640
Susan S. Bell, Mark S. Fonseca, and Nathaniel B. Stafford
conduct landscape analyses. For example, software
programs will readily perform spatial interpolation
of vegetation cover from even a few, widely dispersed data points, but the interpretation of the output cover surface must take into account that the information is very likely highly inaccurate (especially
for patchy cover types). Kriging and other forms of
geostatistical analysis provide reportable error estimates by taking spatial autocorrelation of data into
account, making their application highly desirable
for spatial analyses. This becomes especially important when spatial representations are being compared
and/or become the basis for management decisions.
Lastly, given the many issues related to scale as
noted in the literature, how can and should such considerations be applied to seagrass landscape ecology? For example, a recent paper by Balestri et al.
(2003) reported that both large (10 km) scale and
very small (1 m) scale variation in morphology and
growth of Posidonia oceanica was recorded, illustrating that some environmental factors may have
impacts over multiple scales. There are few seagrass
studies that specifically utilize geostatistics (Fonseca, 1996) or that directly assess scale issues (Bell
et al., 1995; Fonseca et al., 2002). Otherwise, the
problem of non-independence between seemingly
independent samples due to spatially coherent variability (spatial autocorrelation) has been most directly addressed in mensurative ecological studies,
such as those using GIS. As a general rule, some
knowledge of the spatial scale of environmental variation in abiotic and biotic factors (e.g. sediment
composition, plant density, and composition, currents, temperature, etc.) within the landscape extent
at a given sampling resolution must be established,
and adjustment of sampling effort made accordingly. Moreover, relying on randomization or lattice
designs to account for systematic trends in variation also stands little chance of detecting anisotropic
(direction-dependent) sources of measurement error
(Fonseca, 1996). Thus, orientation of the experimental layout may also carry embedded sources of bias,
which need to be considered in the design and analysis of landscape level studies.
VIII. Future Directions
A number of topics emerge from our review above as
logical topics worthy of future investigation. From a
plant perspective, the study of seagrass population
biology in a spatial context remains a critical need.
Information on how growth patterns of rhizomes and
patterns of seed recruitment contribute to the spatial heterogeneity of seagrass landscapes is sorely
lacking. Data from restoration studies or recovery of
propeller-scarred seagrass areas may provide some
unique insight into these topics.
From a landscape perspective, an explanation of
the mechanisms of patch origin and the processes
that control their shape and arrangement would be
a major advancement. Within seagrass landscapes,
information on both biotic and abiotic factors that affect the maintenance of boundaries would improve
our understanding of mechanisms responsible for
changes in landscape structure over time and space.
For example, are the edges of patches highly influenced by bioturbation and is this consistent over
the landscape? Are patch boundaries resilient to extreme storm events and does resiliency vary among
patches of different sizes of different species? How
much change occurs at edges vs. interior gap areas
of seagrass patches? Are changes in patch boundaries accompanied by variation in faunal activity or
movements? These questions, building upon results
from previous investigations, incorporate a new focus on edges of patches and are logical extensions
to studies of patch dynamics (see also Duarte et al.,
Chapter 11).
Faunal relationships with seagrass landscape features are often inferred from sampling regimes
that provide limited data on behavior of organisms
within patches of different characteristics, due in
part to a fluid medium that imposes limits on visibility. Similarly, quantitative information on movement patterns of fauna throughout landscapes is
also sorely lacking. Much of the excellent data on
fauna–landscape associations in terrestrial systems
has emerged from visual observations of animal use
of habitat (e.g. nesting sites) and by tracking of
movements. Techniques to acquire such information in seagrass landscapes are a necessity if we
are to improve our interpretive power concerning
faunal–landscape relationships. Video camera systems that capture detailed records of animal activities may serve as one potential source of behavioral
information to evaluate landscape percolation and
connectivity. Acoustic telemetry may also be a useful tool for tracking animal movement in seagrass
landscapes.
Recently, Wu and Hobbs (2002) argued that
while much attention in landscape ecology has been
Susan S. Bell, Mark S. Fonseca, and Nathaniel B. Stafford
conduct landscape analyses. For example, software
programs will readily perform spatial interpolation
of vegetation cover from even a few, widely dispersed data points, but the interpretation of the output cover surface must take into account that the information is very likely highly inaccurate (especially
for patchy cover types). Kriging and other forms of
geostatistical analysis provide reportable error estimates by taking spatial autocorrelation of data into
account, making their application highly desirable
for spatial analyses. This becomes especially important when spatial representations are being compared
and/or become the basis for management decisions.
Lastly, given the many issues related to scale as
noted in the literature, how can and should such considerations be applied to seagrass landscape ecology? For example, a recent paper by Balestri et al.
(2003) reported that both large (10 km) scale and
very small (1 m) scale variation in morphology and
growth of Posidonia oceanica was recorded, illustrating that some environmental factors may have
impacts over multiple scales. There are few seagrass
studies that specifically utilize geostatistics (Fonseca, 1996) or that directly assess scale issues (Bell
et al., 1995; Fonseca et al., 2002). Otherwise, the
problem of non-independence between seemingly
independent samples due to spatially coherent variability (spatial autocorrelation) has been most directly addressed in mensurative ecological studies,
such as those using GIS. As a general rule, some
knowledge of the spatial scale of environmental variation in abiotic and biotic factors (e.g. sediment
composition, plant density, and composition, currents, temperature, etc.) within the landscape extent
at a given sampling resolution must be established,
and adjustment of sampling effort made accordingly. Moreover, relying on randomization or lattice
designs to account for systematic trends in variation also stands little chance of detecting anisotropic
(direction-dependent) sources of measurement error
(Fonseca, 1996). Thus, orientation of the experimental layout may also carry embedded sources of bias,
which need to be considered in the design and analysis of landscape level studies.
VIII. Future Directions
A number of topics emerge from our review above as
logical topics worthy of future investigation. From a
plant perspective, the study of seagrass population
biology in a spatial context remains a critical need.
Information on how growth patterns of rhizomes and
patterns of seed recruitment contribute to the spatial heterogeneity of seagrass landscapes is sorely
lacking. Data from restoration studies or recovery of
propeller-scarred seagrass areas may provide some
unique insight into these topics.
From a landscape perspective, an explanation of
the mechanisms of patch origin and the processes
that control their shape and arrangement would be
a major advancement. Within seagrass landscapes,
information on both biotic and abiotic factors that affect the maintenance of boundaries would improve
our understanding of mechanisms responsible for
changes in landscape structure over time and space.
For example, are the edges of patches highly influenced by bioturbation and is this consistent over
the landscape? Are patch boundaries resilient to extreme storm events and does resiliency vary among
patches of different sizes of different species? How
much change occurs at edges vs. interior gap areas
of seagrass patches? Are changes in patch boundaries accompanied by variation in faunal activity or
movements? These questions, building upon results
from previous investigations, incorporate a new focus on edges of patches and are logical extensions
to studies of patch dynamics (see also Duarte et al.,
Chapter 11).
Faunal relationships with seagrass landscape features are often inferred from sampling regimes
that provide limited data on behavior of organisms
within patches of different characteristics, due in
part to a fluid medium that imposes limits on visibility. Similarly, quantitative information on movement patterns of fauna throughout landscapes is
also sorely lacking. Much of the excellent data on
fauna–landscape associations in terrestrial systems
has emerged from visual observations of animal use
of habitat (e.g. nesting sites) and by tracking of
movements. Techniques to acquire such information in seagrass landscapes are a necessity if we
are to improve our interpretive power concerning
faunal–landscape relationships. Video camera systems that capture detailed records of animal activities may serve as one potential source of behavioral
information to evaluate landscape percolation and
connectivity. Acoustic telemetry may also be a useful tool for tracking animal movement in seagrass
landscapes.
Recently, Wu and Hobbs (2002) argued that
while much attention in landscape ecology has been