’s Information Criterion (AIC) scores (Burnham and Anderson 2002). In some cases, model fit improved when the interaction between a local-scale seagrass variable (e.g., shoot biomass) and level of development was included (a multiplicative model). When significant, such interactions were further explored by fit-ting linear mixed effect models between each fish variable and seagrass variable, at different levels of development (Quinn and Keough 2002). R2 values of the fixed part in the mixed models were obtained using the r.squaredGLMM() function from the MuMIn package for R (Barton 2013). Significance levels were set as α = 0.05, and each transect was used as a replicate (N = 60).

To explore the relative importance of the different vari-ables on the multivariate fish assemblage structure (based on densities of individual species), constrained ordination was applied using the vegan package for R (Oksanen et al. 2013). Model selection was done based on AIC scores, with the best fit including four predictor variables: (1) dis-tance to coral reef, (2) wave exposure, (3) level of devel-opment and (4) seagrass shoot biomass. A constrained analysis of proximities (CAP), based on Bray–Curtis dis-similarities, was run and constrained to the four predictor variables. Significance assessment of each term and axis was carried out using a permutation procedure with 999 permutations.

2399Mar Biol (2014) 161:2395–2405

1 3

Results

Site characteristics

A total of 1,182 fish individuals belonging to 52 differ-ent taxa, of which 43 were identified to the species level, were recorded during the study (see families and species list in Appendix 2). These figures are relatively low for seagrass communities compared to previous studies in the WIO region (e.g., Gullström et al. 2008). The average fish density across all 36 random samples was 0.34 ± 0.07

individuals m−2 (when including also the non-random samples, N = 60, the average fish density was 0.39 ± 0.06 individuals m−2), with major differences between sites (Table 1; for data including also non-random samples see Table 1S in Appendix 1). Variability among sites was also observed for most predictor variables, includ-ing depth, distance to coral reefs, distance to mangroves, level of development, wave exposure, seagrass shoot bio-mass, shoot density and epiphytes biomass (Table 2; for data including also non-random samples see Table 2S in Appendix 1).

Table 1 Fish density (individuals m−2) in each of the six study sites including only the six randomly placed sampling points (mean ± SE, n = 6)

Fish variable Chumbe Chwaka Fumba Mbweni Nungwi Prison

Total fish 0.37 ± 0.12 0.59 ± 0.15 0.69 ± 0.28 0.15 ± 0.13 0.20 ± 0.09 0.07 ± 0.02

Juvenile fish 0.26 ± 0.08 0.28 ± 0.10 0.35 ± 0.12 0.11 ± 0.10 0.15 ± 0.10 0.02 ± 0.01

Adult fish 0.10 ± 0.04 0.31 ± 0.11 0.34 ± 0.21 0.04 ± 0.03 0.05 ± 0.02 0.05 ± 0.02

Algal herbivore fish 0 0.04 ± 0.01 0.02 ± 0.01 0 0 0

Seagrass herbivore fish 0.12 ± 0.05 0.06 ± 0.03 0.17 ± 0.08 0 0 0.01 ± 0.00

Invertebrate feeder fish 0.16 ± 0.06 0.17 ± 0.08 0.12 ± 0.03 0.01 ± 0.00 0.01 ± 0.00 0.02 ± 0.01

Invertebrate feeder/piscivore fish 0.05 ± 0.04 0.15 ± 0.07 0.03 ± 0.01 0.01 ± 0.01 0.13 ± 0.10 0.04 ± 0.02

Omnivore fish 0.04 ± 0.03 0.15 ± 0.08 0.32 ± 0.23 0.13 ± 0.13 0.05 ± 0.03 0.01 ± 0.01

Table 2 Predictor variables: depth (m), distance to coral reefs (km), distance to mangroves (km), level of development (number of houses in a 3 km radius), effective fetch (km), seagrass biomass (g DW m−2), shoot density (no. of shoots m−2) and epiphyte biomass (g DW epiphyte/g DW seagrass), at each sampling site including only the six randomly placed sampling points (n = 6)

DW dry weight. Mean values and range (minimum/maximum)

Variables Chumbe Chwaka Fumba Mbweni Nungwi Prison

Depth

Mean 2.9 1.75 3.95 2.4 1.8 1.75

Min/max 2.2/3.5 1.5/2.5 3.4/4.5 1.5/2.75 1.5/2 1.5/2.6

Distance to reef

Mean 0.28 10.37 0.28 1.16 0.92 0.14

Min/max 0.26/0.31 10.4/10.4 0.25/0.31 1.12/1.20 0.87/0.96 0.08/0.18

Distance to mangroves

Mean 5.46 0.82 3.33 0.34 48.5 5.03

Min/max 5.43/5.48 0.79/0.85 3.30/3.34 0.31/0.38 48.5/48.6 4.99/5.04

Level of development

Mean 13 608 0 3,471 0 14

Min/max – 607/609 – 3,439/3,495 – –

Effective fetch

Mean 6.97 0.98 35.95 45.75 117.28 25.60

Min/max 6.79/7.07 0.93/1.01 35.91/35.97 45.70/45.79 117.27/117.3 25.53/25.65

Seagrass shoot biomass

Mean 53.95 179.3 44.32 107.85 6.90 146.24

Min/max 21.12/111.5 0/772.32 0/161.28 33.76/414.9 0/20.8 25.28/546.24

Shoot density

Mean 42.67 178.67 58.67 106.67 0 224

Min/max 0/192 0/576 0/352 0/560 – 0/624

Epiphytes biomass

Mean 0.09 0.32 0.23 0.16 0.09 0.34

Min/max 0.00/0.32 0/1.18 0/0.54 0.02/0.42 0/0.25 0.01/0.72

2400 Mar Biol (2014) 161:2395–2405

1 3

Variables affecting fish density and diversity

Total fish density (pooling all species and life stages) was positively influenced by depth and distance to coral reef, and negatively influenced by level of development (Table 3, Fig. 2). Meanwhile, density of herbivores and density of seagrass herbivores were negatively affected by level of development and wave exposure, and positively influenced

by seagrass biomass and, in the case of seagrass herbi-vores, depth (Table 3, Appendix 1: Fig. 1S). In both cases, there was a significant interaction between seagrass bio-mass and level of development, with seagrass biomass posi-tively affecting fish densities at low levels of development, but this effect weakened as level of development increased, eventually becoming a negative effect at very high levels of development (Appendix 1: Fig. 2S). Density of juveniles

Table 3 Results of linear mixed effect models showing which variables significantly predict fish densities (total fish, herbivore fish, seagrass herbivore fish, juveniles, and adults), species richness and diversity (H′)

Blank cells refer to non-significant models. Predictor coefficients >0.5 or less than −0.5 indicate a strong influence on the response variable. ‘LD × SG’ refers to the interaction between seagrass biomass and level of development that was included in some of the models

Dependent variable Predictor coefficients R2

Depth (D) Distance to coral reef (DC)

Level of development (LD)

Wave exposure (W)

Seagrass biomass (SG)

Interaction (LD × SG)

Total density 0.331 0.449 −0.359 0.27

Herbivores density −0.481 −0.386 0.358 −0.284 0.56

Seagrass herbivores density

0.278 −0.481 −0.268 0.405 −0.357 0.58

Juveniles density −0.593 0.226 0.33

Adults density 0.447 0.317 −0.258 0.22

Species richness 0.585 0.815 −0.509 −0.212 0.59

Species diversity (H′) 0.552 0.874 −0.499 0.173 −0.299 0.53

-1.0

-0.5

0.5

Tot

al fi

sh d

ensi

ty (

cube

roo

t)F

ish

dive

rsity

-1.0

-0.5

0.0

0.5

1.0

10-1-2

Depth (square root) Distance to reef (log10)

-1.0 -0.5 0.0 0.5 1.0

Level of development (log10)

-1.0 -0.5 0.0 0.5 1.0 -1.5

Seagrass biomass (log10)

-1.0 -0.5 0.0 0.5 1.0 1.5-1.5

y = 0.331x + 0.077

y = 0.173x + 0.076y = -0.499x + 0.076y = 0.874x + 0.076y = 0.552x + 0.076

y = -0.359x + 0.077y = 0.449x + 0.077

0.0

1.0

Fig. 2 Conditional effect plots for the predictors of cube root total fish density (ind. m−2) (R2 = 0.27) and fish diversity (Shannon–Weaver diversity index; R2 = 0.53). The variables included are square root (x + 2) depth (m), log10 distance to reef (km), log10 (x + 1) level

of development (no. of houses) and log10 (x + 1) seagrass biomass (g DW m−2). The dotted lines are 95 % confidence intervals. DW dry weight

2401Mar Biol (2014) 161:2395–2405

1 3

was negatively affected by level of development, and posi-tively affected by seagrass biomass (Table 3, Appendix 1: Fig. 3S). Adult fish density was positively affected by both depth and distance to reef, and negatively affected by wave exposure (Table 3, Appendix 1: Fig. 4S).

In terms of diversity, both species richness and Shan-non’s H′ were positively affected by depth and distance to reef, and negatively affected by level of development. Spe-cies richness was also negatively affected by wave exposure (Table 3, Appendix 1: Fig. 5S), while species diversity was positively affected by seagrass biomass (Table 3, Fig. 2). In the case of species diversity, the model included an inter-action term between seagrass biomass and level of devel-opment: There was a strong positive effect of seagrass bio-mass on species diversity at low levels of development, but this effect weakened with increasing level of development, and even became negative at very high levels of develop-ment (Fig. 3).

No significant models were found for any of the other functional groups (algal herbivores, invertebrate feeders, invertebrate and fish feeders, and omnivores).

Variables affecting fish assemblage structure

The multivariate CAP analysis showed that fish assemblage structure was influenced by four variables: level of devel-opment (F = 2.62, p = 0.001), wave exposure (F = 2.42, p = 0.001), distance to coral reef (F = 1.89, p = 0.002) and seagrass biomass (F = 1.55, p = 0.015; Fig. 4a). Constrained ordination allowed the sampling points to be divided into three general clusters. Two of these corre-sponded to individual sites—Chwaka Bay and Mbweni—while the third cluster encompassed the other four sam-pling sites (Fig. 4b). This division was mainly driven by level of development, which was highest in Chwaka Bay

and Mbweni. The density of herbivores was positively affected by seagrass biomass, even though the importance of this predictor seemed to be secondary when consider-ing the whole fish assemblage. Herbivore fish were also divided in two distinct clusters, one comprising algal feed-ers (that were positively influenced by distance to coral reefs), and another comprising three species of seagrass feeders, which were negatively affected by level of devel-opment (Fig. 4a). Most invertebrate and fish feeder species densities were also negatively influenced by level of devel-opment, as well as by distance to coral reefs. For the other functional groups (invertebrate feeders and omnivores), no patterns were found.

Discussion

Our study shows that local- and regional-scale variables alone and together affect seagrass-associated fish assem-blages in seagrass beds around Zanzibar, Tanzania. The level of development and distance to coral reefs, which vary on a regional scale, seem to be two of the most influential variables, whereas seagrass biomass played a significant but subordinate role. Importantly, the positive local effect of seagrass biomass also decreased with increasing level of development, highlighting the fundamental importance of assessing not only the relative but also joint (interactive) effects of local and regional variables for fish communities.

Effects of local-scale variables

Seagrass biomass and water depth positively influenced herbivore and seagrass herbivore fish densities, as well as juvenile fish density and species diversity. These results concur with many earlier studies, showing that fish density

Fig. 3 Conditional effect plot for the interaction between level of development (no. of houses) and seagrass biomass (g DW m−2) as predictors of fish diversity (Shannon–Weaver diversity index). The shaded areas are 95 % confidence intervals. DW dry weight

-1.5 - 1.0 - 0.5 0.0

Seagrass Biomass (Log10)

1.00.5 1.5

-10

12

Fis

h di

vers

ity

Low (-1.5)Medium (0)High (1.5)

Level of Development (log10)

2402 Mar Biol (2014) 161:2395–2405

1 3

can increase with seagrass complexity (e.g., Sogard et al. 1987; Edgar and Shaw 1995) and depth (e.g., Hovel et al. 2002; Gullström et al. 2008). Interestingly, the results con-trast with those of Gullström et al. (2008), who found no effect of seagrass biomass even though they worked in one of the same study areas (Chwaka Bay, see Fig. 1). Sea-grass biomass has in fact been questioned as an appropriate variable in terms of reflecting seagrass complexity, since it only gives a species–area relationship (Attrill et al. 2000). However, in our case, unlike in the study by Gullström et al. (2008), all study sites were located in seagrass beds dominated by the same species, Thalassodendron ciliatum. Therefore, areas with higher seagrass shoot biomass also have a higher structural complexity, which may explain why the present study shows such a strong influence of sea-grass biomass (see also Attrill et al. 2000; Sirota and Hovel 2006).

Importantly, the statistical models on herbivore and seagrass herbivore fish densities, as well as fish diversity, improved when the interaction between seagrass bio-mass and a regional variable—level of development—was included. Interestingly, the positive effect of seagrass biomass decreased as the level of development increased, eventually turning into a negative effect at high levels of development. Since level of development had a strong negative effect on fish densities, regardless of seagrass biomass, it seems that the amount of fish in areas with high level of development was so low that fish density could not respond positively to seagrass. Another possible explanation for the interaction could be a reduced preda-tion risk for smaller fish in heavily fished areas, resulting

in a reduced dependence on seagrass as shelter. This would be in accordance with previous studies showing that, in absence of predators, fish modify their habitat use by using more extensively non-vegetated areas, as well as seagrass areas (Sogard and Olla 1993; Jordan et al. 1997). It has often been suggested that one reason for the lower fidel-ity of seagrass fish to local habitat characteristics could be that seagrass fish species are not as territorial and sedentary as coral reef fish (Jackson et al. 2001). Our results indicate that a complementary explanation could be that many stud-ies of seagrass fish communities have been conducted in fished areas, where fish densities may be too low to respond to variation in seagrass complexity.

In the multivariate analysis, seagrass biomass was the only local-scale variable that had a significant effect. As for the univariate variables, seagrass complexity had a weaker effect than the regional-scale variables and seems to play a secondary role in structuring the fish assemblages. Herbi-vores, however, seem to be more affected by seagrass com-plexity than the other functional groups. Among the her-bivore species, two distinct groups occurred: algal feeders (which were more affected by distance to coral reefs) and seagrass feeders (which were mainly influenced by sea-grass biomass and level of development). Our results con-cur with previous studies showing that different fish species may respond to variables at different scales depending on their life history, feeding behavior and/or predation (e.g., Pittman et al. 2004; Grober-Dunsmore et al. 2008; Gull-ström et al. 2011). Our study also confirms the hypothesis that different groups of organisms are influenced by vari-ables at different scales.

-0.1

0.0

0.1

0.2

CA

P2

-0.1 0.0 0.1 0.2CAP1

Distance to Reef

Level of Development

SeagrassBiomass

WaveExposure

Invertebrates and fish feedersInvertebrates feedersSeagrass herbivores

Algae herbivoresOmnivores

1.0

0.5

0.0

-0.5

-0.5 0.0 0.5 1.0

CA

P2

CAP1

FumbaChwakaNungwi

PrisonMbweni

Chumbe

ba

Fig. 4 Results from the constrained ordination analysis (CAP). a CAP biplot where each dot represents a species classified into func-tional trophic groups. Arrows represent the different predictor vari-ables and point in the direction of increasing values of each variable,

their length indicating their relative importance. b CAP biplot where each dot represents the species composition per sample (transect). Dotted lines connect points belonging to the same sampling site

2403Mar Biol (2014) 161:2395–2405

1 3

Effects of regional-scale variables

Several regional-scale variables strongly influenced fish abundance and diversity. Wave exposure had a negative effect on fish density and species richness [as also shown by Edgar and Shaw (1995), Turner et al. (1999)], most likely due to an increased physical disturbance in the area that could interfere with recruitment (Jenkins et al. 1997), move-ment and feeding (Hovel et al. 2002). Importantly, the level of development also had a strong negative effect on almost all the fish variables considered, and showed the strongest influence on fish assemblage structure among the tested pre-dictor variables, mainly by negatively affecting invertivores and piscivores, as well as seagrass herbivores, even though a few species actually seemed to be favored. Similar pat-terns have been found in previous studies (e.g., McClana-han et al. 1994; de Boer et al. 2001) that detected lower fish abundances and species diversity in heavily fished areas. These results contrast with those from Berkström et al. (2013), who found that spatial variation in fishing pressure within one of our sites (Fumba) had a weak effect on most fish community variables compared to fine- and broad-scale habitat variables. However, our study covered a much wider range of fishing pressure levels and a much larger spatial area, making fish migrations between fished and un-fished areas unlikely. This could in turn explain the stronger influ-ence from fishing pressure detected in this study.

Distance to neighboring habitats was expected to influence fish density due to the high mobility of most fish species and their migratory behavior (Parrish 1989; Mumby et al. 2004). It has previously been shown that fish density and diversity in seagrass beds and mangroves can decrease with distance to coral reefs due to the favorable conditions for fish assem-blages created by the structurally complex coral reef habitat (Dorenbosch et al. 2006b; Unsworth et al. 2008). However, contrary to what had been expected, fish densities increased with distance to coral reefs. A possible explanation could be a reduced predation pressure in seagrass beds located away from coral reefs. Parrish (1989) hypothesized that when seagrass beds are located far from coral reefs, daily migra-tions by reef-associated predators are unlikely, resulting in decreased predation pressure and increased juvenile survival. In line with this hypothesis, we find that densities of most invertebrate feeding and piscivorous fish species decreased with distance to coral reefs (see pattern in Fig. 4b). However, this hypothesis requires additional testing, for example by estimating predation pressure on tethered fish.

Conclusion

Our study suggests that at a regional (across-site) scale, sev-eral factors including the level of human development, wave

exposure and distance to other habitats, structure seagrass-associated fish assemblages (see also Gullström et al. 2008), while at the local scale, seagrass structural complexity plays a significant but weaker role. Further, the fish assemblages are influenced by multiple variables that do not only act at different scales but also interact; the positive effect of sea-grass biomass decreased gradually with the level of human development. These findings highlight the necessity of jointly considering seascape connectivity, landscape con-figuration and seagrass structural complexity in conserva-tion and management plans (see also Unsworth and Cullen 2010; Pittman and Brown 2011; Olds et al. 2012). Seagrass fish communities around Unguja Island appear to be strongly influenced by a combination of natural and anthropogenic variables that operate at different scales and link different ecosystems (here, seagrass beds and coral reefs) in the sea-scape. Consequently, fisheries management strategies should ideally encompass the same scales and acknowledge the role of seascape-level interactions. Such strategies would require a more detailed understanding of the variables that determine species distributions (Pittman and Brown 2011), a type of information that this and similar studies can produce.

Acknowledgments The authors wish to thank the staff at IMS (Institute of Marine Sciences, Zanzibar) for providing research facilities and institutional support; Y. Salmin for valuable assis-tance in the field; all our boat operators in Zanzibar; and two anon-ymous reviewers for useful comments. This research was funded by the Swedish International Development Cooperation Agency (Sida) through their Minor Field Study (MFS) program, and a research grant from the Swedish Research Council (VR/Uforsk, grant number SWE-2012-086).

References

Adams SM (1976) The ecology of eelgrass, Zostera marina (L.), fish communities. I. Structural analysis. J Exp Mar Bio Ecol 22:269–291

Attrill MJ, Strong JA, Rowden AA (2000) Are macroinvertebrate communities influenced by seagrass structural complexity? Ecog-raphy 23:114–121

Baelde P (1990) Differences in the structures of fish assemblages in Thalassia testudinum beds in Guadeloupe, French West Indies, and their ecological significance. Mar Biol 105:163–173

Barton K (2013) MuMIn: multi-model inference. R package version 1.9.5

Beck MW (1997) Inference and generality in ecology: current prob-lems and an experimental solution. Oikos 78:265–273

Bell JD, Westoby M (1986) Importance of local changes in leaf height and density to fish and decapods associated with seagrasses. J Exp Mar Bio Ecol 104:249–274

Berkström C, Gullström M, Lindborg R, Mwandya AW, Yahya SAS, Kautsky N, Nyström M (2012) Exploring “knowns” and “unknowns” in tropical seascape connectivity with insights from East African coral reefs. Estuar Coast Shelf Sci 107:1–21

Berkström C, Lindborg R, Thyresson M, Gullström M (2013) Assess-ing connectivity in a tropical embayment: fish migrations and seascape ecology. Biol Conserv 166:43–53

2404 Mar Biol (2014) 161:2395–2405

1 3

Burnham KP, Anderson DR (2002) Model selection and multimodel inference: a practical information-theoretic approach. Springer, New York

Cullen-Unsworth LC, Nordlund LM, Paddock J, Baker S, McKenzie LJ, Unsworth RKF (2014) Seagrass meadows globally as a cou-pled social-ecological system: implications for human wellbeing. Mar Pollut Bull 83:387–397

de Boer WF, van Schie AMP, Jocene DF, Mabote ABP, Guissamulo A (2001) The impact of artisanal fishery on a tropical intertidal benthic fish community. Environ Biol Fishes 61:213–229

de la Torre-Castro M, Rönnbäck P (2004) Links between humans and seagrasses—an example from tropical East Africa. Ocean Coast Manag 47:361–387

Dorenbosch M, Grol MGG, Nagelkerken I, van der Velde G (2006a) Seagrass beds and mangroves as potential nurseries for the threat-ened Indo-Pacific humphead wrasse, Cheilinus undulatus and Caribbean rainbow parrotfish, Scarus guacamaia. Biol Conserv 129:277–282

Dorenbosch M, Grol MGG, Nagelkerken I, van der Velde G (2006b) Different surrounding landscapes may result in different fish assemblages in East African seagrass beds. Hydrobiologia 563:45–60

Dorenbosch M, Verberk W, Nagelkerken I, van der Velde G (2007) Influence of habitat configuration on connectivity between fish assemblages of Caribbean seagrass beds, mangroves and coral reefs. Mar Ecol Prog Ser 334:103–116

Duarte CM, Chiscano CL (1999) Seagrass biomass and production: a reassessment. Aquat Bot 65:159–174

Duarte CM, Kirkman H (2001) Methods for the measurement of sea-grass abundance and depth distribution. In: Short FT, Coles RG (eds) Global seagrass research methods. Elsevier, Amsterdam, pp 141–153

Edgar GJ, Shaw C (1995) The production and trophic ecology of shallow-water fish assemblages in southern Australia III. General relationships between sediments, seagrasses, invertebrates and fishes. J Exp Mar Biol Ecol 194:107–131

Edgar GJ, Shaw C, Watsona GF, Hammond LS (1994) Comparisons of species richness, size-structure and production of benthos in vegetated and unvegetated habitats in Western Port, Victoria. J Exp Mar Biol Ecol 176:201–226

Eklöf JS, de la Torre-Castro M, Nilsson C, Rönnbäck P (2006) How do seaweed farms influence local fishery catches in a seagrass-dominated setting in Chwaka Bay, Zanzibar? Aquat Living Resour 19:137–147

Froese R, Pauly D (2013) FishBase. www.fishbase.orgGell FR, Whittington MW (2002) Diversity of fishes in seagrass beds

in the Quirimba Archipelago, northern Mozambique. Mar Freshw Res 53:115–121

Grace JB, Bollen KA (2005) Interpreting the results from multiple regression and structural equation models. Bull Ecol Soc Am 86:283–295

Grober-Dunsmore R, Frazer TK, Beets JP, Lindberg WJ, Zwick P, Funicelli NA (2008) Influence of landscape structure on reef fish assemblages. Landsc Ecol 23:37–53

Gullström M, de la Torre Castro M, Bandeira SO, Björk M, Dahlberg M, Kautsky N, Rönnbäck P, Öhman MC (2002) Seagrass ecosys-tems in the Western Indian Ocean. Ambio 31:588–596

Gullström M, Bodin M, Nilsson PG, Öhman MC (2008) Seagrass structural complexity and landscape configuration as determi-nants of tropical fish assemblage composition. Mar Ecol Prog Ser 363:241–255

Gullström M, Berkström C, Öhman MC, Bodin M, Dahlberg M (2011) Scale-dependent patterns of variability of a grazing par-rotfish (Leptoscarus vaigiensis) in a tropical seagrass-dominated seascape. Mar Biol 158:1483–1495

Gullström M, Dorenbosch M, Lugendo BR, Mwandya AW, Mgaya YD, Berkström C (2012) Connectivity and nursery function of

shallow-water habitats in Chwaka Bay. In: de la Torre-Castro M, Lyimo TJ (eds) People, nature and research: past, present and future of Chwaka Bay, Zanzibar. WIOMSA, Zanzibar Town, pp 175–192

Håkanson L, Jansson M (1983) Principles of lake sedimentology. Springer, Berlin

Heck KL Jr, Orth RJ (2006) Predation in seagrass beds. In: Larkum AWD, Orth RJ, Duarte CM (eds) Seagrasses: biology, ecology and conservation. Springer, Netherlands, pp 537–550

Hill J, Wilkinson C (2004) Methods for ecological monitoring of coral reefs—a resource for managers, version 1. Australian Insti-tute of Marine Science, Townsville, Australia

Hovel KA, Fonseca MS, Myer DL, Kenworthy WJ, Whitfield PE (2002) Effects of seagrass landscape structure, structural com-plexity and hydrodynamic regime on macrofaunal densities in North Carolina seagrass beds. Mar Ecol Prog Ser 243:11–24

Jackson EL, Rowden AA, Attrill MJ, Bossey SJ, Jones MB (2001) The importance of seagrass beds as a habitat for fishery species. Oceanogr Mar Biol an Annu Rev 39:269–303

Jacquet JL, Zeller D (2007) Putting the “United” in the United Repub-lic of Tanzania: reconstructing marine fisheries catches. In: Zel-ler D, Pauly D (eds) Reconstruction of marine fisheries catches for key countries and regions (1950–2005). Fisheries Centre Research Reports 15(2), Fisheries Centre, University of British Columbia, pp 49–60

Jelbart JE, Ross PM, Connolly RM (2007) Fish assemblages in sea-grass beds are influenced by the proximity of mangrove forests. Mar Biol 150:993–1002

Jenkins GP, Black KP, Wheatley MJ, Hatton DN (1997) Temporal and spatial variability in recruitment of a temperate, seagrass-associ-ated fish is largely determined by physical processes in the pre- and post-settlement phases. Mar Ecol Prog Ser 148:23–35

Jiddawi NS, Öhman MC (2002) Marine fisheries in Tanzania. Ambio 31:518–527

Jordan F, Bartolini M, Nelson C, Patterson PE, Soulen HL (1997) Risk of predation affects habitat selection by the pinfish Lagodon rhomboides (Linnaeus). J Exp Mar Bio Ecol 208:45–56

Lokrantz J, Nyström M, Norström AV, Folke C, Cinner JE (2010) Impacts of artisanal fishing on key functional groups and the potential vulnerability of coral reefs. Environ Conserv 36:327–337

McClanahan TR, Nugues M, Mwachireya S (1994) Fish and sea urchin herbivory and competition in Kenyan coral reef lagoons: the role of reef management. J Exp Mar Bio Ecol 184:237–254

Mumby PJ, Edwards AJ, Arias-González JE, Lindeman KC, Black-well PG, Gall A, Gorczynska MI, Harborne AR, Pescod CL, Ren-ken H, Wabnitz CCC, Llewellyn G (2004) Mangroves enhance the biomass of coral reef fish communities in the Caribbean. Nature 427:533–536

Nagelkerken I, van der Velde G (2002) Do non-estuarine mangroves harbour higher densities of juvenile fish than adjacent shallow-water and coral reef habitats in Curaçao (Netherlands Antilles)? Mar Ecol Prog Ser 245:191–204

Nagelkerken I, van der Velde G, Gorissen MW, Meijer GJ, van’t Hof T, den Hartog C (2000) Importance of mangroves, seagrass beds and the shallow coral reef as a nursery for important coral reef fishes, using a visual census technique. Estuar Coast Shelf Sci 51:31–44

Nordlund L, Erlandsson J, de la Torre-Castro M, Jiddawi N (2010) Changes in an East African social-ecological seagrass system: invertebrate harvesting affecting species composition and local livelihood. Aquat Living Resour 23:399–416

Nordlund LM, de la Torre-Castro M, Erlandsson J, Conand C, Muth-iga N, Jiddawi N, Gullström M (2013) Intertidal zone manage-ment in the Western Indian Ocean: assessing current status and future possibilities using expert opinions. Ambio. doi:10.1007/s13280-013-0465-8

2405Mar Biol (2014) 161:2395–2405

1 3

Ogden JC (1988) The influence of adjacent systems on the structure and function of coral reefs. In: Proceedings of 6th international coral reef symposium, vol 1, pp 123–129

Oksanen J, Blanchet FG, Kindt R, Legendre P, Minchin PR, O’Hara RB, Simpson GL, Solymos P, Stevens MHH, Wagner H (2013) Vegan: community ecology package. R package version 2.0-9

Olds AD, Connolly RM, Pitt KA, Maxwell PS (2012) Primacy of seascape connectivity effects in structuring coral reef fish assem-blages. Mar Ecol Prog Ser 462:191–203

Orth RJ, Heck KL Jr, van Montfrans J (1984) Faunal communities in seagrass beds: a review of the influence of plant structure and prey characteristics on predator–prey relationships. Estuaries 7:339

Orth RJ, Carruthers TJB, Dennison WC, Duarte CM, Fourqurean JW, Heck KL Jr, Hughes AR, Kendrick GA, Kenworthy WJ, Olyarnik S, Short FT, Waycott M, Williams SL (2006) A global crisis for seagrass ecosystems. Bioscience 56:987

Parrish JD (1989) Fish communities of interacting shallow-water hab-itats in tropical oceanic regions. Mar Ecol Prog Ser 58:143–160

Pinheiro J, Bates D, DebRoy S, Sarkar D, R Development Core Team (2013) nlme: linear and nonlinear mixed effects models. R pack-age version 3.1–111

Pittman SJ, Brown KA (2011) Multi-scale approach for predicting fish species distributions across coral reef seascapes. PLoS ONE 6:e20583

Pittman SJ, McAlpine CA, Pittman KM (2004) Linking fish and prawns to their environment: a hierarchical landscape approach. Mar Ecol Prog Ser 283:233–254

Pittman SJ, Kneib RT, Simenstad CA (2011) Practicing coastal sea-scape ecology. Mar Ecol Prog Ser 427:187–190

Pollard DA (1984) A review of ecological studies on seagrass—fish communities, with particular reference to recent studies in Aus-tralia. Aquat Bot 18:3–42

Quinn GP, Keough MJ (2002) Experimental design and data analysis for biologists. Cambridge University Press, Cambridge

R Core Team (2013) R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria

Ralph PJ, Tomasko D, Moore K, Seddon S, Macinnis-Ng CMO (2006) Human impacts on seagrasses: eutrophication, sedimentation, and

contamination. In: Larkum AWD, Orth RJ, Duarte CM (eds) Sea-grasses: biology, ecology and conservation. Springer, Netherlands, pp 567–593

Sirota L, Hovel KA (2006) Simulated eelgrass Zostera marina struc-tural complexity: effects of shoot length, shoot density, and sur-face area on the epifaunal community of San Diego Bay, Califor-nia, USA. Mar Ecol Prog Ser 326:115–131

Sogard SM, Olla BL (1993) The influence of predator presence on utilization of artificial seagrass habitats by juvenile walleye pol-lock, Theragra chalcogramma. Environ Biol Fishes 37:57–65

Sogard SM, Powell GVN, Holmquist JG (1987) Epibenthic fish com-munities on Florida Bay banks: relations with physical param-eters and seagrass cover. Mar Ecol Prog Ser 40:25–39

Stewart KR, Lewison RL, Dunn DC, Bjorkland RH, Kelez S, Halpin PN, Crowder LB (2010) Characterizing fishing effort and spatial extent of coastal fisheries. PLoS ONE 5:e14451

Turner SJ, Hewitt JE, Wilkinson MR, Morrisey DJ, Thrush SF, Cum-mings VJ, Funnell G (1999) Seagrass patches and landscapes: the influence of wind-wave dynamics and hierarchical arrangements of spatial structure on macrofaunal seagrass communities. Estuar-ies 22:1016

UNEP (2001) Eastern Africa atlas of coastal resources—Tanzania. United Nations Environment Programme, Nairobi

Unsworth RKF, Cullen LC (2010) Recognising the necessity for Indo-Pacific seagrass conservation. Conserv Lett 3:63–73

Unsworth RKF, Wylie E, Smith DJ, Bell JJ (2007) Diel trophic struc-turing of seagrass bed fish assemblages in the Wakatobi Marine National Park, Indonesia. Estuar Coast Shelf Sci 72:81–88

Unsworth RKF, Salinas de León P, Garrard SL, Jompa J, Smith DJ, Bell JJ (2008) High connectivity of Indo-Pacific seagrass fish assemblages with mangrove and coral reef habitats. Mar Ecol Prog Ser 353:213–224

Waycott M, Duarte CM, Carruthers TJB, Orth RJ, Dennison WC, Olyarnike S, Calladine A, Fourqurean JW, Heck KL Jr, Hughes AR, Kendrick GA, Kenworthy WJ, Short FT, Williams SL (2009) Accelerating loss of seagrasses across the globe threatens coastal ecosystems. PNAS 106:12377–12381