Modeling the impacts of habitat fragmentation in complex landscapes

1
Modeling the impacts of habitat fragmentation in
complex landscapes Leslie Ries1, Emma Goldberg1
Thomas D. Sisk2 1Department of Biology,
University of Maryland, College Park, MD
20742 2Center for Environmental Sciences and
Education, Northern Arizona University,
Flagstaff, AZ 86011
COULD MALCOLMS MODEL EVER BE USED ON REAL
LANDSCAPES? It is clearly intractable to manually
develop solutions to the model for each unique
shape on a landscape. To make the use of this
model practical, we developed R-scripts that
allow the model to be applied to patches of any
shape. Results for the golden-cheeked warbler
(GCWA) at Ft. Hood in Texas are shown below.
OVERVIEW Habitat fragmentation and local land-use
decisions are two of the major factors driving
habitat quality for threatened and endangered
species. However, much of the theory and
empirical evidence for how landscape structure
impacts habitat quality focuses on highly
simplified landscapes. These landscape
configurations have practical value in trying to
tease apart ecological dynamics within complex
systems, but fail when trying to apply the
principles in real landscapes. This phenomenon
is exemplified by the edge literature, which
tends to focus on ideal edge geometry. Most
focal edges are straight and contain only two
adjacent habitat types. Distance to the closest
edge is usually the key metric applied. In real
landscapes, patches have convoluted, complex
shapes where many different habitat types
converge, and the distance to the nearest edge
may not be the best measure of edge influence. A
model published almost 15 years ago (Malcolm
1994) offers a solution, but has never been used
because of the difficulty of applying the math to
real landscapes. We developed a series of
R-scripts to apply this model to any patch shape.
We show that including more complex geometry
results in subtle, but improved, differences in
predictions over traditional models. We discuss
the utility of this model in multiple systems and
our plans to incorporate a discrete approximation
into our Effective Area Model, which is a
user-friendly tool that is implemented within
ArcGIS.
DOES CONSIDERING COMPLEX GEOMETRY IMPROVE SPATIAL
PREDICTIONS?
For years, researchers have used distance to
closest edge as the primary measure for edge
effects, ignoring issues of complex geometry and
multiple, converging habitat types. Indeed, our
spatial model, the Effective Area Model,
currently uses distance to closest edge because,
at this time, there is simply too little data to
justify including more complicated measures.
This test represents the first field test of
Malcolms model in a landscape with complex
geometry. To test Malcolms model, we selected
several survey locations with complex geometry in
GCWA habitat on Ft. Hood. However, each location
contained only a single edge type within 300 m of
the survey point (reflecting the Dmax estimated
by the model). We then compared observed GCWA
detections to three models those predicted by
Malcolms model, those predicted by the distance
to nearest edge, and a null model where edge
responses are ignored.
Golden-cheek warblers in oak-juniper forest
APPLYING MALCOLMS MODEL RESULTED IN SOME
IMPROVEMENT FOR GCWAs
Incorporating edge responses are critical to
understanding GCWA distributions
These birds are endemic to central Texas and are
federally listed as endangered. The largest
population is on Ft. Hood. These birds are
highly habitat specific and, using data from Ft.
Hoods extensive network of point surveys, we
found that they avoid most other habitat types.
We used surveys near the straightest edges to
parameterize Malcolms equation for ideal edges.
The results suggest that natural variation in
ecological systems tend to largely trump the fine
differences between predictions between the
nearest edge and Malcolms model. However, there
was some improvement. The graph below shows that
as edges go from being more concave to more
convex, bird detections tend to increase as
predicted (with a notable outlier).
GCWAs near scrub edges
c) habitat configuration in a portion of Ft.
Benning
b) habitat configuration in a portion of Ft. Hood
a) an ideal edge
Mean number of detections
Parameter summary Dmax 300m k 1.1 e0
-0.002006
This figures shows the predictions of the null
(dotted line) and nearest edge (solid line)
models, which are easily displayed relative to
edge distance. The predictions of Malcolms
model are shown separately for each survey point
(yellow circles). The first pattern to note is
that the predictions are not substantially
different from the nearest edge model. The
observed detection rate (yellow xs) are clearly
much better predicted by a model that includes
edge responses (compared to the null model). But
does Malcolms model improve predictions even
more?
Distance to closest edge (m)
concave
convex
We then used an optimization program to
parameterize the same model, only using the
actual edge geometry (which was not perfectly
straight or ideal). The parameter values
changed only slightly (Dmax 314m, k 1.1, e0
-0.001983). We also measured responses along
woodland, meadow and large road edges. All
showed significant responses in at least some
years.

• CONCLUSIONS AND FUTURE DIRECTIONS
• In addition to the results we present above, we
  also applied a similar test of Malcolms model to
  black-capped vireos on Ft. Hood and grassland
  birds in Iowa prairies. We found similar
  patterns between all three.
• Including complex geometry modified predictions
  only slightly compared to distance to nearest
  edge measures.
• Observed values were often closer to Malcolms
  predictions than nearest edge predictions,
  suggesting that predictions are somewhat improved
  by this more complex model.
• In many ways these results are encouraging they
  suggest that traditional approaches to measuring
  edge effects are sufficient to capture the
  dominant shifts in habitat quality as landscapes
  are modified. They also confirm that
  incorporating edge effects can substantially
  improve predictions over null models where edge
  effects are ignored.
• There are still several aspects of complex
  geometry that need to be considered. When edge
  effects are very strong, the difference between
  the nearest edge and Malcolm model increases.
  For some species, therefore, the impacts of
  complex geometry may be less trivial. We also
  have yet to extrapolate these results over an
  entire landscape, so it is unclear how these
  differences will accumulate over space. Finally,
  we have continued to ignore the issue of
  multiple, converging habitat types. This topic
  has received no empirical treatment despite
  decades of edge research and hundreds of
  published studies. We hope to pursue this topic
  in the future to determine its impact.
• Finally, it is intractable to include Malcolms
  exact model in the EAM, but we are working on
  developing the model into a discrete version that
  can more easily be incorporated into a grid-based
  Arc analytical framework. Based on our current
  results, we believe that even an approximation
  would provide value additional information for
  interested researchers and be sufficient to
  capture the dynamics of the original model.

LITERATURE CITED Li, Q, J Chen, B Song, JJ
LaCroix, MK Bresee, and JA Radmacher. 2007.
Areas influenced by multiple edges and their
implications in fragmented landscapes. For Ecol
and Mgt 24299-107 Fernandez C, FJ Acosta, G
Abella, F. Lopez, and M. Diaz. 2002. Complex
edge effect fields as additive processes in
patches of ecological systems. Ecol. Model.
149273-83. Fletcher RJ, L Ries, J Battin and AD
Chalfoun 2007. The role of habitat area and edge
in fragmented landscapes definitively distinct
or inevitably intertwined? Can. J. Zool.
851017-1030. Haddad, NM and KA Baum 1999. An
experimental test of corridor effects on
butterfly densities. Ecol. Appl.
9623-633. Malcolm, JR. 1994. Edge effects in
central Amazonian forest fragments. Ecology
752438-45. Ries, L, RJ Fletcher, J Battin, and
TD Sisk 2004. Ecological responses to habitat
edges Mechanisms, models and variability
explained. Ann. Rev. Ecol. Evol. Syst.
35491-522.
WHY FOCUS ON HABITAT AND EDGES? 1) Most species
show distinct habitat preferences that reflect
the quality of resources available in each
habitat type. Edge effects tend to modify the
quality of the habitat in predictable ways
allowing a finer estimation of how shifting
landscapes may modify the population size that
could be supported by a particular habitat
patch. 2) Edge effects are widespread. 32 of all
empirical studies show significant edge responses
(Ries et al. 2004) In addition, edge effects
underlie most area responses (Fletcher et al.
2007) and are linked with isolation effects
(Haddad and Baum 1999). 3) We have developing a
series of models that allow edge responses to be
predicted or measured, then extrapolated over
entire landscapes. These tools are often
applicable even when minimal data are available.
Models that focus on isolation often require data
on movement and mortality through different
matrix types data that are rarely available.
Acknowledgments Many people from Ft. Hoods
Department of Natural Resources, including John
Cornelius, Charles Pekins, Rich Koesteke (TNC)
and Dave Cimprich (TNC), helped quickly oriented
us to the management issues of the base and
shared their data. Spatial data were obtained
from Marion Noble (TNC). Timothy Bricker is
currently reprogramming the EAM and Andra Doherty
developed the maps used for this project.
Funding was supplied by SERDP (Project CS-1100
and SI-1597).
Contact Information Email lries_at_umd.edu OR
Thomas.Sisk_at_nau.edu