Content uploaded by Simon James Pittman
Author content
All content in this area was uploaded by Simon James Pittman
Content may be subject to copyright.
MARINE ECOLOGY PROGRESS SERIES
Mar Ecol Prog Ser
Vol. 427: 259–274, 2011
doi: 10.3354/meps08945 Published April 12
INTRODUCTION
Recent studies have shown that the distribution and
abundance of reef fish can be influenced by seascape
factors such as the areas of adjacent seagrass, soft bot-
tom, and hard bottom in the vicinity (Kendall et al.
2004a, Kendall 2005, Dorenbosch et al. 2006, Grober-
Dunsmore et al. 2007, Pittman et al. 2007, Tuya et al.
2010). Many of these studies relied on particular ben-
thic maps as a source of independent variables with
which to establish relationships between fish and their
surrounding habitat. Benthic maps are, however,
abstract representations of actual seafloor features and
have particular spatial and thematic characteristics
that are profoundly affected by the processes and
source data used to produce them (Turner et al. 1989,
Benson & MacKenzie 1995, Saura 2002, Andréfouët et
al. 2003, Kendall & Miller 2008, Prada et al. 2008). Spa-
tial characteristics include the number of patches, their
size, shape, and edge length. Thematic characteristics
include the number and types of categories used to
describe seafloor features. For coral reef ecosystems,
Andréfouët et al. (2003) found that map-based mea-
surements of coral atolls differed by as much as 28 %
depending on the spatial resolution of satellite data.
Kendall & Miller (2008) found that increasing thematic
resolution greatly increased the number, diversity, and
total edge length of map polygons, whereas changing
the spatial resolution resulted in disproportionate
changes in the area, perimeter, and other values
© Inter-Research 2011 · www.int-res.com*Email: matt.kendall@noaa.gov
Patterns of scale-dependency and the influence of
map resolution on the seascape ecology of reef fish
Matthew S. Kendall1,*, Thomas J. Miller2, Simon J. Pittman1, 3
1National Oceanic & Atmospheric Administration Biogeography Branch, 1305 East West Highway,
Silver Spring, Maryland 20910, USA
2Chesapeake Biological Laboratory, University of Maryland Center for Environmental Science,
PO Box 38, Solomons, Maryland 20688, USA
3National Oceanic & Atmospheric Administration Biogeography Branch, Marine Science Center,
University of the Virgin Islands, St. Thomas, US Virgin Islands 00802, USA
ABSTRACT: Detection and perception of ecological relationships between biota and their surround-
ing habitats is sensitive to analysis scale and resolution of habitat data. We measured strength of uni-
variate linear correlations between reef fish and seascape variables at multiple spatial scales (25 to
800 m). Correlation strength was used to identify the scale that best associates fish to their surround-
ing habitat. To evaluate the influence of map resolution, seascape variables were calculated based on
4 separate benthic maps produced using 2 levels of spatial and thematic resolution, respectively. Indi-
vidual seascape variables explained only 25% of the variability in fish distributions. Length of reef
edge was correlated with more aspects of the fish assemblage than other features. Area of seagrass
and bare sand correlated with distribution of many fish, not just obligate users. No fish variables cor-
related with habitat diversity. Individual fish species achieved a wider range of correlations than
mobility guilds or the entire fish assemblage. Scales of peak correlation were the same for juveniles
and adults in a majority of comparisons. Highly mobile species exhibited broader scales of peak cor-
relation than either resident or moderately mobile fish. Use of different input maps changed percep-
tion of the strength and even the scale of peak correlations for many comparisons involving hard bot-
tom edge length and area of sand, whereas results were consistent regardless of map type for
comparisons involving area of seagrass and habitat diversity.
KEY WORDS: Landscape ecology · Scale · Coral reef · Home range · Habitat
Resale or republication not permitted without written consent of the publisher
Contribution to the Theme Section ‘Seascape ecology’
O
PEN
PEN
A
CCESS
CCESS
Mar Ecol Prog Ser 427: 259–274, 2011260
among feature types. Given the influence of spatial
and thematic resolution on the quantification of sea -
scapes, we hypothesized that map differences could
influence the sensitivity of seascape ecological studies
as well.
Map resolution may affect the detection and mea-
surement of seascape influences on fish distribution in
several ways. The amount of mapped habitat deemed
essential to a particular species can be altered. Small
or rare patches of habitat can be subsumed into larger
features as map resolution is coarsened. Many species
use edges or ecotones between habitat patches (Shul-
man 1985, Sweatman & Robertson 1994, Dorenbosch
et al. 2005, Pittman et al. 2007, Valentine et al. 2007,
Vanderklift et al. 2007), and such boundaries can be
greatly simplified or even removed depending on map
characteristics (Kendall & Miller 2008).
Seascape composition can affect fish ecology at sev-
eral levels of biological organization. At the broadest
level, species diversity, richness, and total abundance
of fish have been partly explained by seascape vari-
ables (Kendall 2005, Grober-Dunsmore et al. 2007,
Pittman et al. 2007). At lower levels of organization de-
fined according to trophic roles or mobility and there-
fore with similar habitat or space requirements, fish
guilds may have greater correlation with seascape ele-
ments when considered separately from the rest of the
fish assemblage (Grober-Dunsmore et al. 2007). Indi-
vidual species would be expected to have even higher
cor relation with seascape features without the added
variability from multiple species of a guild that utilize
slightly different niches or habitats. Highest fish–
seascape correlations are expected for individual life
stages of species considered separately. All such fish
utilize the same discrete spatial scale and habitat types
(Kendall et al. 2003, Grober-Dunsmore et al. 2007), and
correlations would not be reduced by the added vari-
ability associated with the different scales of seascape
utilization and habitat preferences among life stages.
The strength of fish–seascape correlations are likely
scale-dependent and based on fish size, mobility, tax-
onomy, life stage, and habitat requirements (Kramer
& Chapman 1999, Pittman et al. 2004, Kendall 2005,
Grober-Dunsmore et al. 2007). Body size in some reef
fish has been positively correlated to home range size
(Kramer & Chapman 1999, Overholtzer & Motta 1999).
Similarly, juveniles of a given species, by virtue of their
relatively smaller size, are expected to interact with
adjacent seascape features at shorter distances than
adults (Kendall et al. 2003, Grober-Dunsmore et al.
2007). The abundance of those species that utilize a
single rock, coral head, or burrow for most of their life
history, termed resident species, would be expected to
exhibit correlations only with habitat measures for
their immediate vicinity. A good example are fish in
the genus Stegastes, which defend territories of 1 to
5m
2for food and breeding purposes (Itzkowitz 1977,
Luckhurst & Luckhurst 1978). Those species that range
more widely across larger or multiple habitat patches,
termed mobile species, would not be expected to be
correlated with just the habitats in the immediate
vicinity of a focal patch and instead would have cor -
relations with seascape features at distances of 10s
to 100s of meters away. Examples include haemulids
(Tulevech & Recksiek 1994, Burke 1995), acanthurids
(Morgan & Kramer 2004), and scarids (Chapman &
Kramer 2000, Mumby & Wabnitz 2002). Finally, those
species that range widely across the seascape, termed
transient species, would likely have correlations with
seascape features at even greater distances. Such
species include many fish in the families Carangidae
and Lutjanidae (Chapman & Kramer 2000).
The scale of fish– seascape interactions can be iden-
tified by evaluating a local fish assemblage in relation
to the surrounding mosaic of habitat patches (Irlandi
& Crawford 1997, Pittman et al. 2004, Kendall 2005,
Grober-Dunsmore et al. 2007, Vanderklift et al. 2007).
The spatial extent of elements to include in such analy-
sis is critical, and analyses should seek to systemati-
cally vary the spatial scale and distances over which
fish and seascape associations are measured (Addicott
et al. 1987, Wiens 1989, Riitters et al. 1997, Sale 1998,
Kendall 2005). If the spatial extent used is too small,
only weak fish– seascape correlations will be found,
because seascape elements used by the taxa con -
sidered are not included in the analysis. Conversely, if
the analysis is conducted at a spatial extent that is too
broad, weak correlations will again be found, but this
time because too many seascape elements not used by
the taxa under consideration are being included. Cor-
relations will be maximal at an intermediate extent
that matches the scale of habitat use. Once quantified
across a range of scales, correlation strength can be
examined to find the scale that best correlates fish
with their surrounding habitat. Use of this ‘best’ scale
approach to empirically define the ecological scale has
recently emerged in a handful of terrestrial (Pearson
1993, Karl et al. 2000, Ricketts et al. 2001, Steffan-
Dewenter 2003, Holland et al. 2004, 2005) and marine
studies (Kendall 2005, Grober-Dunsmore et al. 2007).
Both the scale and strength of fish–seascape corre-
lations can be influenced by map resolution. If only
the perception of correlation strength is altered, it is
not a serious problem if the objective is merely to
identify the scale of peak correlation. However, it is
of concern if accurate measurement of correlation
intensity is the goal. Of greater concern is when map
type alters both correlation scale and magnitude,
resulting in a complete misperception of a fish–
seascape association.
Kendall et al.: Perception of fish–seascape correlations
The real seascape that fish inhabit and are influ-
enced by is depicted in greatest detail by maps with
very high spatial and thematic resolution. Thus, high-
resolution maps are expected to yield the highest fish–
seascape correlations when an association exists and
also the lowest correlations when no association is
present. Highly detailed maps are, however, costly and
time-consuming to produce. Maps generalized into
coarser thematic and spatial properties are cheaper
and faster to produce, but these changes often have
unknown consequences on the perception of fish–
seascape correlations.
In the present study we investigated several interre-
lated aspects of ecological scale by simultaneously
varying (1) the size of habitat measurements used as
independent variables, (2) both the spatial and the-
matic resolution of map data, and (3) the life stage,
mobility guild, and fish assembly groups used as
dependent variables. Our complementary hypotheses
(1 to 5, below) seek to disentangle the effects of these
issues in detecting and accurately perceiving fish–
seascape relationships:
(1) Reef fish are significantly correlated with
seascape variables (area of sand or seagrass, length of
hard bottom or reef edge, and habitat diversity).
(2) Correlation strength will vary with the spatial
scale of the analysis.
(3) Thematic and spatial resolution of maps will
affect correlation strength and scale. Highest resolu-
tion maps will yield the highest correlations.
(4) Range in correlation strength will be highest for
individual life stages of species followed by guilds, and
lowest for whole community metrics.
(5) Resident fish species will have highest correlation
with seascape variables at shorter distances than
mobile, followed by transient species, and juveniles of
a given species will have highest correlation with
seascape variables at shorter distances than adults.
MATERIALS AND METHODS
This study was based on the fish communities and
seascape around Buck Island Reef National Monument
(BIRNM), US Virgin Islands, which is comprised of
seascape features typical of Caribbean coral reef eco -
systems (Fig. 1). Dependent variables were based on
fish surveys, and independent variables were based on
4 types of benthic maps.
Fish survey data. Underwater visual surveys were
used to census fish on 588 sites on reefs and hard
bottom within and around BIRNM between January
2002 and May 2006. Due to random placement, each
site was surrounded by a unique combination of habi-
tat elements. At each georeferenced site, a diver swam
along a randomly selected compass heading for 15 min
and recorded all fish within 5 cm size classes that were
observed within a 25 ×4 m belt transect (100 m2) to the
lowest possible taxon.
A hierarchical approach was taken in selecting re -
sponse variables to test hypotheses relating fish distri-
bution at several organizational levels from whole
community to particular life stages of individual spe-
cies (Table 1, columns 1 & 2). Variables included total
fish abundance and overall species richness; species
richness and fish abundance within the mobility guilds
of resident (R), mobile (M), and transient (T) (Grober-
Dunsmore et al. 2007); and the abundance by life stage
of 6 common reef fish species. The 6 species were
chosen to include those with (1) representation from
diverse family and trophic groups, (2) known life his-
tory and habitat preferences, and (3) relatively common
occurrence in the study area across a range of seascape
settings. Abundances of these species were also sepa-
rated into juvenile and adult categories, respectively,
for life stage-specific analysis.
Mapping and quantifying seascape structure. Four
maps of the benthic habitat around BIRNM were pro-
duced using 2 levels of spatial and thematic resolution,
respectively (Kendall & Miller 2008). Maps were cre-
ated by visual interpretation of orthorectified aerial
photographs (Kendall et al. 2001). A relatively large
minimum mapping unit (MMU; size of the smallest
feature to be mapped) of 4000 m2and a much smaller
MMU of 100 m2were used. A hierarchical classifica-
tion scheme was used to attribute maps at both spatial
scales into 17 subcategories nested within 3 main cate-
gories in a scheme structurally similar to those used
to produce reef ecosystem maps elsewhere (FMRI
& NOAA 1998, Mumby & Harborne 1999, NOAA
NCCOS 2004). Main categories were unconsolidated
sediment, submerged vegetation, and hard bottom.
Subcategories of unconsolidated sediment were sand
and mud. Subcategories of submerged vegetation
were seagrass and algae in varying degrees of patchi-
ness. Subcategories of hard bottom were patch reefs,
linear reefs, colonized pavement, bedrock, pavement
with sand channels, spur and groove, and scattered
coral/rock. Maps with 17 classes served as high the-
matic resolution maps. Subsequently, we dissolved the
boundaries and aggregated the polygons of these high
thematic resolution maps to the 3 thematic class level
for use in the analyses as maps with low thematic reso-
lution. This process resulted in 4 maps of BIRNM using
the same approach but with different spatial and
thematic characteristics (Fig. 2).
Four variables were selected to quantify seascape
structure that were representative of broad classes of
‘landscape pattern metrics’ and suspected to play a role
in reef fish community structure: (1) area of seagrass or
261
submerged aquatic vegetation (SAV), (2) length of hard
bottom edge, (3) area of sand or unconsolidated sedi-
ment (map with high or low thematic resolution, re-
spectively), and (4) habitat diversity (Shannon-Wiener
Index). Area of seagrass around reefs is suspected to
enhance abundance of lutjanids (snappers), haemulids
(grunts), and other fish on reefs (Randall 1965, Ogden
1976, Kendall et al. 2003, Kendall 2005, Grober-Dun-
smore et al. 2007). Reef edges have been implicated as
a key ecotone shaping fish communities on reefs
(Dorenbosch et al. 2005, Vanderklift et al. 2007), includ-
ing piscivores (Shul man 1985, Sweatman & Robertson
1994, Valentine et al. 2007, Vanderklift et al. 2007), her-
bivores (Wern berg et al. 2006), and those species that
use hard bottom as structural refuge but forage over
soft bottom (Ogden 1976, Burke 1995). Sand and sea-
grass bottom provides settlement habitat for many reef
fish species and may therefore eventually impact adult
abundance on nearby reefs (Shulman & Ogden 1987,
Shulman 1985, Parrish 1989). Diversity of habitat types
may be positively correlated with the diversity of the
fish community (Ward et al. 1999) and has been sug-
gested as a surrogate for overall biodiversity in marine
reserve selection (National Research Council 2001).
Mar Ecol Prog Ser 427: 259–274, 2011262
0 1000 2000 3000 4000500
Meters
Land
Patchy macroalga
Patchy seagrass
Col. bedrock
Col. pavement
Cont. macroalgae
Cont. seagrass
Linear reef
Mangrove
Mud
Patch reef
Reef rubble
Sand
Scattered C&R
17° 48’
17° 47’
17° 46’
N
64
°
40’ W 64
°
39’ 64
°
38’ 64
°
37’ 64
°
36’ 64
°
35’ 64
°
34’
Buck Island study area
St.Croix
-
Atlantic Ocean
Caribbean Sea
04812162km
Green
Key Point
Udall
75
°
0’0’’W 70
°
0’0’’ 65
°
0’0’’ 60
°
0’0’’ 55
°
0’0’’
N
Bahamas
Cuba
St.Croix
25
°
0’0’’
20
°
0’0’’
30
°
0’0’’
Fig. 1. Buck Island study area, St. Croix, U.S. Vir-
gin Islands. Upper panel shows locations of fish
survey sites denoted by black dots. A set of 14 ran-
domly selected, non-overlapping sites with their
corresponding analysis radii (25, 50, 100, 200, 400,
800 m) are shown. Backdrop is the benthic map
with high spatial and thematic resolution. Lower
panel shows the study area location
Kendall et al.: Perception of fish–seascape correlations
Multiscale analysis. The dominant scales at which
components of the fish assemblage are correlated with
their surrounding habitat mosaic were identified using
a multi-scale approach. The seascape pattern metrics
were calculated around each of the 588 fish survey
sites within circular sample units
for all 4 map types, respectively
(Fig. 1). To determine which
analysis scale ‘best’ related to
a fish community variable, sea -
scape metrics were calculated at
a range of distances from very
small, including only seascape
elements directly at the fish
census site, through broad scales
that included the mosaic of
habitat elements beyond the
distance that fish were likely to
be influenced. The smallest dis-
tance was 25 m to incorporate
each 4 by 25 m fish survey.
Habitat metrics were also calcu-
lated at increasing distances of
50, 100, 200, 400, and 800 m
radii around each survey (Ken -
dall 2005) (Fig. 1).
Univariate linear regression
was used to determine the basic
relationship between a given fish
variable and the seascape vari-
ables at each analysis distance.
The strength of the relation -
ship as a function of spatial scale
was evaluated using the Pearson
correlation coefficient (r) that
resulted from the regression.
This was chosen over other re -
gression-based statistics, such as
R2, to characterize relationships
because it ranges from –1 to +1,
and therefore both positive and
negative relationships be tween
variables can be perceived.
To maintain independence
among samples, especially where
larger analysis diameters (e.g.
800 m) would result in very simi-
lar seascape values around adja-
cent survey sites but maximize
use of the data, a resampling ap-
proach was taken using the soft-
ware Focus v2.1 (Holland et al.
2004). Based on distances be-
tween survey points, a non-over-
lapping subset of random survey
sites was repeatedly drawn from the entire pool of 588
surveys (1 subset shown in Fig. 1). Preliminary analysis
revealed that 1000 iterations of the resampling proce-
dure and their corresponding correlation statistics pro-
duced a stable and unimodal set of r values with very low
263
Habitat Seagrass/ Sand/sedi- Hard bottom
diversity SAV area ment area edge length
Fish variables Suppl 1 Suppl 2 Suppl 3 Suppl 4 Panel
Whole community
Fish abundance – – – – a
Species richness – 800
a800
bb 800
ab b
Mobility guild
Resident fish abundance – – 25
bb –c
Resident species richness – 800
a– 800 d
Mobile fish abundance – – – – e
Mobile species richness – – – 50
bbb f
Transient fish abundance – – 800
bb –g
Transient species richness – – 800
bb –h
Species (abundance)
Acanthurus coeruleus
Juvenile – – – 800
aaa i
Adult – – – 800
aab j
Total – – – 800
aaa k
Cephalopholus fulvus
Juvenile – 800
a25
bb 800
ab l
Adult – 800 100
bb 800
ab m
Total – 800 25
bb 800
ab n
Haemulon flavolineatum
Juvenile – – – – o
Adult – 800 – 50
abb p
Total – 800 – 50
abb q
Ocyurus chrysurus
Juvenile – 800
bbb ––r
Adult – – – – s
Total – 100
abb ––t
Sparisoma viride
Juvenile – 800
a– 800
aab u
Adult – 400
a– 800
abb v
Total – – – 800
aa w
Stegastes planifrons
Juvenile – 800
a– 800
ab x
Adult – 800 – 800
ay
Total – 800
a– 800
ab z
No. of comparisons with 0 14 7 17
at least one map type
with |r| > 0.2
No. of times one of the na 8 0 22
other 3 map types had a
peak |r| at the same scale
but a significantly lower
value. aeach occurrence
No. of times one of the na 5 14 17
other 3 map types had
a peak |r| at a different
scale. beach occurrence
Table 1. Scale (m) of maximum fish– seascape correlation among all map types. SAV:
submerged aquatic vegetation; (–) variable pairs with no correlations above |r| = 0.2 for
any map type. Bottom rows summarize the changes in perception of fish–seascape
correlation due to map type. aand bare defined in the bottom 2 rows and denote 2 types
of altered perception. All 104 correlation plots are provided in the supplement at
www.int-res. com/ articles/suppl/m427p259_supp.pdf where numbers (Suppl. 1–Suppl.
4) and letters (Panels a– z) indicate seascape and fish variables, respectively (e.g. habitat
diversity and fish abundance is Supplement 1, Panel a).
Mar Ecol Prog Ser 427: 259–274, 2011
standard error for all variables and analysis scales. The
sampling process was conducted at all 6 analysis scales
for each combination of xand yvariables, respectively.
The mean and standard error of the correlation
coefficients from each scale were plotted and the scale of
greatest correlation (maximum |r|) was identified for
each pair of fish and seascape variables. The resampling
analysis was conducted for each of the 4 map types, and
the results for each fish and seascape variable were plot-
ted on the same chart to visualize the effect of map type
on fish– seascape correlations.
Hypothesis testing. Four outcomes were possible for
each fish–seascape combination in Table 1. The sim-
plest case was when no correlation was found between
a fish variable and seascape variable for any map type
or analysis scale. Another possibility was that a signifi-
cant correlation existed at one or more scales and all
map types yielded similar results. It was also possible
that one or more analysis scales yielded a significant
correlation, but the results were different depending on
the map type. In this case maximum |r| value among
map types could occur at the same scale but achieve
significantly different values, or maximum |r| values
among map types could occur at entirely different
scales. The results of each of the 104 fish and landscape
comparisons were tallied into one of these 4 categories
using the rules defined in the hypotheses below.
Hypothesis 1. Reef fish are significantly correlated
with seascape variables (area of sand or seagrass,
length of hard bottom or reef edge, and habitat diver-
sity): With Bonferroni correction for testing 6 scales at
once, nearly all mean r values were significantly non-
zero due to the very narrow standard error of the mean
(SEM) values. Even r values between +0.1 and –0.1,
which would account for <1% of the variability in the
fish– seascape relationship, were statistically signifi-
cant. To infer ecological relevance, a much higher and
more conservative |r| of 0.2 was therefore selected as a
cutoff for identifying more important ecological rela-
tionships and to reduce the probability of Type I errors.
Hypothesis 2. Correlation strength will vary with the
spatial scale of the analysis: A simple ANOVA evalu-
ated whether all 6 analysis scales yielded a maxi mum r
value (|r| > 0.2) at the same scale for each fish–seascape
variable pair and map type. Where a significant
ANOVA was found, a Tukey’s-type multiple means
comparison determined which scales differed. A more
conservative α= 0.001 was used to define significant
differences due to the narrow SEM values that resulted
from the resampling procedure.
264
Fig. 2. Four map types of the study region. Counterclockwise from upper left is the map with low spatial and thematic resolution,
low spatial but high thematic resolution, high spatial and thematic resolution, and high spatial but low thematic resolution. Grey
denotes land. White denotes unmapped area beyond the shelf edge. Green tones denote seagrass/submerged aquatic vegetation
categories. Tans denote sand/sedi ment categories. Reds denote coral reef/hard bottom categories. See Kendall & Miller (2008)
for quantitative differences among maps
Kendall et al.: Perception of fish–seascape correlations
Hypothesis 3. Thematic and spatial resolution of
maps will affect correlation strength and scale. Highest
resolution maps will yield the highest correlations: To
evaluate the possibility that peak r values occurred at
the same scale but had significantly different values
among map types, the scale with the highest r value
was identified and the mean r values among map types
were tested for significant differences using a conserv-
ative Tukey’s-type multiple means comparison proce-
dure (α= 0.001). To evaluate the possibility that peak r
values occurred at different scales for different map
types, peaks in |r| values by map type were identified
visually. The relative frequencies of these possible out-
comes were tabulated for each seascape variable at the
bottom of Table 1.
To determine if there was a map type that consis-
tently had the highest or lowest |r| values for each of
the 4 seascape variables, comparisons with significant
results were evaluated further. When a |r| > 0.2 was
present, maps yielding significantly higher or lower |r|
than the rest were tallied for each seascape variable.
Hypothesis 4. Range in correlation strength will be
highest for individual life stages of species followed by
guilds and lowest for whole community metrics: Maxi-
mum |r| values for each of the 104 variable combina-
tions were grouped and plotted by those that tested: (1)
abundance of individual life stages (juvenile or adult)
of particular species, (2) total abundance of particular
species, (3) abundance or species richness of the
mobility guilds, and (4) the whole assemblage vari-
ables of overall abundance or species richness. Plotted
|r| values were separated by map type and coded by
seascape variables. The range and distribution of
extreme values was compared among levels
of organization of the fish variables for each
map type.
Hypothesis 5. Resident fish species will
have highest correlation with sea scape vari-
ables at finer scales than mobile, followed by
transient species, and juveniles of a given
species will have highest correlation with
seascape variables at finer scales than adults:
The scale of peak correlation between juve-
niles and a given landscape variable was
identified for each map type and simply com-
pared to the scale of peak correlation for
adults. The distance of peak correlation of
juveniles relative to adults was described as
one of the following: juvenile < adult, adult <
juvenile, or when the scale of peak correlation
was the same for both of these life stages,
juvenile = adult. The hypo thesis that resident
fish have higher correlations with seascape
variables at finer scales than mobile, followed
by transient fish, was evaluated in similar
fashion. Habitat diversity was not evaluated because r
values were very low across all scales and no clear
peaks in correlation were ob served.
RESULTS
Hypothesis 1— Fish are correlated with seascape
variables
Linear correlations between fish and individual
seascape variables were low overall and explained a
low percentage of the variability in fish distributions
(see the supplement at www.int-res.com/ articles/ suppl/
m427p259_supp.pdf). Of the 2496 mean r values calcu-
lated, only 220 (~11%) exceeded the selected signifi-
cance level of |r| = 0.2. The strongest association was |r|
= 0.5 between the abundance of Cephalopholus fulvus,
a small grouper and the amount of hard bottom edge.
Habitat diversity was not significantly correlated with
any fish variable at any scale (Table 1) (e.g. Fig. 3).
Despite the lack of strong correlations between indi-
vidual seascape variables and fish distributions, some
relationships between fish and specific attributes of the
seascape were found. Length of hard bottom edge was
correlated with more of the fish variables (17) than any
other seascape feature (Table 1). Species richness, res-
ident species richness, mobile species richness, and
abundance of all species except for Ocyurus chrysurus
were related to length of hard bottom edge. Highest |r|
values found with hard bottom edge length were for
Acanthurus coeruleus adult and total abundance,
Cephalopholis fulvus adult and total abundance (neg-
265
Fig. 3. Correlations of habitat diversity and fish species richness by analy-
sis distance for all 4 map types. HH: map with high spatial and thematic
resolution. LH: map with low spatial and high thematic resolution. HL:
map with high spatial and low thematic resolution. LL: map type with low
spatial and thematic resolution. Observations between the horizontal
dashed lines (|r| < 0.2) denote non-significant results
Mar Ecol Prog Ser 427: 259–274, 2011
ative correlations), and Sparisoma viride juve-
nile and total abundance. Area of seagrass/
SAV was correlated with total species rich-
ness, species richness of resident fish, and
abundance of at least one life stage of all
species considered except for A. coeruleus
(Table 1). The highest |r| values found with
seagrass/SAV were for Stegastes planifrons
juvenile and total abundance, and C. fulvus
juvenile, adult, and total abundance (negative
correlations). Species richness and abundance
of transients, abundance of residents, and
over all species rich ness all had positive corre-
lations with area of sand/ sediment (Table 1).
Hypothesis 2— Correlation strength varies
with analysis radius
This hypothesis was accepted for all com-
parisons with significant r values, although
the results differed by seascape variable (Table 1).
Most peak correlations involving area of seagrass/SAV
occurred at the broadest scale considered (800 m, e.g.
Fig. 4). Transient richness and abundance had highest
correlation at 800 m with sand/sediment area, whereas
resident abundance had highest correlation at 25 m.
Patterns for correlations between individual species
and seascape variables were less predictable (Table 1).
The abundance of adult Cephalopholis fulvus had
highest correlation with sand/sediment area at 100 m,
whereas overall and juvenile abundance had highest
correlation at 25 m. Most peak correlations with length
of hard bottom edge occurred at the 800 m scale except
for total and adult abundance of Haemulon flavolinea-
tum and species richness of mobile species that oc -
curred at 50 m. All peak |r| values showed positive rela-
tionships except for C. fulvus, which showed strong
negative correlations.
Hypothesis 3— Map resolution affects correlation
strength and scale
Support for this hypothesis was equivocal among
seascape variables, with spatial and thematic resolu-
tion influencing some results but not others. For fish–
habitat diversity comparisons, all 4 map types yielded
similar results, with |r| values rarely exceeding 0.1
across all scales (e.g. Fig. 3). For seagrass/SAV area,
the scale of highest correlation was the same for all 4
map types in all but one of the 14 comparisons with sig-
nificant results. In contrast, all sand/sediment area
results were strongly influenced by map type. Maps of
the same spatial resolution resulted in similar r values
at all spatial scales, whereas maps with differing spa-
tial resolution resulted in very different values across
scales (e.g. Fig. 5). More specifically, the abundance of
adult Cephalopholis fulvus had highest correlation
with sand/sediment area at 100 m, whereas overall and
juvenile abundance had highest correlation at 25 m.
These were perceived as positive relationships only
when maps with low thematic resolution were used.
For length of hard bottom edge, map type significantly
influenced the results for all but one of the 17 compar-
isons with at least one |r| > 0.2. Only species richness of
residents was consistently correlated with hard bottom
edge by all 4 map types (Fig. 6). For the 16 other com-
parisons, use of different map types resulted in either
significantly lower r at the same scale or even a peak in
r at an entirely different scale.
Changes in perception of fish–seascape correlation due
to map type are summarized at the bottom of Table 1
(i.e. either peak in correlation at different scale or peak
at the same scale but different strength). Map type had
no effect on correlations involving habitat diversity
with no significant correlations observed with any map.
When map type had an effect on the sand/sediment
results, maximum |r| value occurred at completely dif-
ferent scales (14 occurrences) rather than simply peak-
ing at the same scale but at a significantly lower value
(0 occurrences). In contrast, seagrass/SAV and hard
bottom edge relationships showed some of each type of
difference. Overall, the 2 types of differences occurred
with approximately equal frequency.
The number of times that each map type had the
highest or lowest |r| value when a |r| > 0.2 was present
is tallied in Table 2. Only hard bottom edge compar-
isons yielded a consistent pattern. Maps with high spa-
266
Fig. 4. Correlations of seagrass/submerged aquatic vegetation and Haemu-
lon flavolineatum adult abundance by analysis distance for all 4 map types.
Where a significant correlation was found (|r| > 0.2 as indicated by the
horizontal dashed lines), the scale with the strongest relationship is noted
with the black arrow. Vertical lines adjacent to the legend denote map types
that have correlations that are not significantly different from each other at
the scale with highest correlation. See Fig. 3 for abbreviations
tial and thematic resolution most often had |r| values
significantly higher than other maps. Maps with low
spatial and thematic resolution also had significantly
lowest |r| values in the most comparisons. No single
map type consistently differed from the others for the
other 3 seascape variables.
Hypothesis 4— Range in correlation strength varies
by life stage, guild, and whole community variables
Maximum |r| values showed similar minima, max-
ima, and ranges among individual life stages of the 6
focal species and when all life stages were grouped
together. Values for mobility guilds and whole fish
community results were also similar to each other but
quite different from those based on individ-
ual species (Fig. 7). Findings were therefore
grouped into these 2 broader categories,
respectively. Of the 104 fish– seascape combi-
nations tested, at least 11 of the highest max-
imum |r| values were for species-level analy-
ses. This was true for all map types except for
high spatial and low thematic resolution,
which had only 4 of the highest values
(Fig. 7d). Species-level analyses also had a
higher range of values (~0.4), much higher
than the range for guild or community com-
parisons (~0.2) (Fig. 7a– c). The exception was
again for analyses based on maps with high
spatial and low thematic resolution, which
differed from this pattern in that the range
of values was lower (~0.3) (Fig. 7d). Also of
note, nearly all of the highest |r| values were
for comparisons involving hard bottom edge
length and seagrass/SAV, whereas nearly all
of the lowest values were for correlations
between habitat diversity and individual fish
species.
Hypothesis 5— Mobility guilds and life
stage will affect distance of peak correlation
Overall, 38% (9 of 24) of the comparisons
had maximum correlations at the same scale
for resident, mobile, and transient species
(Table 3). The next most common result (30%,
7 of 24 of the comparisons), occurred when
transient species had a larger scale of correla-
tion than either resident or mobile species
(which had a common scale of peak correla-
tion). The expected result of r value trends:
resident < mobile < transient, occurred in only
one of the 24 comparisons evaluated. Also of
note, no fish– seascape correlations based on mobility
yielded the same results for all 4 map types, and differ-
ences were unpredictable and inconsistent.
Overall, 56% (40 of the 72) of the comparisons eval-
uated had maximum correlation at the same scale for
both adults and juveniles of a given species (Table 4).
Juveniles had a finer scale of peak correlation in only
15% (11 of 72) of the comparisons, whereas adults had
finer scales of peak correlation in 30% (21 of 72) of the
comparisons. Of note, when a difference was found in
comparisons involving seagrass/SAV, it was always
that adults had a finer scale of peak correlation than
juveniles. All 4 map types generally resulted in the
same patterns. Exceptions to this were for Ocyurus
chrysurus and Cephalopholis fulvus. For O. chrysurus,
use of maps with high spatial resolution resulted in
Kendall et al.: Perception of fish–seascape correlations 267
Fig. 5. Correlations of sand/sediment area and resident fish abundance
by analysis distance for all four map types. See Figs. 3 & 4 for definitions
and abbreviations
Fig. 6. Correlations of hard bottom edge and resident species richness
by analysis distance. See Figs. 3 & 4 for definitions and abbreviations
Mar Ecol Prog Ser 427: 259–274, 2011
juveniles having finer scales of peak
correlation than adults. When low
spatial resolution was used, the in -
verse pattern was perceived. For C.
fulvus, use of maps with high spatial
resolution resulted in adults having
finer scales of peak correlation than
juveniles. When low spatial resolution
was used, the inverse pattern was per-
ceived. It should be noted that infer-
ence regarding scales of peak correla-
tion by life stage are limited to only
the 6 focal species tested, whereas
results for mobility guilds a more
broadly robust and are based on all
species observed.
268
Map resolution Habitat Seagrass/ Sand/ Hard bottom
diversity SAV sediment edge length
Significantly highest |r| value
High spatial high thematic 0 0 0 4
High spatial low thematic 0 0 0 0
Low spatial high thematic 0 0 0 1
Low spatial low thematic 0 1 0 0
Significantly lowest |r| value
High spatial high thematic 0 0 0 0
High spatial low thematic 0 0 0 3
Low spatial high thematic 0 0 0 0
Low spatial low thematic 0 0 0 4
Table 2. Tally of the number of times each map type had the highest or lowest |r|
value when a significant result was present. SAV: submerged aquatic vegetation
Maximum |r|
All life stages
n = 24
Individual life stages
n = 48
Mobility guild
n = 24
Whole c ommun ity
n = 8
0
0.1
0.2
0.3
0.4
0.5
Maximum |r|
0
0.1
0.2
0.3
0.4
0.5
0.1
0.2
0.3
0.4
0.5
0.1
0.2
0.3
0.4
0.5
Range: 0.41 0.39 0.21 0.17 Range: 0.43 0.45 0.24 0.23
All life stages
n = 24
Individual life stages
n = 48
Mobility guild
n = 24
Whole c ommun ity
n = 8
All life stages
n = 24
Individual life stages
n = 48
Mobility guild
n = 24
Whole c ommun ity
n = 8 All life stages
n = 24
Individual life stages
n = 48
Mobility guild
n = 24
Whole c ommun ity
n = 8
Range: 0.40 0.47 0.21 0.20 Range: 0.31 0.34 0.25 0.15
ab
cd
Fig. 7. Maximum |r| values for all 104 xand yvariables investigated in the study using maps with (a) high spatial and thematic res-
olution, (b) low spatial and high thematic resolution, (c) low spatial and thematic resolution, and (d) high spatial and low thematic
resolution. Results are grouped by level of organization of the fish variables. Symbols denote the landscape variables associated
with each |r| value. d: habitat diversity; s: hard bottom edge length; z: seagrass/SAV; y: sand/sediment
Kendall et al.: Perception of fish–seascape correlations
DISCUSSION
A central question asked in the present study is
‘How much of the pattern in fish distribution can be
explained using landscape variables?’ The seascape
pattern metrics selected for study were considered to
be among those with the greatest explanatory power
over fish distributions based on prior research. Results
suggest, however, that each of the seascape variables
studied here explain only a relatively small proportion
(~25%) of the variability in the distribution of fish in
coral reef systems.
Similar studies in a variety of sys-
tems have generally yielded a similar
range of correlation values to those
found here. Landscape variables ex -
plained 2 to 64% of the variability in
bird and insect communities (Ricketts
et al. 2001, Pearman 2002, Steffan-
Dewenter 2003, Holland et al. 2004),
and although less-studied, findings
from multiscale studies of reef fish are
similarly wide ranging, with 11 to 94%
of the variability explained between
seascape and fish variables (Kendall
2005, Grober-Dunsmore et al. 2007).
Linear correlation between fish species
richness on sand sites with area of
nearby hard bottom reached maximum
values of r = 0.33 in a separate study at
BIRNM (Kendall 2005). Grober-Duns -
more et al. (2007) reported linear cor-
relations between reef fish community
variables and area of seagrass as high
as 0.97 and were often in the range of
~0.5 to 0.6 around the nearby island of
St. John, U.S. Virgin Islands. These dif-
ferences in results for ecologically sim-
ilar coral reef ecosystems from the
same geographical area are likely
the result of differences in sampling
design between the 2 studies. Studies
by Kendall (2005) and the results here
were based on a large number of
spatially random survey sites, whereas
findings of Grober-Dunsmore et al.
(2007) were based on a subset of se -
lected coral reef sites chosen specifi-
cally to quantify the effects of variation
in the amount of nearby seagrass cover
and to minimize confounding variables
such as differences in coral cover, ru -
gosity, depth, and distance from shore.
In contrast, our study provides a more
comprehensive, ecosystem-wide mea-
sure of the strength of the relationships across the
complete range of coral reefs in the study area.
What seascape variables had the highest or most cor-
relations with the fish variables? Habitat diversity has
been considered as a proxy for fish diversity in the
selection of marine reserves (National Research Coun-
cil 2001). Terrestrial studies have shown a relationship
between habitat diversity and biotic diversity for a
range of taxa (Kohn & Walsh 1994, Kerr & Packer 1997,
Ricklefs & Lovette 1999, Fox & Fox 2000). However,
our results suggest that habitat diversity is a very poor
predictor of fish species richness or indeed any compo-
269
High High Low Low
spatial spatial spatial spatial
high low high low
thematic thematic thematic thematic
Hard bottom edge length
Fish abundance R = M < T M < R = T R = M < T R = M < T
Species richness R = M = T R = M = T R = M < T R = M = T
Seagrass/SAV
Fish abundance R = M = T R = M < T M < R < T R < M < T
Species richness R = M = T R = M < T R < M = T R < M = T
Sand/sediment
Fish abundance R = M = T T < R = M R = M = T R = M < T
Species richness M < R = T T < R = M R = M = T R = M = T
Table 3. Relative scale of maximum |r| values for resident (R), mobile (M), and
transient (T) fish within the 4 map types. SAV: submerged aquatic vegetation
High High Low Low
spatial spatial spatial spatial
high low high low
thematic thematic thematic thematic
Hard bottom edge length
Acanthurus coeruleus J = A J = A J = A A < J
Cephalopholis fulvus J = A J = A J = A J = A
Haemulon flavolineatum J = A J = A J = A J = A
Ocyurus chrysurus J < A J < A A < J A < J
Sparisoma viride J = A A < J J = A A < J
Stegastes planifrons J < A J = A J = A J = A
Seagrass/SAV
A. coeruleus A < J A < J A < J A < J
C. fulvus J = A J = A J = A J = A
H. flavolineatum J = A J = A J = A J = A
O. chrysurus A < J A < J A < J A < J
S. viride A < J A < J A < J A < J
S. planifrons J = A J = A J = A J = A
Sand/sediment
A. coeruleus A < J A < J J = A J = A
C. fulvus A < J A < J J < A J < A
H. flavolineatum J = A J = A J < A J < A
O. chrysurus J = A J = A J = A J = A
S. viride J = A J = A J = A J = A
S. planifrons J < A J < A J < A J < A
Table 4. Relative scale of maximum |r| values for juveniles (J) versus adults (A) of each
of the 6 focal species within the 4 map types. SAV: submerged aquatic vegetation
Mar Ecol Prog Ser 427: 259–274, 2011
nent of the fish community considered. A possible
explanation is that benthic maps of the type used here
may not capture the aspects of habitat diversity to
which fish respond. It is also possible that, although we
evaluated a wide range of variables representing the
fish assemblage, the species and assemblages consid-
ered may be habitat generalists or have considerable
plasticity in suitable habitats (Ricklefs & Lovette 1999).
Our results bolster growing evidence against using
habitat diversity at the seascape scale, as depicted in
benthic maps, as a proxy for predicting overall fish and
biotic diversity (Donaldson 2002, Pittman et al. 2007,
Grober-Dunsmore et al. 2008).
Area of seagrass/SAV was correlated with several of
the fish community variables, including at least one life
stage of most species tested. This confirms prior studies
on species suspected of association with seagrass/SAV
and further quantifies those relationships (Dorenbosch
et al. 2005, 2006, Kendall 2005, Grober-Dunsmore et al.
2007, Valentine et al. 2007). Correlations were also
found for species not previously thought to be related to
area of seagrass (e.g. Cephalopholis fulvus, Sparisoma
viride, and Stegastes planifrons). This demonstrates the
importance of seagrass/SAV as an influence on fish dis-
tribution on reefs generally, not just those considered
obligate users. It also indicates that a variety of direct
and indirect mechanisms can operate that influence
abundance of particular species or guilds. Sand/sedi-
ment area predicted several of the fish variables, al-
though not as many as expected given this bottom
type’s role in settlement and foraging of many species.
Length of hard bottom edge was correlated with
more of the fish variables than any other landscape
feature. This underscores its role as an important habi-
tat margin to a diversity of reef fish (Dorenbosch et al.
2005, 2006, Valentine et al. 2007). Edges between reef
types and soft bottom often have high rugosity that
offers structural refuge supporting a diversity of reef
species (Pittman et al. 2007).
Correlations were found between diverse elements of
the fish community and seascape features at a wide
range of distances. Systematically changing the size of
the analysis window and comparing fit among the mod-
els allowed the neighborhood that explains the highest
amount of variability (highest |r|) in the fish data to be
identified. The distance or neighborhood with the
strongest correlation has been interpreted as the most
ecologically influential or relevant scale for each com-
bination of biotic and seascape variable (Holland et al.
2004, Kendall 2005). For many comparisons no signifi-
cant relationships were found for any fish variables at
any scale. In these instances, a number of factors may
be responsible. Fish variables may be more closely re-
lated to a seascape variable not tested in this study, or
the relationship may be non-linear and insensitive to
our linear-regression-based approach. It could also be
that the seascape maps did not adequately capture the
necessary detail of the seascape parameters that were
tested. Fish may even be responding to seascape fea-
tures beyond our maximum analysis distance.
Ecologically meaningful explanations are present for
many of the observed patterns in neighborhood dis-
tance and associations with particular landscape vari-
ables. Species richness of fish was positively correlated
with area of sand/sediment, area of seagrass/SAV, and
length of hard bottom edge. Correlation with these
variables increased with analysis distance such that
maximum r values occurred at the 800 m scale, a
broader scale of peak correlation than identified by
prior research (400 m by Kendall 2005; 500 m by
Grober-Dunsmore et al. 2007). It has long been
thought that the area of surrounding seagrass in -
creases the number of fish species on hard bottom sites
by providing foraging areas for some species (Randall
1965, Ogden 1976, Nagelkerken et al. 2000), transfer
of energy to reefs (Meyer et al. 1983, Meyer & Shultz
1985), nursery habitat (Dorenbosch et al. 2005, Adams
et al. 2006, Dorenbosch et al. 2007, Verweij et al. 2008),
and enhanced recruitment (Shulman & Ogden 1987,
Cocheret de la Morinière et al. 2002). Similarly, area of
surrounding sand bottom may result in enhanced
recruitment to nearby hard bottom sites of the many
species that initially settle in sand habitat to avoid reef
and reef edge predators (Helfman et al. 1982, Shulman
1985, Shulman & Ogden 1987). Species richness may
be enhanced by length of hard bottom edge through
several mechanisms. Hard bottom edge must be tran-
sited for juvenile fish undergoing ontogenetic shifts
following settlement in sand or seagrass (Shulman 1985,
Shulman & Ogden 1987, Cocheret de la Morinière et
al. 2002), it is a preferred hunting ground of some
pisci vores (Helfman et al. 1982, Quinn & Ogden 1984,
Sweatman & Robertson 1994), and is the optimum
location to seek structural refuge to minimize travel
distance from reef to soft bottom for species that
undergo such daily foraging migrations (Kendall et al.
2003, Tuya et al. 2010). Hard bottom edge represents a
key ecotone habitat for many species (Wernberg et al.
2006, Valentine et al. 2007, Vanderklift et al. 2007),
and also indicates the presence of bathymetric com-
plexity between reef types or reef and soft bottom,
which has been positively correlated with species rich-
ness of fish (Luckhurst & Luckhurst 1978, Gratwicke &
Speight 2005a,b, Pittman et al. 2007).
Ecologically meaningful correlations were also found
between individual species and seascape variables.
Strong negative correlations were observed be tween
Cephalopholis fulvus and length of hard bottom edge
and area of seagrass/SAV. In both cases r values
steadily decreased with analysis distance to a maxi-
270
Kendall et al.: Perception of fish–seascape correlations
mum at the 800 m scale. This species utilizes flat hard
bottom often sparsely colonized by corals, sponges, and
gorgonians (Pittman et al. 2008), a bottom type often
described as pavement that typically covers broad ar-
eas (Kendall et al. 2004b). Hard bottom edges or a large
area of seagrass nearby would mean that there is less of
their preferred flat hard bottom habitat. Logical eco -
logical correlations were also observed between
seascape variables and Haemulon flavolineatum adult
and overall abundance. This species feeds solitarily
over seagrass and soft bottom at night but schools over
reefs and hard bottom during the day (Randall 1965,
Ogden 1976). Area of seagrass positively influenced
abundance on reef sites by providing a large foraging
area (Burke 1995, Nagelkerken et al. 2000, Kendall et
al. 2003), especially at long analysis distances that may
correspond to a broad foraging range. High correlation
with hard bottom edge, especially at very short analysis
distances, is logical, too, because optimality theory pre-
dicts that H. flavolineatum will utilize reef sites near
reef edges (Kendall et al. 2003). Such proximity mini-
mizes energy costs and daily travel time from resting
sites on reefs to adjacent seagrass foraging areas. This
relationship was apparent only when maps with high
spatial and thematic resolution were used.
More difficult to explain were the strong correlations
observed between other variables. For example, a pos-
itive correlation was observed between Stegastes
planifrons and both area of seagrass/SAV and length
of hard bottom edge. High correlations were measured
at the 800 m analysis scale. This highly resident spe-
cies settles directly onto reefs (Tolimieri 1995, Gutier-
rez 1998) and spends its benthic life associated with
the same coral head or <~1 m2territory (Luckhurst
& Luckhurst 1978, Robertson et al. 1981). That either
of these landscape variables or this analysis distance
have a direct influence on fish abundance is doubtful.
These seascape variables may instead be correlated
with some other environmental factor, some indirect
effect may be responsible, and we are re minded that
correlation need not be obviously linked to causation.
In many comparisons, use of different input maps
resulted in a changed perception of either the strength
of peak correlation at a given scale, or the scale at
which peak correlations occurred. The latter case rep-
resents a more serious problem in that both the spatial
dimensions as well as the intensity of the relationship
are perceived differently. Such events call for the most
careful consideration of the consequences of relying on
a particular map type. These 2 types of misperception
occurred with different frequency depending on the
seascape feature tested. Studies relying on the amount
of hard bottom edge length and area of sand need to
be cautiously interpreted due to the large number of
cases where map type changed the perception of the
fish– seascape correlation. Spatial resolution of maps
often completely changed the perceived relationships
between fish and their area of surrounding sand/sedi-
ment. In all cases, use of high spatial resolution maps
resulted in lower r values or even negative r values
compared to low spatial resolution maps at the same
analysis scale. Perception of correlation strength be -
tween fish and hard bottom edge also depended on the
type of input maps used. While the general patterns of
increasing correlation with scale were similar among
all 4 map types, the values of the correlation were often
significantly different. Maps of the study site exhibited
a doubling of edge length for hard bottom features
when high spatial resolution was used to create them
(Kendall & Miller 2008). Many reef edges that fish
interact with, such as small patch reefs in sand and
sand channels in hard bottom, only appeared when
high spatial resolution was used. In contrast, results
were quite consistent for seagrass/SAV area regard-
less of map type. Continuous seagrass beds were char-
acterized quite consistently at the 2 map scales used in
the present study, but patchy beds showed large dif -
ferences (Kendall & Miller 2008). For habitat diversity,
all 4 map types performed similarly in that none had
significant correlations with any fish variables.
Is there a particular map type that is best for sea -
scape ecological studies of reef fish? Our results sug-
gest that the answer depends on the seascape vari-
ables of interest. Maps with high spatial and thematic
resolution had most of the significantly highest correla-
tions for comparisons involving hard bottom edge,
whereas maps with low spatial and thematic resolution
were often lowest. This indicates that results of studies
using hard bottom edge are likely inaccurate when
using lower spatial or thematic resolution maps. In
contrast, all 4 map types performed similarly for sea-
grass/SAV, indicating that even simple, inexpensive to
produce maps do just as well as highly detailed,
expensive, time-consuming maps in studies involving
this variable. Also of relevance are the plots of maxi-
mum r values by level of organization of fish variables.
All map types yielded a similar range of results except
for maps with high spatial but low thematic resolution.
This map type had lower sensitivity to detecting the
highest and lowest peak correlations that were ob -
served more consistently among the other map types.
This indicates that mapping only a few bottom types
with great spatial detail may be least effective in
seascape ecological studies. Why such maps would
perform more poorly than those with low thematic as
well as low spatial resolution is unclear.
Maximum correlations between seascape variables
and individual species achieved a wider range and
more extreme values (highest and lowest) than com-
parisons involving either guilds or the entire fish
271
Mar Ecol Prog Ser 427: 259–274, 2011272
assemblage. Variables representing more than a single
species had more moderate peak correlations. This is
likely because the habitat preferences and scales of
movement of the many species included in such vari-
ables get averaged together and limit extreme values.
In contrast, individual species had both highest and
lowest values since each species interacts with a more
discrete set of habitats at similar scales. This pattern
did not however, separate the results of individual life
stages from all individuals of the focal species, as was
expected, nor did it distinguish between mobility
guilds and whole community metrics.
Scales of peak correlation were the same for juve-
niles and adults in over half of the comparisons. The
expectation based on terrestrial literature (Holling
1992, Gehring & Swihart 2003, Holland et al. 2005),
that juveniles would have a shorter distance of maxi-
mum correlation than adults, rarely occurred (but see
Grober-Dunsmore et al. 2007). This suggests that
seascape influences on the distribution of juvenile fish
may operate at scales often as broad as those for their
adult stages. Typical scales of seascape interaction for
mobility guilds were somewhat more in line with
expectations (Pearman 2002), in that transients had
broader scales of peak correlation than either resident
or mobile fish in a large number of comparisons. Still,
however, scale of influence was the same for all 3
mobility guilds in many comparisons, again indicating
that in many cases even resident fish are influenced by
their surrounding seascape at distances as broad as
those for transients. Despite peak correlation at similar
scales, the mechanisms responsible are almost cer-
tainly indirect given present understanding of the very
small home range of resident species and juveniles of
the 6 focal species (Itzkowitz 1977, Luckhurst & Luck-
hurst 1978, Overholtzer & Motta 1999, Bell & Kramer
2000, Watson et al. 2002). Map type generally did not
influence the results of peak scale for adult versus
juvenile fish. In contrast, results of mobility guild
analysis differed in unpredictable ways depending on
map type, again indicating that caution be used when
studying mobility guilds using a single map type.
Most prior seascape ecological studies base results
on one type of map; whatever is available. Little con-
sideration appears to have been given to the influence
of map type on the conclusions reached. Terrestrial
investigations have shown that the characteristics of
input maps can influence results of landscape ecology
studies (Stohlgren et al. 1997, Karl et al. 2000). Results
here also suggest that use of a single map type in the
marine environment can lead to an incomplete or even
incorrect perception (i.e. undetected, weakly mea-
sured, inversely signed, thought to occur at the wrong
scale) of habitat utilization and scale at which organ-
isms interact with their seascape.
Based on the findings here, the following advice can
be given to those interested in mapping coral reef
ecosystems to study seascape ecology of reef fish, to
model species distributions, or in making spatially ex-
plicit management decisions using benthic maps. Hard
bottom should be mapped with high spatial resolution
above all else since this most affects reef edge depic-
tions. Time and money permitting, hard bottom should
be mapped with high thematic resolution as well and
separated into its various reef types. Many studies are
presently concerned with hard bottom edge and prox-
imity to hard bottom habitat (Sweatman & Robertson
1994, Dorenbosch et al. 2005, Wernberg et al. 2006,
Valentine et al. 2007, Vanderklift et al. 2007, Tuya et al.
2010). Extrapolating their mostly in situ studies to
seascape scales using benthic maps carries with it par-
ticular concerns. Sand should be mapped with high
spatial resolution to pick up key features such as sand
channels in hard bottom and halos separating hard bot-
tom from seagrass (Kendall & Miller 2008). In contrast
to these bottom types, seagrass mapped at coarse the-
matic and spatial resolution appear to effectively evalu-
ate the seascape ecology of a variety of fish species and
will result in similar values when more detailed maps
are used. Given these findings, prior seagrass studies
probably do not need to be concerned about their re-
sults changing if different map types were used (e.g.
Pittman et al. 2004, Kendall 2005, Grober-Dunsmore et
al. 2007). Results involving hard bottom or sand, how-
ever, could change measurably were different maps to
be used as input. Habitat diversity, as measured by the
type of benthic maps used here, is simply not represen-
tative of fish diversity or any other measure of the fish
community at any scale and should not be considered
as a surrogate or proxy variable for overall biodiversity.
To keep these recommendations in perspective, how-
ever, seascape variables that were used here were for
common bottom features. Habitat specialists that are
obligate users of a particular reef type, for example,
would need to be studied with a map of sufficient spa-
tial and thematic complexity to capture such features.
Acknowledgements. This is contribution No. 4474 of the Uni-
versity of Maryland Centre for Environmental Science and
was drawn in part from M.S.K.’s dissertation.
LITERATURE CITED
Adams AJ, Dahlgren CP, Kellison GT, Kendall MS and others
(2006) Nursery function of tropical back-reef systems. Mar
Ecol Prog Ser 318:287– 301
Addicott JF, Aho JM, Antolin MF, Padilla DK, Richardson JS,
Soluk DA (1987) Ecological neighborhoods: scaling envi-
ronmental patterns. Oikos 49:340–346
Andréfouët S, Robinson JA, Hu C, Feldman GC, Salvat B,
Payri C, Muller-Karger FE (2003) Influence of the spatial
➤
➤
Kendall et al.: Perception of fish–seascape correlations
resolution of SeaWiFS, Landsat-7, SPOT, and Interna-
tional Space Station data on estimates of landscape para-
meters of Pacific Ocean atolls. Can J Rem Sens 29:210– 218
Bell T, Kramer DL (2000) Territoriality and habitat use by
juvenile blue tangs, Acanthurus coeruleus. Environ Biol
Fishes 58:401–409
Benson BJ, MacKenzie MD (1995) Effects of sensor spatial
resolution on landscape structure parameters. Landscape
Ecol 10:113–120
Burke NC (1995) Nocturnal foraging habitats of French and
bluestriped grunts, Haemulon flavolineatum and H. sciu-
rus, at Tobacco Caye, Belize. Environ Biol Fishes 42:
365– 374
Chapman MR, Kramer DL (2000) Movements of fishes within
and among fringing coral reefs in Barbados. Environ Biol
Fishes 57:11–24
Cocheret de la Morinière E, Pollux BJA, Nagelkerken I, van
der Velde G (2002) Post-settlement life cycle migration
patterns and habitat preference of coral reef fish that use
seagrass and mangrove habitats as nurseries. Estuar Coast
Shelf Sci 55:309– 321
Donaldson TJ (2002) High islands versus low islands: a com-
parison of fish faunal composition of the Palau Islands.
Environ Biol Fishes 65:241–248
Dorenbosch M, Grol MGG, Nagelkerken I, van der Velde G
(2005) Distribution of coral reef fishes along a coral reef –
seagrass gradient: edge effects and habitat segregation.
Mar Ecol Prog Ser 299:277–288
Dorenbosch M, Grol MGG, Nagelkerken I, van der Velde G
(2006) Different surrounding landscapes may result in dif-
ferent fish assemblages in east African seagrass beds.
Hydrobiol 563:45–60
Dorenbosch M, Verberk WCEP, 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
FMRI NOAA (Florida Marine Research Institute and National
Oceanic and Atmospheric Administration) (1998) Benthic
habitats of the Florida Keys. Fla Mar Res Inst Tech Rep
TR-4
Fox BJ, Fox MD (2000) Factors determining mammal species
richness on habitat islands and isolates: habitat diversity,
disturbance, species interactions and guild assembly
rules. Glob Ecol Biogeogr 9:19–37
Gehring TM, Swihart RK (2003) Body size, niche breadth, and
ecologically scaled responses to habitat fragmentation:
mammalian predators in an agricultural landscape. Biol
Conserv 109:283–295
Gratwicke B, Speight MR (2005a) The relationship between
fish species richness, abundance and habitat complexity
in a range of shallow tropical marine habitats. J Fish Biol
66:650– 667
Gratwicke B, Speight MR (2005b) Effects of habitat complex-
ity on Caribbean marine fish assemblages. Mar Ecol Prog
Ser 292:301–310
Grober-Dunsmore R, Frazer TK, Lindberg WJ, Beets J (2007)
Reef fish and habitat relationships in a Caribbean
seascape: the importance of reef context. Coral Reefs 26:
201–216
Grober-Dunsmore R, Frazer TK, Beets J, Lindberg WJ, Zwick
P, Funicelli NA (2008) Influence of landscape structure on
reef fish assemblages. Landscape Ecol 23:37– 53
Gutierrez L (1998) Habitat selection by recruits establishes
local patterns of adult distribution in two species of
damselfishes: Stegastes dorsopunicans and S. planifrons.
Oecologia 115:268–277
Helfman GS, Meyer JL, McFarland WN (1982) The ontogeny
of twilight migration patterns in grunts (Pisces: Haemuli-
dae). Anim Behav 30:317–326
Holland JD, Bert DG, Fahrig L (2004) Determining the spatial
scale of species response to habitat. Bio Sci 54:227–233
Holland JD, Fahrig L, Cappuccino N (2005) Body size affects
the spatial scale of habitat-beetle interactions. Oikos 110:
101–108
Holling CS (1992) Cross-scale morphology, geometry, and
dynamics of ecosystems. Ecol Monogr 62:447–502
Irlandi EA, Crawford MK (1997) Habitat linkages: the effect
of intertidal saltmarshes and adjacent subtidal habitats on
abundance, movement, and growth of an estuarine fish.
Oecologia 110:222–230
Itzkowitz M (1977) Spatial organization of the Jamaican dam-
selfish community. J Exp Mar Biol Ecol 28:217–241
Karl JW, Heglund PJ, Garton EO, Scott JM, Wright NM, Hutto
RL (2000) Sensitivity of species habitat-relationship model
performance to factors of scale. Ecol Appl 10:1690–1705
Kendall M (2005) A method for investigating seascape ecol-
ogy of reef fish. Gulf Caribb Fish Ins 56:355–366
Kendall MS, Miller TJ (2008) The influence of spatial and the-
matic resolution on maps of a coral reef ecosystem. Mar
Geod 31:75–102
Kendall MS, Kruer CR, Buja KR, Christensen JD, Finkbeiner
M, Warner R, Monaco ME (2001) Methods used to map the
benthic habitats of Puerto Rico and the U.S. Virgin Islands.
NOAA NOS NCCOS CCMA Tech Rep 152. National
Oceanic and Atmospheric Administration, Silver Spring,
MD
Kendall MS, Christensen JD, Hillis-Starr Z (2003) Multi-scale
data used to analyze the spatial distribution of French
grunts, Haemulon flavolineatum, relative to hard and soft
bottom in a benthic seascape. Environ Biol Fishes 66:
19–26
Kendall MS, Buja KR, Christensen JD, Kruer CR, Monaco ME
(2004a) The seascape approach to coral ecosystem map-
ping: an integral component of understanding the habitat
utilization patterns of reef fish. Bull Mar Sci 75:225–237
Kendall MS, Christensen JD, Caldow C, Coyne M and others
(2004b) The influence of bottom type and shelf position on
biodiversity of tropical fish inside a recently enlarged
marine reserve. Aquatic Cons Mar Freshwat Ecosys 14:
113–132
Kerr JT, Packer L (1997) Habitat heterogeneity as a determi-
nant of mammal species richness in high-energy regions.
Nature 385:252–254
Kohn DD, Walsh DM (1994) Plant species richness—the effect
of island size and habitat diversity. J Ecol 82:367–377
Kramer DL, Chapman MR (1999) Implications of fish home
range size and relocation for marine reserve function.
Environ Biol Fishes 55:65–79
Luckhurst BE, Luckhurst K (1978) Analysis of the influence of
substrate variables on coral reef fish communities. Mar
Biol 49:317–323
Meyer JL, Shultz ET (1985) Migrating haemulid fishes as
a source of nutrients and organic matter on coral reefs.
Limnol Oceanogr 30:146–156
Meyer JL, Shultz ET, Helfman GS (1983) Fish schools: an
asset to corals. Science 220:1047–1049
Morgan IE, Kramer DL (2004) The social organization of adult
blue tangs, Acanthurus coeruleus, on a fringing reef, Bar-
bados, West Indies. Environ Biol Fishes 71:261–273
Mumby PJ, Harborne AR (1999) Development of a systematic
classification scheme of marine habitats to facilitate
regional management and mapping of Caribbean coral
reefs. Biol Conserv 88:155–163
273
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤➤
➤
➤
Mar Ecol Prog Ser 427: 259–274, 2011
Mumby PJ, Wabnitz CCC (2002) Spatial patterns of aggres-
sion, territory size, and harem size in five sympatric
Caribbean parrotfish species. Environ Biol Fishes 63:
265–270
Nagelkerken I, Dorenbosch M, Verberk WCEP, Cocheret de
la Morinière E, van der Velde G (2000) Day-night shifts of
fishes between shallow-water biotopes of a Caribbean
bay, with emphasis on the nocturnal feeding of Haemuli-
dae and Lutjanidae. Mar Ecol Prog Ser 194:55–64
National Research Council (2001) Marine protected areas:
tools for sustaining ocean ecosystems. National Academy
Press, Washington, DC
NOAA NCCOS (National Oceanic and Atmospheric Adminis-
tration National Centers for Coastal Ocean Science) (2004)
Shallow-water benthic habitats of American Samoa,
Guam, and Commonwealth of the Northern Mariana
Islands. Technical Memorandum NOS NCCOS 8, Bio-
geography Team, Silver Spring, MD. (CD ROM)
Ogden JC (1976) Some aspects of herbivore-plant relation-
ships on Caribbean reefs and seagrass beds. Aquat Bot
2:103–116
Overholtzer KL, Motta PJ (1999) Comparative resource use by
juvenile parrotfishes in the Florida Keys. Mar Ecol Prog
Ser 177:177–187
Parrish JD (1989) Fish communities of interacting shallow-
water habitats in tropical oceanic regions. Mar Ecol Prog
Ser 58:143–160
Pearman PB (2002) The scale of community structure: habitat
variation and avian guilds in tropical forest understory.
Ecol Monogr 72:19– 39
Pearson SM (1993) The spatial extent and relative influence of
landscape-level factors on wintering bird populations.
Landscape Ecol 8:3–18
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, Christensen JD, Caldow C, Menza C, Monaco ME
(2007) Predictive mapping of fish species richness across
shallow-water seascapes in the Caribbean. Ecol Model
204:9–21
Pittman SJ, Hile SD, Jeffrey CFG, Caldow C, Kendall MS,
Monaco ME, Hillis-Starr Z (2008) Fish assemblages and
benthic habitats of Buck Island Reef National Monument
(St. Croix, U.S. Virgin Islands) and the surrounding
seascape: a characterization of spatial and temporal pat-
terns. NOAA Technical Memorandum NOS NCCOS 71,
Silver Spring, MD
Prada MC, Appeldoorn RS, Rivera JA (2008) The effects of
minimum map unit in coral reefs maps generated from
high resolution side scan sonar mosaics. Coral Reefs
27:297–310
Quinn TP, Ogden JC (1984) Field evidence of compass orien-
tation in migrating juvenile grunts (Haemulidae). J Exp
Mar Biol Ecol 81:181–192
Randall JE (1965) Grazing effect on sea grasses by herbivo-
rous reef fishes in the West Indies. Ecol 46:255–260
Ricketts TH, Daily GC, Ehrlich PR, Fay JP (2001) Countryside
biogeography of moths in a fragmented landscape: biodi-
versity in native and agricultural habitats. Conserv Biol
15:378– 388
Ricklefs RE, Lovette IJ (1999) The roles of island area per se
and habitat diversity in the species-area relationships of
four Lesser Antillean faunal groups. J Anim Ecol 68:
1142–1160
Riitters KH, O’Neill RV, Jones KB (1997) Assessing habitat
suitability at multiple scales: a landscape-level approach.
Biol Conserv 81:191–202
Robertson DR, Hoffman SG, Sheldon JM (1981) Availability of
space for the territorial Caribbean damselfish Eupoma-
centrus planifrons. Ecol 62:1162–1169
Sale PF (1998) Appropriate spatial scales for studies of reef-
fish ecology. Aust J Ecol 23:202–208
Saura S (2002) Effects of minimum mapping unit on land
cover data spatial configuration and composition. Int J
Remote Sens 23:4853– 4880
Shulman MJ (1985) Recruitment of a coral reef fish: effects of
the spatial distribution of predators and shelter. Ecol 66:
1056–1066
Shulman MJ, Ogden JC (1987) What controls tropical reef fish
populations: recruitment or benthic mortality? An exam-
ple in the Caribbean reef fish Haemulon flavolineatum.
Mar Ecol Prog Ser 39:233–242
Steffan-Dewenter I (2003) Importance of habitat area and land-
scape context for species richness of bees and wasps in
fragmented orchard meadows. Conserv Biol 17: 1036–1044
Stohlgren TJ, Chong GW, Kalkhan MA, Schell LD (1997)
Multiscale sampling of plant diversity: effects of minimum
mapping unit size. Ecol Appl 7:1064–1074
Sweatman H, Robertson DR (1994) Grazing halos and preda-
tion on juvenile Caribbean surgeonfishes. Mar Ecol Prog
Ser 111:1–6
Tolimieri N (1995) Effects of microhabitat characteristics on
the settlement and recruitment of a coral reef fish at two
spatial scales. Oecologia 102:52– 63
Tulevech SM, Recksiek CW (1994) Acoustic tracking of adult
white grunt, Haemulon plumieri, in Puerto Rico and
Florida. Fish Res 19:301–319
Turner MG, O’Neill RV, Gardner RH, Milne BT (1989) Effects
of changing spatial scale on the analysis of landscape pat-
tern. Landscape Ecol 3:153–162
Tuya F, Vanderklift MA, Hyndes GA, Wernberg T, Thomsen
MS, Hanson C (2010) Proximity to rocky reefs alters the
balance between positive and negative effects on seagrass
fauna. Mar Ecol Prog Ser 405:175–186
Valentine JF, Heck KL Jr, Blackmon D, Goecker ME and oth-
ers (2007) Food web interactions along seagrass– coral reef
boundaries: effects of piscivore reductions on cross-habi-
tat energy exchange. Mar Ecol Prog Ser 333:37–50
Vanderklift MA, How J, Wernberg T, MacArthur LD, Heck
KL Jr, Valentine JF (2007) Proximity to reef influences
density of small predatory fishes, while type of seagrass
influences intensity of their predation on crabs. Mar Ecol
Prog Ser 340:235–243
Verweij MC, Nagelkerken I, Hans I, Ruseler SM, Mason PRD
(2008) Seagrass nurseries contribute to coral reef fish pop-
ulations. Limnol Oceanogr 53:1540–1547
Ward TJ, Vanderklift MA, Nicholls AO, Kenchington RA
(1999) Selecting marine reserves using habitats and spe-
cies assemblages as surrogates for biological diversity.
Ecol Appl 9:691–698
Watson M, Munro JL, Gell FR (2002) Settlement, movement
and early juvenile mortality of the yellowtail snapper
Ocyurus chrysurus. Mar Ecol Prog Ser 237:247 –256
Wernberg T, Vanderklift MA, How J, Lavery PS (2006) Export
of macroalgae from reefs adjacent to seagrass beds.
Oecologia 147:692– 701
Wiens JA (1989) Spatial scaling in ecology. Funct Ecol 3:
385– 397
274
Submitted: May 29, 2010; Accepted: November 19, 2010 Proofs received from author(s): March 3, 2011
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤
➤➤
➤
