Chapter 26 A Landscape Perspective
629
at a 1 m scale leaves the reader unclear on whether it
is the grain size or the extent that is represented by 1
m. Besides the basic problem of differences in opinion on appropriate scale for an organism of study,
the choice of scale in a study is, by default, autocorrelated with results. This intertwining of scale and
results arises because of the simple premise that units
of X become less similar as a function of distance
or time that separates them. We generally recognize
that a given unit X, when sampled over time, is not
a replicate sample unless we invoke special tools for
dealing with repeated measures; that is, the sample
at time (t + 1) is strongly influenced by the state
of the unit at time (t) [i.e. they are autocorrelated,
(Goodchild, 1987)]. What is often not addressed in
ecological studies is that these same attributes apply to sampling units in space as well as time. For
example, it is probable that two adjacent square metre plots of seagrass are more likely to be similar
to each other than a comparison of two plots separated by 100 m. Therefore, resolution will influence
observed variance between samples.
The notion that two units of seagrass have declining similarity as a function of distance has an
important implication: if two samples are related as
a function of distance apart, then samples taken too
closely together are not independent, which violates
most assumptions used in statistical analyses (Fonseca, 1996). Thus, the choice of spatial grain and
extent may render invalid the apparent results because of spatial autocorrelation of data. Discussions
of tools that can be employed to address such spatial autocorrelation of data are available in Cressie
(1991) and Rossi et al. (1992), and these are useful
general references for spatial analyses of seagrass
landscapes.
III. Historical Overview
Here we survey the literature that has applied concepts of landscape ecology as presented above to
seagrass ecosystems. Excluded from this historical overview are the many valuable studies reporting estimates of seagrass coverage and change, i.e.
chronosequences (e.g. Orth and Moore, 1983; Pulich
and White, 1991; Fletcher and Fletcher, 1995). Although many of these studies were done at or beyond
what might be considered the landscape scale, by
simply tracking changes in the areal extent of seagrass beds they do not specifically apply landscape
theory or metrics to their analyses, and hence these
studies will not be addressed here. The papers that do
apply landscape theory to seagrasses fall largely into
two groups: those that examine (1) factors that modify seagrass landscape structure and persistence, and
(2) relationships between seagrass landscape characteristics and ecosystem components. We first review
the papers dealing with seagrass vegetation as the
primary patch type, and then discuss those that address how the size, shape, and arrangement of seagrass patches affect other ecosystem components.
A. Investigating Seagrass Landscapes
1. Patch Structure and Dynamics
The application of landscape principles to seagrass
systems is well represented by studies attempting to
quantify patch structure, arrangement, and dynamics of vegetation (see Duarte et al., Chapter 11 for
a discussion of the clonal nature of seagrass beds
to patch structure and dynamics). One of the first
studies to take a landscape approach to seagrasses
was that of Olesen and Sand-Jensen (1994), which
examined the patch dynamics of Zostera marina in
a semi-enclosed embayment in northern Denmark.
Over 2.5 years a fringe of eelgrass patches bordering a continuous bed was studied (Olesen and
Sand-Jensen, 1994). The authors reported that patch
size was skewed toward smaller patches (50% were
<2 m), but small, young patches were susceptible to
a high mortality rate; lateral expansion rates of patch
edges were not affected by patch size (Olesen and
Sand-Jensen, 1994). In a similar study with somewhat contradictory results, Vidondo et al. (1997)
found lateral expansion rates increased with size of
Cymodocea nodosa patches in the Mediterranean.
They noted the same predominance of small patch
sizes and high rate of patch mortality as was reported
for Z. marina in Denmark, but also found that rate of
patch recruitment exceeded the high mortality rate
(Vidondo et al., 1997). These studies advanced the
concept that seagrass landscapes represent a level of
organization and interaction beyond the spatial scale
of individual clonal units.
Studies investigating temporal dynamics of seagrass patches have considered landscape level questions. A study of Zostera novazelandica patches by
Ramage and Schiel (1999) in New Zealand tracked
patch dynamics at the landscape scale and found
629
at a 1 m scale leaves the reader unclear on whether it
is the grain size or the extent that is represented by 1
m. Besides the basic problem of differences in opinion on appropriate scale for an organism of study,
the choice of scale in a study is, by default, autocorrelated with results. This intertwining of scale and
results arises because of the simple premise that units
of X become less similar as a function of distance
or time that separates them. We generally recognize
that a given unit X, when sampled over time, is not
a replicate sample unless we invoke special tools for
dealing with repeated measures; that is, the sample
at time (t + 1) is strongly influenced by the state
of the unit at time (t) [i.e. they are autocorrelated,
(Goodchild, 1987)]. What is often not addressed in
ecological studies is that these same attributes apply to sampling units in space as well as time. For
example, it is probable that two adjacent square metre plots of seagrass are more likely to be similar
to each other than a comparison of two plots separated by 100 m. Therefore, resolution will influence
observed variance between samples.
The notion that two units of seagrass have declining similarity as a function of distance has an
important implication: if two samples are related as
a function of distance apart, then samples taken too
closely together are not independent, which violates
most assumptions used in statistical analyses (Fonseca, 1996). Thus, the choice of spatial grain and
extent may render invalid the apparent results because of spatial autocorrelation of data. Discussions
of tools that can be employed to address such spatial autocorrelation of data are available in Cressie
(1991) and Rossi et al. (1992), and these are useful
general references for spatial analyses of seagrass
landscapes.
III. Historical Overview
Here we survey the literature that has applied concepts of landscape ecology as presented above to
seagrass ecosystems. Excluded from this historical overview are the many valuable studies reporting estimates of seagrass coverage and change, i.e.
chronosequences (e.g. Orth and Moore, 1983; Pulich
and White, 1991; Fletcher and Fletcher, 1995). Although many of these studies were done at or beyond
what might be considered the landscape scale, by
simply tracking changes in the areal extent of seagrass beds they do not specifically apply landscape
theory or metrics to their analyses, and hence these
studies will not be addressed here. The papers that do
apply landscape theory to seagrasses fall largely into
two groups: those that examine (1) factors that modify seagrass landscape structure and persistence, and
(2) relationships between seagrass landscape characteristics and ecosystem components. We first review
the papers dealing with seagrass vegetation as the
primary patch type, and then discuss those that address how the size, shape, and arrangement of seagrass patches affect other ecosystem components.
A. Investigating Seagrass Landscapes
1. Patch Structure and Dynamics
The application of landscape principles to seagrass
systems is well represented by studies attempting to
quantify patch structure, arrangement, and dynamics of vegetation (see Duarte et al., Chapter 11 for
a discussion of the clonal nature of seagrass beds
to patch structure and dynamics). One of the first
studies to take a landscape approach to seagrasses
was that of Olesen and Sand-Jensen (1994), which
examined the patch dynamics of Zostera marina in
a semi-enclosed embayment in northern Denmark.
Over 2.5 years a fringe of eelgrass patches bordering a continuous bed was studied (Olesen and
Sand-Jensen, 1994). The authors reported that patch
size was skewed toward smaller patches (50% were
<2 m), but small, young patches were susceptible to
a high mortality rate; lateral expansion rates of patch
edges were not affected by patch size (Olesen and
Sand-Jensen, 1994). In a similar study with somewhat contradictory results, Vidondo et al. (1997)
found lateral expansion rates increased with size of
Cymodocea nodosa patches in the Mediterranean.
They noted the same predominance of small patch
sizes and high rate of patch mortality as was reported
for Z. marina in Denmark, but also found that rate of
patch recruitment exceeded the high mortality rate
(Vidondo et al., 1997). These studies advanced the
concept that seagrass landscapes represent a level of
organization and interaction beyond the spatial scale
of individual clonal units.
Studies investigating temporal dynamics of seagrass patches have considered landscape level questions. A study of Zostera novazelandica patches by
Ramage and Schiel (1999) in New Zealand tracked
patch dynamics at the landscape scale and found