Tropical seagrass fish assemblage
composition: importance of edge effect and seascape context
Gustav Palmqvist
Department of Ecology, Environment and Plant Sciences
Masters Degree 60 HE credits
Marine Ecology
Autumn term 2012/Spring term 2013
Main supervisor: Martin Gullström
Co-supervisors: Regina Lindborg, Lina Mtwana Nordlund, Narriman
Jiddawi and Anders Knudby
Page 2 of 33
Tropical seagrass fish assemblage
composition: importance of edge effect and seascape context
Gustav Palmqvist
Abstract
In this study, variations in fish assemblage composition were studied with a unique combination of multiple
spatial scales and detailed in situ measurements as well as a comparison of interiors and edges of
Thalassodendron ciliatum seagrass beds in Zanzibar, Tanzania. Findings successfully show how different fish
response variables vary on different spatial scales. The variables included abundance of adult and juvenile fish,
fish species diversity, species composition and size distribution. Predictors included seagrass structural
complexity, habitat configuration in the beds’ edges and large-scale seascape configuration using satellite
images. Juvenile fish and fish species diversity increased with structural complexity of interiors of T. ciliatum
beds. Further, habitat configuration in the edges of T. ciliatum beds, especially coral habitat and overall
heterogeneity of habitats, affected adult fish abundance and fish species diversity positively in the edges of beds,
likely due to adults’ larger home ranges and/or more complex migratory habits compared to juvenile fish.
Seascape configuration affected fish assemblages via its ability to affect overall connectivity in the seascape,
where distance between land and deep water is short; adult fish are more abundant compared to large shallow
areas with large continuous habitat patches. Multivariate analyses of species composition are well in line with
results from abundance and diversity. With a unique combination of width in spatial scales and detail in
measurements, this study has improved the understanding of what is determining fish assemblage composition.
Furthermore, the study yields a holistic view and has a high degree of general applicability, making it interesting
for further development, and potentially useful for research as well as local management.
Keywords
Seascape; Tropical; East Africa; Fish; Seagrass; Edge effect; Connectivity; Landscape ecology; Remote sensing
Page 3 of 33
Contents
Introduction............................................................................................... 4
Material and methods ................................................................................. 6
Study area .............................................................................................. 6
Focal habitat ........................................................................................... 7
Field data collection ................................................................................. 7
Fish census .......................................................................................... 8
Collection of seagrass characteristics ....................................................... 8
Edge habitat mapping ........................................................................... 9
Benthic substrate mapping ..................................................................... 9
Data analysis ........................................................................................ 10
Results .................................................................................................... 11
Effects of seagrass structural complexity .................................................. 11
Effects of edge habitat and heterogeneity ................................................. 12
Effects of seascape habitat configuration .................................................. 13
Multivariate analyses of fish species composition ....................................... 14
Discussion ............................................................................................... 15
Effects of seagrass structural complexity .................................................. 15
Effects of edge habitat and heterogeneity ................................................. 15
Effects of seascape habitat configuration .................................................. 16
Biases in the method ............................................................................. 17
Impacts of the study .............................................................................. 18
Conclusions .......................................................................................... 19
Acknowledgements ................................................................................... 19
References .............................................................................................. 20
Appendix 1: Fish species list ...................................................................... 24
Appendix 2: Benthic substrate map ............................................................ 27
Northern region ..................................................................................... 29
Western region ...................................................................................... 30
Eastern region (Chwaka Bay) .................................................................. 31
Southern region (Menai Bay) .................................................................. 32
Page 4 of 33
Introduction
Tropical marine shallow waters are one of the most biologically diverse areas on the planet (Moberg
and Folke, 1999), with around 30 % of the entire world’s fish species found on coral reefs (McAllister,
1991) and huge areas (e.g. the Caribbean and Malay archipelagos, Micronesia and Polynesia)
classified as global “biodiversity hotspots” (Cincotta et al., 2000). The so called “tropical seascape” is
a complex mosaic of habitat patches; the three main types being the mangrove forests close to shore,
coral reefs bordering the open sea and seagrass beds in the lagoon in between (Ogden, 1988). These
shallow-water habitats exist in a kind of equilibrium where mangroves and seagrass beds act as a
shield from fresh-water runoff and a sink for the organic material that it brings, creating oligotrophic
and clear waters that benefit coral reefs (Moberg and Folke, 1999; Harborne et al., 2006). The coral
reefs in turn offer an effective protection against waves and currents from the ocean, and create a calm
environment needed for mangroves and seagrasses to grow successfully (Moberg and Folke, 1999;
Harborne et al., 2006).
Seagrasses often form dense beds that cover extensive areas in the shallow-water seascape, and are
among the most productive habitats in the biosphere (Duarte and Chiscano, 1999). Seagrass beds are
often described as “ecosystem engineers”, as they constitute a large structure that slow down water
movements, increase sedimentation rates, and stabilize the sediment (Bos et al., 2007). In addition,
seagrass beds are structurally complex systems with high biodiversity (Gullström et al., 2002; Boström
et al., 2006). A seagrass bed is most often dominated by only one or a few seagrass species and
diversity is mainly made up of faunal communities and algal flora inside the bed (Hemminga and
Duarte, 2000). However, seagrass systems may be made up of a patchwork of different types of beds,
and thus on a large scale, a seascape may exhibit a relatively high diversity also of seagrass species
(Boström et al., 2006). Seagrass beds constitute the main habitat for many species of fish and
invertebrates and play a key role at some points in the life cycle or as a part of a multi-habitat use of
even more species (Parrish, 1989; Boström et al., 2006; Nagelkerken, 2009a; Berkström, 2012). They
are important as nursery grounds, hideout and feeding grounds for many fish and invertebrate species
and a key medium for safe movements across the seascape as a whole (Boström et al. 2006;
Nagelkerken, 2009a; Berkström et al., 2012).
Fish assemblages of seagrass beds vary due to many different factors at different scales (Boström et
al., 2006; Conolly and Hindell, 2006; Dorenbosch, 2006; Gullström et al., 2008, 2011). At a small
scale, physical complexity (e.g. canopy height or shoot density) of seagrass beds has been suggested to
increase abundance of many fish species (Orth et al., 1984; Boström et al., 2006) due to changes in
food-availability and hiding-possibilities for prey species. This pattern has been showed to be
especially important for juvenile fish, thus attributing seagrass beds as nursery grounds (Nagelkerken
et al., 2000; Dorenbosch, 2006; Dorenbosch et al., 2007; Gullström et al., 2008; Nagelkerken, 2009a).
However, while many studies have found seagrass beds to be more important nursery grounds than
unvegetated areas, few have found differences between seagrass beds and other structurally complex
habitats such as oyster reefs or macroalgal belts (Heck et al., 2003). It may be that structure per se, not
seagrass specifically, increases juvenile abundance.
Fish assemblage composition in seagrass beds have also been studied at larger seascape scales.
Variations have been suggested to depend on factors such as seagrass patch size and shape (Salita et
al., 2003), seagrass beds’ spatial relationship to nearby coral reefs and coastline shape (Dorenbosch et
al., 2006a, 2006b) and seagrass beds’ distance to nearby mangroves and coral reefs (Gullström et al.,
Page 5 of 33
2008, 2011; Unsworth et al., 2008). Distribution, size and spatial arrangement of seagrass beds have
also been studied as explanatory variables for fish assemblage composition on coral reefs (Grober-
Dunsmore et al., 2006; Dorenbosch et al., 2007; Berkström et al., 2013). The effect of such seascape-
scale factors varies depending on focal species and focal habitat, in particular the species’ home range,
distance and frequency of migrations, and degree of success of ontogenetic shifts, and also with
temporal factors such as tidal movements, seasonal oceanographic and climatic variations. Even the
relative importance of seagrass beds as nurseries has been suggested to depend on surrounding
seascape configuration (Dorenbosch et al., 2007), emphasizing the complexity of the seascape and the
importance of site-specific context. When studied at a seascape scale, ecological connectivity is often
mentioned as an underlying mechanism for variations in fish assemblage composition. Ecological
connectivity is a concept derived from terrestrial landscape ecology that aims at mapping and
describing processes interlinking habitat patches in a landscape mosaic (Sheaves, 2009). Connectivity
links take form in biogeochemical processes, flows of detritus, planktonic transportation of eggs, seeds
and larvae, as well as active migration of organisms for e.g. foraging, predator-prey interactions,
reproduction, egg-laying or permanent ontogenetic shifts in preferred habitat (Nagelkerken, 2009b).
For example, connectivity links between seagrass beds, mangroves and coral reefs give rise to more
planktivore- and herbivore-dominated fish assemblages in seagrass beds close to mangroves, and more
omnivore- and predator-dominated assemblages in seagrass beds close to coral reefs (Unsworth et al.,
2008). Furthermore, connectivity links in the form of egg-laying, nursery habitat use and ontogenetic
shifts, creates higher abundances of these specific fish species on coral reefs closer to seagrass beds
and mangroves (Dorenbosch et al., 2004; Berkström et al., 2013). In contrast, seagrass beds in closer
proximity to coral reefs harbor higher abundances of reef fish species that utilizes seagrass beds as
nurseries compared to seagrass beds farther away from reefs (Dorenbosch et al., 2006b).
East African seagrass beds are extensively used by local populations as a resource for food and income
through fisheries, collection of invertebrates and substrate for seaweed farming (Gullström et al.,
2002; de la Torre-Castro and Rönnbäck, 2004). Seagrass beds are important fishing grounds in
themselves but also increase coral reef fish production through their function as nurseries and foraging
grounds for reef fish (Unsworth and Cullen, 2010). Apart from edibles, collection includes fishing-
bait, manure, medicine and curios for sale to tourists, and seagrass beds have also been showed to be
important for the fulfillment of aesthetical, cultural and spiritual needs (de la Torre-Castro and
Rönnbäck, 2004). In addition, seagrass beds’ sediment stabilization provides clearer waters and makes
them efficient nutrient and carbon sinks (Bos et al., 2007; Kennedy et al., 2010).
However, seagrasses worldwide suffer major declines due to anthropogenic pressures, with as much as
7 % of its global area per year (Unsworth and Cullen, 2010). Threats directly affecting seagrass beds
are mainly eutrophication due to human settlements and development (Gullström et al., 2012). Small-
scale anthropogenic stressors include destructive fishing methods such as drag-net fishing and clearing
of seagrass to make way for seaweed farming (de la Torre-Castro and Lindström, 2010). Also, extreme
outbreaks of sea urchins can locally cause overgrazing of seagrass. The underlying reasons of these
outbreaks have not been fully investigated, but have so far been linked to the overfishing of predatory
fish that feed on sea urchins (Eklöf et al., 2009).
Because of the major ecological as well as social and economical services seagrass ecosystems
provide, their health and resilience must be considered highly important (Unsworth and Cullen, 2010).
A study design that includes a landscape scale is crucial to holistically understand ecosystem
processes, including connectivity. The application of terrestrial landscape ecological ideas on marine
systems was pioneered by Robbins and Bell (1994), and gave rise to the field of “seascape ecology”.
Even though seascape ecology is rising as topic, conclusive evidence is still scarce and results often
Page 6 of 33
contradictory. The effects of seascape configuration and connectivity on fish assemblage composition
have been severely under-studied (this issue is discussed in several reviews; e.g. Boström et al. (2006),
Connolly and Hindell (2006), Grober-Dunsmore et al. (2009), Berkström et al. (2012)). Understanding
of ecological connectivity is currently one of the most critical issues for conservation managers to
make meaningful management decisions. However, oddly enough, it is one of the least understood
concepts in the marine environment (Grober-Dunsmore et al., 2009).
In the present study, I aimed at investigating how tropical fish assemblages vary in middles compared
to edges of seagrass beds and in relation to multiple factors at different spatial scales, including
seagrass structural complexity, habitat composition in the direct vicinity of the seagrass beds and
configuration of the surrounding seascape. Fish assemblages were characterized by measuring life-
stage composition and abundance of adult and juvenile fish, fish species diversity and fish species
composition. The focal seagrass species was Thalassodendron ciliatum.
Material and methods
Study area
The present study was conducted in the shallow waters around Zanzibar Island between November
2012 and February 2013. The Zanzibar archipelago is located just off the east African coast in the
western Indian Ocean. It consists of two main islands; Unguja (also known as Zanzibar Island) and
Pemba, as well as many smaller islets, and is a semi-autonomous region of the United Republic of
Tanzania. Zanzibar Island is oblong and north-south oriented, approximately 85 km long and 35 km
wide, with an area of 1530 km2 (Shaghude, 2001). The climate is tropical with two main seasons; the
clear, calm and warm north east monsoon from November to April, and the rainy, windy and cool
south east monsoon from March to October (McClanahan, 1988). Zanzibar Island exhibits
archetypical tropical seascapes (Berkström et al., 2012) and its coastline is densely dotted by small
rural villages, which makes it well suited as a study area for ecological or socio-ecological studies
with a seascape perspective.
The seascape plays a vital role in the day-to-day life of many Zanzibaris, and fisheries are
exceptionally important. There are estimated to exist about 20 000 full-time fishers along the entire
coastline of mainland Tanzania, while along the considerably shorter coastline of Zanzibar, the
estimation concluded about 23 000 (Jiddawi and Öhman, 2002). Seagrass beds are habitat for many
species of high commercial value, the most important fish families being rabbitfish (Siganidae),
parrotfish (Scaridae), emperors (Lethrinidae) and surgeonfish (Acanthuridae) (Jiddawi and Stanley,
1997; Jiddawi and Öhman, 2002). Fishers use traditional canoes or sailing boats (“Ngalawa” or
“Dhow”) and gillnets, handlines and/or bottom-lying basket traps (“Dema”) (Gullström et al. 2002;
Jiddawi and Öhman, 2002). Invertebrates are also collected, either by swimming equipped with spears,
or walking in the intertidal zone at low tide; selecting cockles, cowries, sea cucumbers, sea stars and
octopus, among others, or small fish trapped in intertidal pools (Gullström et al., 2002; Jiddawi and
Öhman, 2002). Walking collectors have been shown to prefer larger, denser and more continuous
seagrass beds (Nordund et al., 2010). In addition to fishing and collection, seagrass areas are
sometimes used as substrate for seaweed farms. The seaweeds cultivated are brown algae, later sold
for production of carageenans (de la Torre-Castro and Rönnbäck, 2004).
Page 7 of 33
Focal habitat
The seagrass species Thalassodendron ciliatum (henceforth shortened “Tc”) was chosen as the focal
habitat of this study because of its great importance as habitat for fish and invertebrates, thus also for
seascape connectivity and consequently for local fisheries. Tc is a large species of seagrass that often
form tall, dense and extensive beds that are common in the region. Edges of the beds are sharp and
easy to define, which allows for advanced method designs. Fifteen large Tc dominated areas were
chosen as sampling areas (henceforth called “seascapes”), scattered all around Zanzibar Island (see
Figure 1). To standardize the selection of sampling areas, all beds should at some point in the tidal
cycle be at least 3 m deep, and the centroids of each seascape at least 500 meters apart. Apart from
this, the beds varied heavily in physical structure and the seascapes varied heavily in terms of habitat
configuration, bathymetry and oceanographic conditions.
Figure 1. Map of Zanzibar Island and the locations of the 15 seascapes (stars).
Field data collection
The data-collection was divided into four parts; (i) fish census, (ii) collection of within-patch
variables, (iii) edge habitat mapping, and (iv), the creation of a benthic substrate map. The fish census
served to collect the focal response variables; fish species- and life stage composition, species
diversity and abundance. The within-patch variables comprised the bed’s physical structure and served
as small-scale predictor variables. The edge habitat mapping served as a set of high-detailed medium-
scale predictor variables, while the benthic substrate map had a lower spatial resolution but covered a
larger area.
Page 8 of 33
Fish census
In each seascape, twelve belt transects of 25 m in length were placed in the Tc bed, by means of a
“structurized randomness”-design, meaning complete random placement, but within a set of pre-set
conditions. Transects were placed at least 10 m apart from each other, and in as continuous Tc beds as
possible (avoiding too patchy areas). Six transects were placed in the middle of the bed and the
remaining six transects in the edge. “Middle transects” were placed at least 10 m “inside” the edge of
the bed while “edge transects” were angled parallel to the edge, at a maximum of 3 m from the edge.
Some Tc areas consisted of many small patches instead of continuous beds. In these cases, it was
difficult to place middle transects in continuous Tc and to keep them away from the edges. Edge
transects suffered from inaccuracy as well when the Tc was patchy, or when edges undulated.
However, if disregarding patchiness and only considering the overall “Tc dominated area”, conditions
for transect placement were always met.
Visual census was carried out with one observer swimming on the surface along a transect, observing
all fish within 2 m of each side of the transect, totaling a 25 x 4 m belt (100 m2) during 5 minutes. The
observer swam down to the seagrass canopy two to three times in each transect and/or when a closer
observation was needed for identification. All fish were counted, identified to the lowest taxonomic
level possible and total length estimated using ten size classes; 0<4 cm, 4<8 cm, 8<12 cm, 12<16 cm,
16<20 cm, 20<30 cm, 30<40 cm, 40<50 cm, 50<60 cm and 60< cm. When large schools were
encountered, taxonomy, size and numbers were estimated visually. Censusing was carried out at a
shallow depth, aiming for 3 m, but in the end varying between 2 and 5 m. However, 3.0 ± 0.5 m were
true in more than 80 % of the cases. Because of the near constant depth, but varying topography,
transects were censused at different times in the tidal cycle. To avoid the strongest tidal currents,
censuses were only carried out during days with a difference between high and low tide of <3 m. Two
well coordinated observers worked together and performed half of the fish censuses each.
Coordinates were collected at both ends of each transect using a GPS and a compact camera with
built-in GPS. Coordinates were stored in (and when needed converted to) UTM UPS. Coordinates
with obvious GPS-inaccuracy were removed.
Information was collected from Froese and Pauly (2005) regarding maximum total length (in a few
cases standard length was used due to lack of information) in centimeters for each fish species.
According to Nagelkerken and van der Velde (2002), fish of < ⅓ of maximum total length can be
considered juveniles, fish of ⅓ < ⅔ of maximum total length subadults, and fish of > ⅔ of maximum
total length adults. In this study, however, subadults and adults were merged into a single adult
category. When the juvenile-adult-border was located “inside” one of the size classes used in the
census, juvenile-adult proportions were calculated, assuming even distribution of sizes within the size
class range. For example, if a species’ juvenile-adult border was 9 cm, it belonged to the 8<12 cm size
class that had a range of four one-cm classes (8<9, 9<10, 10<11 and 11<12); thus, 25 % of all fish in
this size class were considered juveniles and 75 % adults.
Collection of seagrass characteristics
Along each transect, seagrass canopy height and shoot density were measured. For canopy height, ten
subsamples were collected in each transect consisting of random Tc plants measured from sediment to
the tip of the leaves, plant stretched. The two highest values were excluded and the final estimate of
Page 9 of 33
the transect was the mean of the eight remaining subsamples. Shoot density was measured by
randomly (but always inside the bed) placing a single 25 x 25 cm quadrate in each transect, and each
individual living shoot in the quadrate counted. Branching shoots were counted as one shoot. Canopy
height was consistently measured by one and the same observer, while shoot density was consistently
measured by the other observer.
Edge habitat mapping
For each edge transect, edge habitat was mapped, using six subsamples in the form of visual 1 x 1 m
quadrates evenly distributed (at 0, 5, 10, 15, 20 and 25 m) along the transect. From the transect, the
quadrates were moved perpendicular to the nearest edge of the Tc bed. The quadrates bordered the bed
with one of their sides, but did not contain any Tc. The habitat composition of each quadrate was
described by means of a pre-defined set of habitat classes (algae, coral, coral rubble, pavement, rock,
sand and seagrass) and coverage percentages (5, 10, 25, 50, 75, 90 and 100). From the subsamples,
mean percentage cover of respective habitat class was calculated for each seascape, and one
dominating habitat was assigned. The edge habitat mapping was consistently carried out by the same
observer. Furthermore, edge habitat heterogeneity was estimated for each transect, using the habitat
coverage values. An even mix of habitats resulted in high heterogeneity while a dominating single
habitat resulted in low heterogeneity. This was done by subtracting the highest habitat coverage value
from a constant (the number 100).
Benthic substrate mapping
A benthic habitat map was created over the entire Zanzibar Island using a random forest classifier
(Breiman, 2001; Liaw and Wiener, 2002). Field data was collected in the form of coordinate points in
UTM UPS paired with a classification of the habitat, where it should be similar approximately 15 m in
all directions. In addition, photos of the habitat were taken, depth was measured, and time of day and
date were noted to be able to calculate water depth at the coordinates at the time when the satellite
imagery was taken. Nine habitat classes were used and divided into two groups; the water-classes:
algae, coral, sparse seagrass (10<40 % cover), dense seagrass (>40 % cover), deep water, pavement
(all consolidated, unvegetated substrates) and sand (all unconsolidated, unvegetated substrates), and
the land-classes: mangrove forest and terrestrial vegetation. The data collection was distributed over
the entire Zanzibar Island between November 2012 and February 2013, and 721 data points were
collected.
An existing data set collected using the same principles in September to December 2007 (Knudby et
al., 2010) was added to increase the sample size. These data points were visually compared to the
current Landsat images (from 2011-2012), and areas where changes have occurred since the data
collection were removed (remaining n = 391). Finally, additional data were derived from visual
interpretation for undersampled classes that were easy to identify in the Landsat images (Deep water: n
= 113; Land: n = 30; Sand: n = 54). This gave a final sample size of 1309.
Twenty Landsat 7 images taken 2011-2012 were used for the random forest classification. The first
step was to identify five basic classes in all 20 images; cloud, shadow of cloud, NA-data (Not
available data), land and water. The next step was to add the field data points onto all 20 images to
produce a more detailed classification. Spectral features (such as color and brightness) were extracted
Page 10 of 33
from pixels with an overlapping field data point. A water classifier was trained using all the data
points that had been both observed as a water class, and initially classified as water. This classifier was
then applied to all pixels initially classified as water. A similar process was used to classify all pixels
initially classified as land. This was repeated for all 20 images. All pixels were thus classified 20
times, and the most frequently recurring class except cloud, shadow or NA, was the class the pixel
finally was classified as.
For the geographic analyses, ESRI ArcMap 10.1 was used and the benthic substrate map was imported
as a raster. The coordinates from the fish census were imported as point shape files and for each
seascape, a mean coordinate was calculated. Around the mean coordinate points, circular polygons of
500 meter radius were generated. These circles determined the spatial extent of the seascapes in the
benthic habitat map. All pixels of each respective pixel value (habitat class) was then counted and
summed within each circle (seascape) and the results translated into new variables (“amount of X in
the seascape”, where X is a given habitat class).
Additionally, a “seascape heterogeneity” variable was created for each seascape by subtracting the
lowest score in any of the pixel counts from the highest, and then subtracting this result from 1000. A
high value meant that the difference between the highest and lowest pixel count was little, and thus the
seascape consisted of an even mix of habitats. A low value meant that the difference between the
highest and lowest pixel count was large and that one or a few habitats dominated the seascape.
The benthic substrate map received a final accuracy of 69 %, and evidently, generation of the map
proved challenging. The main reason being the large variations within, and similarities between,
habitats. For example, the sand and pavement habitats encompassed many different bottom types,
being rarely completely free of seagrass, algae or coral bommies. Further, the seagrass, algae and coral
habitats were often a combination of two or all three habitats, making boundaries between classes fluid
and sometimes hard to distinguish.
Landsat 7 images have a 30 x 30 m spatial resolution. It is possible to argue that this is a too large
grain-size, since large areas of homogenous habitat are rare. On the other hand, with for instance 4 x 4
m resolution (e.g. IKONOS satellite imagery), the risk of mapping wrong pixels due to GPS
inaccuracy will likely increase. In addition, an advantage of using Landsat imagery is the low cost and
relatively quick and easy analyses that follow; important factors for the usefulness of regional and
local management.
The map in its entirety as well as detailed views of all 15 seascapes can be seen in Appendix 2.
Data analysis
All organization of data was made in Microsoft Excel 2007 and all statistical analyses were made in R
2.15.2 with the aid of Tinn-R 2.4.1.5. In R, the packages car, AEG, vegan and MASS were used. In all
analyses, the number of replicates was 15, one data point for each seascape. However, two separate
data sets were created, first using only middle transects to calculate values for all variables, and for the
second data set, only edge transects were used. Throughout analysis, all tests were performed twice,
once with each data set.
Species that were observed in large schools and only roughly estimated were completely removed
from all abundance and species composition analyses. Even species that were often counted
Page 11 of 33
accurately, and only occasionally estimated, were excluded. However, all species were included in
diversity analyses.
The distribution of all variables were explored and some variables transformed; mean abundance of
adult as well as juvenile fish were square rooted. Number of land pixels was square rooted as well; and
number of dense seagrass pixels was logarithm-transformed. Predictor variables were tested in a
pairplot for collinearity. None were found and analyses were carried out intially with multiple
regressions. Though, no significant results could be found, so simple linear regressions were used
instead. Since this meant quite a large number of regressions, a Bonferroni correction was considered,
but rejected since Type I errors were not deemed dangerous enough; even non-significant trends could
prove interesting.
In analyses to test different edge habitat compositions’ effects on fish assemblage structure in the
seagrass beds, data points were grouped based on the dominating edge habitat in the seascape, and fish
response variables compared between groups using ANOVA’s and Tukey HSD post-hoc tests.
A species composition-matrix was created for all seascapes and analyzed with non-metric
multidimensional scaling (nMDS) using the metaMDS function from the vegan package in R. The
metaMDS-function uses the Bray-Curtis dissimilarity distance. Significance of predictor variables was
investigated using the BioEnv, ANOSIM and CCA functions, respectively. Trying to analyze all
predictor variables at once proved overwhelming and no comprehensive results could be extracted. To
solve this, predictor variables were divided into three groups based on scale; local scale (canopy height
and shoot density), edge scale (the edge habitat variables) and seascape scale (the benthic substrate
map variables).
Results
Effects of seagrass structural complexity
No significant patterns were found in any relationships between fish response variables and within-
patch predictor variables, neither in the middle nor in the edge transects. However, although not
significant, in two cases clear trends could be discerned (see Figure 2). Canopy height seems to affect
mean abundance of juvenile fish (p = 0.065, r2 = 0.180) and shoot density seems to be related to mean
number of fish species (p = 0.073, r2 = 0.167). Both relationships were positive and exclusive for the
middles of beds.
Page 12 of 33
Figure 2. Relationships between square-rooted mean abundance of juvenile fish and mean canopy height (cm)
(a); and mean number of fish species and mean shoot density (number of shoots/0.0625m2) (b) in the middle of
Tc beds. Each dot represents a seascape and all values are the mean of the six middle transects in each seascape.
Effects of edge habitat and heterogeneity
Coral-dominated Tc edges showed higher mean abundance of adult fish compared to seagrass-
dominated Tc edges, and higher mean number of fish species compared to seagrass- as well as sand-
dominated Tc edges (Figure 3; Table 1).
Figure 3. Mean abundance of adult fish (a), mean abundance of juvenile fish (b) and mean number of fish
species (c), in edges of Tc beds in seascapes dominated by different edge habitats (algae, coral, sand and
seagrass). Each box represents a group of seascapes, and vary in sample size depending on how many seascapes
that were dominated by the given habitat; algae (n = 3), coral (n = 2), sand (n = 4) and seagrass (n = 6). The bold
lines represent the medians of the groups’ distributions, the boxes the interquartile range and the whiskers the 95
% confidence interval.
Mean abundance of
adult fish
Mean abundance of
juvenile fish
Mean number of fish
species
Full ANOVA model 0.013 0.105 0.020
Post-hoc tests:
Coral - Algae 0.608 0.312 0.118
Sand - Algae 0.515 0.998 0.892
Seagrass - Algae 0.079 0.825 0.669
Sand - Coral 0.102 0.222 0.034
Seagrass - Coral 0.014 0.072 0.014
Seagrass - Sand 0.575 0.875 0.974
Table 1. P-values of full ANOVA models and post-hoc tests (TukeyHSD) of Figure 3. Significant values (p <
0.05) are shown in bold.
The heterogeneity of edge habitats showed a strong positive relationship with mean abundance of
adult fish (p = 0.004, r2 = 0.437; Figure 4a), a significant but clearly weaker positive relationship with
mean abundance of juvenile fish (p = 0.033, r2 = 0.250; Figure 4b) and again a strong positive
relationship with mean number of fish species (p = 0.007, r2 = 0.402; Figure 4c). All fish response
variables are from the edge transects data.
Page 13 of 33
Figure 4. Relationships between mean edge habitat heterogeneity and square-rooted mean abundance of adult
fish (a), square-rooted mean abundance of juvenile fish (b) and mean number of fish species (c), in the edges of
Tc beds. Edge habitat heterogeneity was calculated as 100-X, where X equals the largest value in the edge
habitat mapping (see Material and methods). Each dot represents a seascape and values are the mean of the six
edge transects in each seascape.
Effects of seascape habitat configuration
Some significant relationships were found between fish abundance and seascape habitat classes,
derived from the benthic substrate map. Mean abundance of adult fish in the middle of Tc beds was
related to number of deep water-pixels (p = 0.038, r2 = 0.235; Figure 5a) as well as number of land-
pixels (p = 0.027, r2 = 0.271; Figure 5b). Number of dense seagrass-pixels, on the other hand, showed
a negative relationship, although non-significant, with mean abundance of adult fish in the middles (p
= 0.108; 0.125; Figure 5c).
Figure 5. Relationships between mean abundance of adult fish in the middle of Tc beds, and number of deep
water pixels (a), square-rooted number of land pixels (b) and square-rooted number of dense seagrass pixels (c).
The numbers of pixels are derived from analysis of the benthic substrate map, in which each seascape is a cricle
of 500 m radius. Each dot represents a seascape and abundance-values are the mean of the six middle transects in
each seascape.
Page 14 of 33
Multivariate analyses of fish species composition
Neither the BioENV nor the ANOSIM-analyses of the species composition matrix yielded any
signifcant results. However, all three CCA-models (Figure 6) showed significant p-values (the
seascape model: p = 0.005; Figure 6c; the local- and edge models: p-values = 0.01; Figure 6a-b).
However, when performing ANOVA’s on the CCA-models, shoot density was the only significant
variable (p = 0.015), from any of the models. The CCA plots (Figure 6) show two-dimensional
visualizations of the multivariate analyses of the species composition matrix.
Figure 6. CCA-plots of species occurance similarities (crosses) and seascape species composition similarities
(circles). Distance between points measures similarity. Arrows illustrate how predictor variables affect species
occurance and seascape species composition; directions illustrate similiraties and length measures strength of the
effect. Predictor variables are divided into three spatial scales; local scale: seagrass physical structure (a), edge
scale: edge habitat (b) and seascape scale: seascape configuration (c).
Page 15 of 33
Discussion
This study shows how fish assemblage compositions are clearly affected by multiple predictor
variables on several spatial scales. The strongest relationships were found between edge heterogeneity
and adult fish abundance and fish species diversity, respectively, as well as between seascape habitat
configuration and fish species composition; indicating that large-scale factors are the most important
predictors for fish assemblage compositions.
Juvenile fish abundance showed a positive trend toward higher canopy in the middle of Tc beds, while
fish species diversity shows a positive trend toward denser middles of Tc beds. Adult and juvenile fish
abundance as well as fish species diversity were affected positively by edge heterogeneity. Coral-
dominated edges showed higher abundance of adult fish than seagrass-dominated edges and higher
fish species diversity than both seagrass- and sand-dominated edges. Adult fish abundance was also
positively affected by the amount of land and deep water in the seascape, and showed a negative trend
in relation to the amount of dense seagrass in the seascape.
Effects of seagrass structural complexity
The results showed that there were no overall differences in abundance or diversity of fish between
middle and edges of Tc beds, meaning no significant edge effect was found. Though, clear trends
show that variations in physical structure of the studied seagrass beds may have affected fish
assemblage composition differently in the middles compared to the edges, i.e. a higher Tc canopy
seems to attract higher numbers of juvenile fish in the middles, but not in the edges. Similarly, a
higher Tc shoot density seems to attract more species of fish in the middles, while not in the edges.
Compared to the edges, the middle of Tc beds exhibits sheltered habitats, and higher and denser
seagrass offers more food and better hiding possibilities for many fish species. Predation pressure is
also lower in seagrass beds, compared to deeper, more exposed habitats (Nagelkerken, 2009a). The
findings of this study seem to be in line with such ecological theory, meaning that they are in line with
previous claims of seagrass beds acting as nursery habitats for fish (Nagelkerken et al., 2000;
Dorenbosch et al., 2004; Gullström et al., 2008). Though, this study cannot add proof in any direction
to the question whether seagrass beds per se attract juvenile fish, rather than other structurally
complex habitats. The current findings suggest that more structurally complex middles of Tc beds
likely attract more juvenile fish, and fish species overall, than less structurally complex middles of Tc
beds.
The local scale CCA-plot shows that the canopy height and shoot density arrows create an angle of
approximately 90 degrees, thus together dividing the graph into one half of high structural complexity
(the lower right) and the other half of low structural complexity. This could mean that in addition to
increasing juvenile fish abundance and species diversity, seagrass structural complexity affects fish
species composition in the Tc beds in one coherent direction.
Effects of edge habitat and heterogeneity
This study showed that edges with different habitat compositions exhibit different abundances and
diversity of fish. Specifically, coral-dominated edges were more abundant in adult fish compared to
Page 16 of 33
seagrass-dominated edges, and showed higher fish species diversity than both seagrass- and sand-
dominated edges. The algae edge habitat showed slightly higher fish abundances and diversity
compared to the sand- and seagrass habitats, although there were no significance. Seagrass-, sand- and
algae-dominated edges showed no significant differences between themselves.
Coral bommies and reefs are structurally complex and host highly diverse fish assemblages (Moberg
and Folke, 1999). Likely, presence of coral habitat in the edges of Tc increases abundance and
diversity of fish simply due to a spillover-effect (Dorenbosch et al., 2005). Algae habitats are also
attractive habitat for many species of fish (Ornellas and Coutinho, 1998) and it is possible that a
spillover effect would exist in some algae-dominated edges as well. This kind of spillover effect is
absent from seagrass- and sand habitat, thus making the coral edges more abundant and diverse. An
open space of sand is barren and exposed, making it unattractive and even dangerous for many fish
species. In addition, most species of seagrass found in the edges (except Enhalus acoroides) are
shorter, structurally simpler and offer less hiding-possibilities than Tc, thus also creating an edge that
exposes fish. Another mechanism that may increase abundance and diversity in edges is
“complementary resource distribution” (Ries and Sisk, 2004), where species actively utilize resources
from both habitats and are therefore attracted by the edges per se. The seagrass edge habitat, being
relatively similar to the Tc bed itself in terms of physical structure and faunal communities might lack
a strong complementary resource distribution effect in comparison to coral habitat (and possibly algae
habitat) that differs a lot from the Tc beds in these aspects.
As this study shows, habitat heterogeneity in the seagrass edge zones seems to be important for
abundance of adult and juvenile fish as well as fish species diversity. A more heterogenous edge
habitat configuration generates a more complex habitat with more niches to be filled and thus a higher
abundance and diversity of fish (Gratwicke and Speight, 2005).
In terms of abundance, the patterns of edge habitat and heterogeneity were stronger for adult fish
compared to juvenile fish. Juvenile fish often inhabit shallow, sheltered areas such as seagrass beds
and mangroves, avoiding deep, exposed areas (Parrish, 1989; Watson et al., 2002). As fish matures,
many species increases their home ranges and/or permanently migrate to new habitats, some adopting
a complex multi-habitat use (Berkström et al., 2012). The increased complexity in migratory habits as
fish mature are likely an important explanation to why structurally complex edges affect the
abundance of adult fish in such a strong way, while having a relatively small effect on the abundance
of juvenile fish.
In the edge scale CCA-plot, the sand and seagrass arrows are attuned, and opposed to algae, coral and
edge heterogeneity, meaning these two groups of variables are each other’s antitheses and
determinants for visibly different species compositions. The results in this study show how coral- (and
to a lesser extent algae-) dominated edges and heterogeneous edges exhibit higher abundances and
species diversity of fish compared to seagrass- and sand-dominated and homogenous edges. The CCA-
plot may be interpreted as these two groups of predictor variables affecting not only abundance and
diversity, but fish species compositions as well, in two separate directions.
Effects of seascape habitat configuration
The positive relationships between abundance of adult fish and amount of land and deep water in the
seascape might seem surprising since seagrass-associated fish would logically find land and deep
water highly unattractive. Nonetheless, the underlying mechanisms for this pattern might be explained
Page 17 of 33
by a strong influence of seascape connectivity. It is well known that many fish species exhibit a multi-
habitat use (Berkström et al., 2012) and that close proximity between habitat patches facilitates
migrations and increases connectivity (Grober-Dunsmore et al., 2009). Large proportions of land
and/or deep water in a certain seascape means that the distance between land and fringing reef is short,
and the shallow area and habitat patches are restricted in size. This results in short distances between
habitat patches and high connectivity, which makes this environment relatively risk-free for organisms
to freely move across, and therefore favorable for many fish species, and especially so for adult fish
because of their larger home ranges and more complex multi-habitat use compared to juvenile fish.
The negative trend found between abundance of adult fish and the amount of dense seagrass was non-
significant and exhibited a low r2-value. Nevertheless, this trend could be explained in a similar way as
the coupling between adult fish and amount of land/deep water, with amount of dense seagrass acting
as the antithesis of the amount of land/deep water. In this study, seascapes with large proportions of
dense seagrass were generally characterized by a large shallow area, relatively far from either land or
the deep sea. These seascapes often displayed large, continuous habitat patches and movements
between them are likely time-consuming, costly and risky for organisms (Grober-Dunsmore et al.,
2009). However, whether amount of seagrass-pixels can be used as a generally applicable indicator for
large shallow-water seascapes and a measure of low connectivity, or if this pattern was a unique
finding, cannot be concluded.
The seascape-scale CCA-model was particularly strong, suggesting that seascape habitat configuration
is of considerable importance as determinant for species composition of fish in Tc beds. The species
composition of the seascape scale CCA-plot is in line with results of fish abundance, e.g. the amount
of deep water- and land-arrows are placed opposite to amount of dense seagrass. Amount of sand and
amount of algae also points in the same direction as dense seagrass, likely because large sand flats and
algae belts are also typical characteristics of the large shallow-water seascapes. Sparse seagrass seems
to affect fish species composition similarly as land and deep water. However, it seems likely that it is
not the amount of sparse seagrass per se that affects species composition, but that it simply correlates
with amount of land and deep water without causality. Seascape heterogeneity affects fish species
composition in a similar way as amount of land and deep water. Seascapes with large proportions of
land and deep water show high values of seascape heterogeneity due to their steep slopes and large
scope of habitats in small areas. Amount of coral habitat in the seascapes also contribute to the
separation of fish species composition. Amount of coral habitat does not seem to correlate with
amount of land/deep water, and it has indeed been shown before that fish species composition in
seagrass beds are affected by presence of nearby coral reefs (Dorenbosch et al., 2006b).
Biases in the method
Performing visual fish census in seagrass beds by snorkelling at the surface has some major biases that
are near-impossible to avoid. Tc is a high and dense seagrass where fish easily hide and it is often
difficult to “visually penetrate” the canopy. Therefore, both abundance and diversity of fish were
likely underestimated, and especially of small and cryptic species. In addition, some species or life
stages are more easily scared off than others, and therein lies a second bias in the observed species
and/or life stage composition.
Upon return from the field, we discovered that the observers in the fish census had not been as well
syncronised as first thought. There was a significant difference in the size-distribution of fish between
the two observers. There was, however, no skewed distribution in selection of transects that were
Page 18 of 33
censused by the two observers. Hence, the error was randomly distributed over the data set, and not
systematic in any way; and the error was deemed insignificant enough to be disregarded.
Patchiness of habitats is also an issue to consider since some Tc beds consisted of many small patches
instead of continuous beds, which may in some cases have increased the effect of edge. However, on a
seascape scale, looking at the larger “Tc dominated area”, the middle transects were always accurately
placed in the middle and the edge transects were always accurately placed close to and parallel to the
edges. In addition, patchiness was measured and in analyses revealed no effect. Therefore, it can be
concluded that patchiness of Tc beds did not cause any noticeable problem.
As described above (see Material and Methods), the benthic substrate map achieved an accuracy of 69
% and consequently, this means some errors in how seascape habitat configuration was depicted.
Exactly how this affects results and conclusions on fish assemblage composition and dynamics is
unknown at this point, but it is possible that the results from the regressions of number of pixels as
well as seascape scale CCA would slightly change in relation to map accuracy.
Impacts of the study
Connecting this study with previous studies, it can be suggested that seagrass fish assemblage
composition depend on small-scale structural complexity of the seagrass bed (Orth et al., 1984;
Gullström et al., 2008) and that seagrass beds are attributed as fish nursery habitat (e.g. Nagelkerken et
al., 2000; Dorenbosch et al., 2004; Nagelkerken, 2009a). Seagrass fish assemblage composition has
also previously been explained by large-scale factors such as distances to neighboring habitats
(Gullström et al., 2008; Unsworth et al., 2008), connectivity (Dorenbosch et al., 2007; Unsworth et al.,
2008) and seascape configuration (Dorenbosch et al., 2006b). Edge effects on fish assemblages in
seagrass beds have also been studied quite well, e.g. edge effects on coral reefs (Dorenbosch et al.,
2005), effects of seaward edges compared to shoreward edges (Smith et al., 2008), edge effects in
seagrass patches of different sizes (Jelbart et al., 2006) and degree of fragmentation (MacReadie et al.,
2010. However, to combine multiple spatial scales, in addition to fish life stage separation, can be
considered quite unique, especially within the Western Indian Ocean region (Gullström et al., 2002;
2012).
This study has highly detailed, continuous quantification of edge habitat and seascape configuration.
In contrast, previous studies mostly use single categories of edge habitat, such as coral-seagrass edges
(Dorenbosch et al., 2005) or sand-seagrass edges (Smith et al., 2008; MacReadie et al., 2010) and also
categories of seascape configurations (Dorenbosch et al., 2006b) rather than continuous
measurements. Using site-specific categories instead of continuous measurements is a simplification
that sometimes might generate clearer results more easily. But to be able to produce generally
applicable frameworks for practical use in management, site-specific categories are not sustainable in
the long run, but generally applicable, holistic research- and monitoring-designs need to be created.
Remote sensing paired with Geographical Information Systems (GIS) have been attributed as suitable
large-scale and cost-efficient tools in seascape ecology and management-related research (Roff and
Zacharias, 2011; Grober-Dunsmore et al., 2009). Seagrass beds have been analyzed with remote
sensing methods (Ferguson et al., 1993; Mumby et al., 1997; Gullström et al., 2006; Knudby and
Nordlund, 2011; Knudby et al., in review) and the benthic substrate map in this study is a good
example of how remote sensing can be used in seascape ecology research. This mapping method used
Page 19 of 33
in this study carries great potential as a tool for understanding fish assemblage composition, and for
further development and use as a tool in future study designs.
Conclusions
This study successfully shows how fish assemblage composition is affected by different predictor
variables on different spatial scales. On a small scale, juvenile fish abundance and overall species
diversity seem to be positively affected by structural complexity in the middle of Tc beds, due to an
increased survival of juveniles and overall quality of habitat for seagrass-associated fish in structurally
complex beds. On a larger scale, adult fish abundance and fish species diversity is higher in Tc edges
bordering structurally complex or heterogeneous habitats, compared to structurally simple or
homogenous habitats. Further, adult fish abundance was positively affected by higher overall seascape
connectivity. Adult fish were affected by edge and seascape habitat composition rather than juveniles
due to their more complex migratory habits and multi-habitat uses. On all scales, predictors’ effects on
fish species compositions mirror the abovementioned ecological patterns; bringing further coherence
and credibility to the results of this study.
An interesting difference can be found when comparing the results of the edge-scale- and seascape-
scale analyses. In the edge, coral habitat and heterogeneity clearly affected fish abundance and
diversity positively. At seascape-scale, however, neither coral habitat nor heterogeneity had any
effects whatsoever. The reason for the difference is likely that ecological processes operate on
different scales, e.g. the amount of coral and heterogeneity seem to be affecting fish composition on a
relatively small scale or have a restricted radius of influence (Dorenbosch et al., 2005).
With this study, a novel design for understanding fish assemblage composition is presented. With a
unique combination of spatial scales and details in measurements, it yields a holistic view and has a
high potential for general applicability and use in further research as well as local management.
Acknowledgements
I would like to give my warmest thanks to:
Alan Koliji and Karolina Wikström for assistance in the field, advice, discussions and support. Martin
Gullström, Regina Lindborg and Lina Mtwana Nordlund for design development, advice, discussions
and feedback. Narriman Jiddawi and the Institute of Marine Science (IMS) in Zanzibar, for support
and guidance in the field. Anders Knudby for the idea, design and creation of the benthic substrate
map. Olle Hjerne for help with statistical analyses. Sustainable Management of Land and Environment
(SMOLE II) in Zanzibar, for providing aerial photographs. National Land Survey of Sweden for
coordinate conversions. The Swedish International Development Cooperation Agency (Sida), and
Martin Gullström, for funding this project. Isabell Stenson for support and feedback.
Page 20 of 33
References
Berkström, C., Gullström, M., Lindborg, R., Mwandya, A.W., Saleh, S.A.S., Kautsky, N., Nyström, M., 2012. Exploring
‘knowns’ and ‘unknowns’ in tropical seascape connectivity with insights from East African coral reefs. Estuarine, Coastal
and Shelf Science, 107:1-21.
Berkström, C., Lindborg, R., Thyresson, M., Gullström, M., 2013. Assessing connectivity in a tropical embayment: Fish
migrations and seascape ecology. Biological Conservation, 166:43-53.
Bos, R.A., Bouma, T.J., de Kort, G.L.J., van Katwijk, M.M., 2007. Ecosystem engineering by annual intertidal seagrass beds:
Sediment accretion and modification. Estuarine, Coastal and Shelf Science, 74:344-348
Boström, C., Jackson, E.L. and Simenstad, C.A., 2006. Seagrass landscapes and their effects on associated fauna: A review.
Estuarine, Coastal and Shelf Science, 68:383-403.
Breiman, L., 2001. Random forests. Machine Learning, 45:5-32.
Cincotta, R.P, Wisnewski, J., Engelman, R., 2000. Human population in the biodiversity hotspots. Nature, 404:990-992.
Conolly, R.M., Hindell, J.S., 2006. Review of nekton patterns and ecological processes in seagrass landscapes. Estuarine,
Coastal and Shelf Science, 68:433-444
de la Torre Castro, M., Rönnbäck, P., 2004. Links between humans and seagrasses – an example from tropical East Africa.
Ocean and Coastal Management, 47:361-387.
de la Torre-Castro, M., Lindström, L., 2010. Fishing institutions: addressing regulative, normative, and cultural-cognitive
elements to enhance fisheries management. Marine Policy, 34:77-84.
Dorenbosch, M., (ed.), 2006. Connectivity between fish assemblages of seagrass beds, mangroves and coral reefs - Evidence
from the Caribbean and western Indian Ocean -. 1st edn. Nijmegen: Radboud University.
Dorenbosch, M., Grol, M.G.G., 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., Grol, M.G.G., Nagelkerken, I., van der Velde, G., 2005. Distribution of coral reef fishes along a coral reef –
seagrass gradient: edge effects and habitat segregation. Marine Ecology Progress Series, 299:277–288.
Dorenbosch, M., Nagelkerken, I., Verberk. W.C.E.P., van der Velde, G., 2006a. Habitat configuration influences connectivity
of fishes between Caribbean seagrass beds, mangroves and coral reefs. In: Dorenbosch, M., (ed.), 2006. Connectivity between
fish assemblages of seagrass beds, mangroves and coral reefs - Evidence from the Caribbean and western Indian Ocean -.
Nijmegen: Radboud University. Ch. 7.
Dorenbosch, M., van Riel, M.C., Nagelkerken, I., van der Velde, G., 2004. The relationship of reef fish densities to the
proximity of mangrove and seagrass nurseries. Estuarine Coastal and Shelf Science, 60:37-48.
Dorenbosch, M., Verberk, W.C.E.P., Nagelkerken, I., van der Velde, G., 2007. Influence of habitat configuration on
connectivity between fish assemblages of Caribbean seagrass beds, mangroves and coral reefs. Marine Ecology Progress
Series, 334:103-116.
Duarte, C.M., Chiscano, C.L., 1999. Seagrass biomass and production: a reassessment. Aquatic Botany, 65:159-174.
Eklöf, J.S., de la Torre-Castro, M., Gullström, M., Uku, J., Muthiga, N., Lyimo, T., Bandeira, S.O., 2008. Sea urchin
overgrazing of seagrasses: A review of current knowledge on causes, consequences, and management. Estuarine, Coastal
and Shelf science, 79:569-580.
Ferguson, R., Wood, L., Graham, D., 1993. Monitoring spatial change in seagrass habitat with aerial photography.
Photogrammetric Engineering and Remote Sensing, 59:1033–1038.
Page 21 of 33
Froese, R., Pauly, D., (eds.), 2011. FishBase. World Wide Web electronic publication. www.fishbase.org, (10/2013 ).
Gratwicke, Speight, M.R., 2005. The relationship between fish species richness, abundance and habitat complexity in a range
of shallow tropical marine habitats. Journal of Fish Biology, 66(3):650-667.
Grober-Dunsmore, R., Frazer, T.K., Lindberg, W.J., Beets, J., 2006. Reef fish and habitat relationships in a Caribbean
seascape: the importance of reef context. Coral Reefs, 26:201-216
Grober-Dunsmore, R., Pittman, S.J., Caldow, C., Kendall, M.S., Frazer, T.K., 2009. A Landscape Ecology Approach for the
Study of Ecological Connectivity Across Tropical Marine Seascapes. In: Nagelkerken, I., (ed.), 2009. Ecological connectivity
among tropical coastal ecosystems. 1st edn. Dordrecht: Springer. Ch. 14.
Gullström, M., Berkström, C., Öhman, M.C., Bodin, M., Dahlberg, M., 2011. Scale-dependent patterns of variability of a
grazing parrotfish (Leptoscarus vaigiensis) in a tropical seagrass-dominated seascape. Marine Biology, 158:1483–1495.
Gullström, M., Bodin, M., Nilsson, P.G., Öhman, M.C., 2008. Seagrass structural complexity and landscape configuration as
determinants of tropical fish assemblage composition. Marine Ecology Progress Series, 363:241-255.
Gullström, M., de la Torre-Castro, M., Bandeira, S.O., Björk, M., Dahlberg, M., Kautsky, N., Rönnbäck, P., Öhman, M.C.,
2002. Seagrass Ecosystems in the Western Indian Ocean. Ambio, 31(7/8):588-596.
Gullström, M., Lundén, B., Bodin, M., Kangwe, J., Öhman, M., Mtolera, M., Björk, M., 2006. Assessment of changes in the
seagrass-dominated submerged vegetation of tropical Chwaka Bay (Zanzibar) using satellite remote sensing. Estuarine,
Coastal and Shelf Science, 67:399-408.
Gullström, M., Lyimo, T.J., Eklöf, J.S., Björk, M., Semesi, I.S., de la Torre-Castro, M., 2012. Seagrass Meadows in Chwaka
Bay: Socio-ecological and Management Aspects. In: de la Torre-Castro, M., Lyimo, T.J. (eds.), 2012. People, Nature and
research in Chwaka Bay, Zanzibar, Tanzania. WIOMSA. 346p. (Synthesis book).
Harborne, A.R., Mumby, P.J., Micheli, F., Perry, C.T., Dahlgren, C.P., Holmes, K.E., Brumbaugh, D.R., 2006. The
Functional Value of Caribbean Coral Reef, Seagrass and Mangrove Habitats to Ecosystem Processes. Advancec in Marine
Biology, 50:59-192.
Heck, K.L.Jr., Hays, G., Orth, R.J., 2003. Critical evaluation of the nursery role hypothesis for seagrass meadows. Marine
Ecology Progress Series, 253: 123-136.
Hemminga, M.A., Duarte, C.M., 2000. Seagrass ecology. 1st edn. Cambridge: Cambridge University Press.
Jelbart, J.E., Ross, P.M., Connolly, R.M., 2006. Edge effects and patch size in seagrass landscapes: an experimental test
using fish. Marine Ecology Progress Series, 319: 93-102.
Jiddawi, N.S., Stanley, R.D., 1999. A study of the artisanal fishery landings in the villages of Matemwe and
Mkokotoni, Zanzibar. In: Jiddawi, N.S., Stanley, R.D (eds). Fisheries Stock Assessment in the Traditional Fishery
Sector: the Information Needs. pp 50-70. National Workshop on the Artisanal Fisheries Sector, Zanzibar. September,
1997, Institute of Marine Sciences, University of Dar-es-Salaam, Zanzibar, Tanzania.
Jiddawi, N.S., Öhman, M.C., 2002. Marine fisheries in Tanzania. Ambio, 31(7): 518-527.
Kennedy, H., Beggins, J., Duarte, C.M., Fourqurean, J.W., Holmer, M., Marbà, N., Middelburg, J.J., 2010. Seagrass
sediments as a global carbon sink: Isotopic constraints. Global Biogeochemical Cycles, 24, GB4026,
doi:10.1029/2010GB003848.
Knudby, A., Newman, C., Shaghude, Y., Muhando, C., 2010. Simple and effective monitoring of historic changes in
nearshore environments using the free archive of Landsat imagery. International Journal of Applied Earth Observation and
Geoinformation, 12(1):116-122.
Knudby, A., Nordlund, L., 2011. Remote sensing of seagrasses in a patchy multi-species environment. International Journal
of Remote Sensing, 32(8):2227-2244.
Page 22 of 33
Knudby, A., Nordlund, L.M., Palmqvist, G., Wikström, K., Koliji, A., Lindborg, R., Gullström, M. Using multiple Landsat
scenes in an ensemble classifier reduces classification error in a stable nearshore environment. Submitted to International
Journal of Applied Earth Observation and Geoinformation.
Liaw, A., Wiener, M., 2002. Classification and Regression by randomForest. R news, 2(3), 18-22.
McAllister, D.E., 1991. What is the status of the world's coral reef fishes? Sea wind, 5:14-18.
Macreadie, P.I., Hindell, J.S., Keough, M.J., Jenkins, G.P., Connolly, R.M., 2010. Resource distribution influences positive
edge effects in a seagrass fish. Ecology, 91(7):2013-2021.
McClanahan, T., 1988. Seasonality in East Africa’s coastal waters. Marine Ecology Progress Series, 44: 191-199.
Moberg, F., Folke, C., 1999. Ecological goods and services of coral reef ecosystems. Ecological Economics, 29:215-233.
Mumby, P.J., Green, E.P., Edwards, A.J., Clark, C.D., 1997. Coral reef habitat mapping: how much detail can remote
sensing provide? Marine Biology, 130:193-202.
Nagelkerken, I., 2009a. Evaluation of Nursery function of Mangroves and Seagrass beds for Tropical Decapods and Reef
fishes: Patterns and Underlying Mechanisms. In: Nagelkerken, I., ed., 2009. Ecological connectivity among tropical coastal
ecosystems. 1st edn. Dordrecht: Springer. Ch. 10.
Nagelkerken, I., ed., 2009b. Ecological connectivity among tropical coastal ecosystems. 1st edn. Dordrecht: Springer.
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 Curacao (Netherlands Antilles)? Marine Ecological Progress Series, 245:191-204.
Nagelkerken, I., van der Velde, G., Gorissen, M.W., Meijer, G.J., 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. Estuarine, Coastal and Shelf Science, 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. Aquatic Living Resources, 23:399-416.
Ogden, J.C., 1988. The influence of adjacent systems on the structure and function of coral reefs. In: Proceedings of the Sixth
International Coral Reef Symposium, Townsville, Australia, pp. 123-129.
Ornellas, A.B., Coutinho, R., 1998. Spatial and temporal patterns of distribution and abundance of a tropical fish assemblage
in a seasonal Sargassum bed, Cabo Frio Island, Brazil. Journal of Fish Biology, 53:198–208.
Orth, R.J., Heck, K.L.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(4A):339-350.
Parrish, J.D., 1989. Fish communities of interacting shallow-water habitats in tropical oceanic regions. Marine Ecology
Progress Series, 58:143-160.
Ries, L., Sisk, T.D., 2004. A predictive model of edge effects. Ecology, 85(11): 2917-2926.
Robbins, B.D., Bell, S.S., 1994. Seagrass landscapes: a terrestrial approach to the marine subtidal environment. TREE,
9(8):301-304.
Roff, J., Zacharias, M. (eds.), 2011. Marine Conservation Ecology. 1st edn. Washington D.C.: Earthscan.
Salita, J.T., Ekau, W., Saint-Paul, U., 2003. Field evidence on the influence of seagrass landscapes on fish abundance in
Bolinao, northern Philippines. Marine Ecology Progress Series, 247: 183-195.
Shaghude, Y. W., Wannäs, K. O., Petersson, C. H., 2001. General characteristics of the intertidal and lagoonal sediments
around Unguja Island, Zanzibar, Tanzania. Marine Science Development in Tanzania and Eastern Africa, 95-105.
Sheaves, M., 2009. Consequences of ecological connectivity: the coastal ecosystem mosaic. Marine Ecology Progress Series,
391:107-115.
Page 23 of 33
Smith, T.M., Hindell, J.S., Jenkins, G.P., Connolly, R.M., 2008. Edge effects on fish associated with seagrass and sand
patches. Marine Ecology Progress Series, 359: 203-213.
Unsworth, R.K.F, Cullen, L.C., 2010. Recognising the necessity for Indo-Pacific seagrass conservation. Conservation
Letters, 3: 63–73.
Unsworth, R.K.F., Salinas de León, P., Gerrard, S.L., Jompa, J., Smith, D.J., Bell, J.J., 2008. High connectivity of Indo-
Pacific seagrass fish assemblages with mangrove and coral reef habitats. Marine Ecology Progress Series, 353: 213–224.
Watson, M., Munro, J.L., Gell, F.R., 2002. Settlement, movement and early juvenile mortality of the yellowtail snapper
Ocyurus chrysurus. Marine ecology. Progress Series, 237, 247-256.
Page 24 of 33
Appendix 1: Fish species list
In the following appendix, a complete list of all fish species found in transects are presented. Fish
species are organized alphabetically according to family. When species names are not given, the
following rules were applied; for example: “Acanthuridae sp.” was identified to family-level but not
any further, i.e. it can thus be any member of the family. “Acanthuride sp. 1” and “Acanthuride sp. 2”
are distinguished from all other species, but could not be identified well enough so that a species name
could be decided. An asterisk (*) means that the species was observed as schools and removed from
abundance and multivariate analyses, but included in diversity analyses.
Acanthuridae (surgeonfishes, tangs, unicornfishes) Apogonidae (cardinalfishes)
Acanthuridae sp. Apogon cyanosoma
Acanthuridae sp. 1 Apogon nigripes
Acanthuridae sp. 2 Apogonidae sp. 1
Acanthurus blochii Apogonidae sp. 2
Acanthurus leucosternon Archamia fucata
Chlorurus sordidus Archamia mozambiquensis
Ctenochaetus striatus Cheilodipterus artus
Ctenochaetus strigosus Cheilodipterus quinquelineatus
Naso brevirostris Balistidae (triggerfishes)
Zebrasoma scopas Balistidae sp. 1
Aulostomidae (trumpetfishes) Ballistapus undulatus
Aulostomus chinensis Caesionidae (fusiliers)
Blenniidae (combtooth blennies) Caesio lunaris
Blenniidae sp. 1 Pterocaesio tile
Exallias brevis Carangidae (jacks)
Meiacanthus mossambicus Gnathanodon speciosus
Plagiotremus tapeinosoma Chaetodontidae (butterflyfishes)
Centriscidae (shrimpfishes, snipefishes) Chaetodon auriga
Aeoliscus punctulatus Chaetodon falcula
Cirrhitidae (hawkfishes) Chaetodon guttatissimus
Paracirrhites forsteri Chaetodon kleinii
Fistularidae (cornetfishes) Chaetodon lunula
Fistularia commersonii Chaetodon melannotus
Gerreidae (mojarras) Chaetodon meyeri
Gerres oyena Chaetodon trifascialis
Gobiidae (gobies) Chaetodon trifasciatus
Gobiidae sp. 1 Chaetodon xanthocephalus
Gobiidae sp. 2 Heniochus acuminatus
Hemiramphidae (halfbeaks) Haemulidae (grunts)
Hemiramphus far Plectorhinchus flavomaculatus
Hyporhamphus dussumieri * Plectorhinchus gaterinus
Holocentridae (soldierfishes, squirellfishes) Plectorhinchus gibbosus
Myrispristis violacea Plectorhinchus playfairi
Page 25 of 33
Sargocentron diadema Plectorhinchus schotaf
Labridae (wrasses) Lethrinidae (emperors)
Anampses lineatus Lethrinidae sp. 1
Bodianus axillaris Lethrinus harak
Cheilinus bimaculatus Lethrinus mahsena
Cheilinus trilobatus Lethrinus obsoletus
Cheilio inermis Lethrinus variegatus
Coris africana Lutjanidae (snappers)
Coris caudimacula Lutjanus fulviflamma
Gomphosus caeruleus Monacanthidae (filefishes)
Halichoeres hortulanus Cantherhines pardalis
Halichoeres nebulosus Monacanthidae sp. 1
Halichoeres nigrescens Monodactylidae (moonyfishes)
Halichoeres scapularis Monodactylus argenteus
Hemigymnus fasciatus Mullidae (goatfishes)
Hemigymnus melapterus Mulloidichtys flavolineatus
Labridae sp. Parupeneus barberinus
Labridae sp. 1 Parupeneus macronema
Labridae sp. 2 Upeneus tragula
Labridae sp. 3 Muraenidae (moray eels)
Labroides bicolor Gymnothorax sp.
Labroides dimidiatus Nemipteridae (threadfin breams)
Novaculichthys macrolepidotus Scolopsis ghanam
Stethojulis albovittata Ophichthidae (worm eels)
Stethojulis strigiventer Myrichtys colubrinus
Thalassoma amblycephalum Ostraciidae (boxfishes)
Thalassoma hardwicke Ostracion cubicus
Thalassoma hebraicum Ostracion meleagris
Thalassoma lunare Pomacanthidae (angelfishes)
Pempheridae (sweepers) Centropyge multispinis
Pempheris adusta Pomacanthidae sp. 1
Plotosidae (eeltail catfishes)
Plotosus lineatus *
Pomacanthus semicirculatus
Ptereleotridae (dartfishes)
Pomacentridae (damselfishes) Ptereleotris evides
Abudefduf sexfasciatus Scaridae (parrotfishes)
Abudefduf sparoides Calotomus carolinus
Abudefduf vaigiensis Chlorurus sordidus
Amphiprion akallopisos Ctenochaetus striatus
Amphiprion chrysopterus Leptoscarus vaigiensis
Chromis atripectoralis Scaridae sp. 1
Chromis dimidiata Scarus ghobban
Chromis nigroanalis Scarus niger
Chromis nigrura Scarus psittacus
Chromis sp. 1 Serranidae (seabasses, groupers)
Chromis ternatensis Cephalopholis argus
Chromis weberi Cephalopholis boenak
Page 26 of 33
Chromis viridis Epinephelus fasciatus
Chrysiptera annulata Serranidae sp. 1
Chrysiptera biocellata Serranidae sp. 2
Chrysiptera leucopoma Siganidae (rabbitfishes)
Dascyllus aruanus Siganus fuscescens
Dascyllus carneus Siganus sutor
Dascyllus trimaculatus Soleidae (soles)
Plectroglyphidodon lacrymatus Pardachirus marmoratus
Pomacentridae sp. Sphyraenidae (barracudas)
Sphyraena sp. * Pomacentridae sp. 1
Pomacentridae sp. 2 Syngnathidae (pipefishes, seahorses)
Pomacentrus caeruleus Chorythoichtys flavofasciatus
Pomacentrus sp. Hippichthys spicifer
Pomacentrus sulfureus Synodontidae (lizardfishes)
Pomacentrus trilineatus Saurida gracilis
Stegastes nigricans Synodus variegatus
Tetraodontidae (puffers) Unknown family
Arothron nigropunctatus Unknown sp.
Canthigaster bennetti Unknown sp. 1
Canthigaster solandri Unknown sp. 2
Canthigaster valentini Unknown sp. 3
Zanclidae (moorish idol) Unknown sp. 4
Zanclus cornutus Unknown sp. 5 *
Unknown sp. 6
Page 27 of 33
Appendix 2: Benthic substrate map
First in this appendix, the benthic substrate map is presented in its entirety including a legend. On the
following pages, more detailed maps at the main regions (northern, western, eastern; Cwaka Bay, and
southern; Menai Bay) are presented together with each individual seascape. The seascapes are
presented as black circles (with radiuses of 500 m). Sources of the names of the seascapes are various
local guides that helped in the fieldwork. Names are derived from a nearby village, island or fishing
site.
Page 28 of 33
Page 29 of 33
Northern region
Page 30 of 33
Western region
Page 31 of 33
Eastern region (Chwaka Bay)
Page 32 of 33
Southern region (Menai Bay)
Page 33 of 33
Stockholm University
SE-106 91 Stockholm
Phone: 08 – 16 20 00
www.su.se