ATBC 2008 - ewers

Worskhop

Quantitative measurement of edge effects in fragmented landscapes

Conveners : Robert Ewers (Imperial College London, Silwood Park Campus, UK) and Raphael Didham (School of Biological Sciences University of Canterbury, New Zealand)

Contact : r.ewers -at- imperial.ac.uk; raphael.didham -at- canterbury.ac.nz

Edges, ecotones or ecological boundaries by any other name have had profound effects on the dynamics of species and communities in human-modified landscapes (Yahner 1988; Sisk & Battin 2002; Ries et al. 2004). Boundaries between habitat patches are typically accompanied by a transition in the diversity and structural complexity of plant communities (Fraver 1994; Laurance et al. 2002; Harper et al. 2005), resulting from the complex interplay of direct human disturbance and indirect changes in a wide range of biotic and abiotic processes (Chen, Franklin & Spies 1995; Murcia 1995). Edges also alter species’ interactions (Fagan, Cantrell & Cosner 1999), the trophic structure of communities (Laurance et al. 2002), the movement of individuals through landscapes (Wiens 1992) and resource flows between habitats (Wiens, Crawford & Gosz 1985; Wiens 1992; Huxel & McCann 1998), thereby modifying ecological processes and dynamics at a wide range of spatial and temporal scales (Wiens 1992; Ries et al. 2004). From a conservation viewpoint, edges can be the focal sites for the invasion of exotic species (Fraver 1994) and reduction in population densities of habitat interior specialists (Temple 1986). At the same time, edges are frequently associated with elevated densities of predators (Sisk & Battin 2002) that invade from the surrounding matrix habitat and may concentrate their foraging along habitat boundaries (Askins 1995). Yet despite widespread recognition of the importance of habitat boundaries in ecology and conservation, and the vast literature covering the subject, there have been relatively few attempts to define statistically the scale at which edge effects operate across ecological boundaries (Chen, Franklin & Spies 1992, 1995; Cadenasso, Traynor & Pickett 1997; Laurance et al. 1998; Didham & Lawton 1999; Brand & George 2001; Harper & MacDonald 2001; Toms & Lesperance 2003; Cancino 2005; Ewers & Didham 2006).

Our goals in this workshop are to:

discuss the design of sampling programs that can be used to measure edge effects
explore statistical methods for quantifying the strength of edge effects
consider how patch-based models of edge effects might be incorporated into a GIS to investigate patterns of biodiversity across landscapes Measuring edge effects

Robust quantification of edge response functions depends on the initial design and implementation of edge sampling protocols. Key points of discussion will centre on the need for comparable sampling at multiple distances on both sides of an edge (patch and matrix), whether it is more appropriate to sample at discrete fixed distances versus random continuous distances, and whether linear versus logarithmic scaling is most appropriate. We will highlight that adequate model comparisons (see below) can only be made if appropriate sampling is employed. Extrapolating local-scale edge effects up to landscape-scale models (see below) can only be attempted if potential spatial and/or environmental variability in edge response functions are considered (Ewers et al. 2007).

Edge response functions

A wide variety of mathematical methods has been proposed for quantifying one-dimensional edge response functions, but there are strengths and weaknesses to the use of each method, and there are unresolved problems that remain to be addressed

arbitrary threshold similarity (e.g. Chen, Franklin & Spies 1992; Brand & George 2001; Hylander 2005)
intersecting confidence intervals (e.g. Laurance et al. 1998; Didham & Lawton 1999; Harper & MacDonald 2001)
breakpoint regression (e.g. Toms & Lesperance 2003; Cancino 2005)
continuous response functions (e.g. Mancke & Gavin 2000; Ewers & Didham 2006)

Key points of discussion will centre on discriminating and quantifying edge effect magnitude versus edge extent, and on how to incorporate asymmetrical responses across boundaries into a new generation of models.

Predictive modelling and scaling local edge effects up to landscape models

Edge response functions can be used to extrapolate patch-scale variation in the distribution and abundance of species up to landscape or regional scale models. This can be useful for conservation management or monitoring of biodiversity loss, but there are inherently a large number of caveats to these approaches
core area model (Laurance & Yensen 1991)
effective area model (Sisk et al. 1997)
D-AEI model (Zheng & Chen 2000)
GIS algorithms (Ewers & Didham 2007)

Key points of discussion will centre on the assumptions underlying various approaches, including the qualitative landscape model being used, the mechanisms of species response to landscape change and the context-dependence of local edge effects.

What participants should bring?

Essential: A willingness to discuss openly the advantages and disadvantages of various conceptual and analytical approaches to the quantification of edge effects

Optional: A laptop computer, if possible, and edge responses data that might be suitable for discussing or analysing. Some scripts will be available for use in the R package.

Optional: ArcGIS with a habitat map that can be used to generate landscape maps of edge effects.

References

SPONSORSHIPS

Société Michelin

ATBC 2008 - ewers_didham