Wetland connectivity: understanding the dispersal of organisms that occur in Victoria’s wetlands draft
|
|  Yüklə 2.04 Mb.
|
Assessing landscape connectivityThe dispersal characteristics of wetland biota are varied and will result in diverse temporal and spatial patterns of connectivity in the landscape. Based on an understanding of the biological processes that connect wetlands, models of wetland connectivity can be generated to help guide the management of wetlands at a landscape scale. Attempts to assess connectivity have so far focused on terrestrial systems; models of wetland connectivity have not yet been attempted. Assessing wetland connectivity requires a multispecies approach to represent the diverse biota of wetland systems. Such an approach would require that patterns of connectivity be developed for key groups that share similar modes of dispersal. These models could then be integrated to provide a system-based connectivity model. Connectivity at a landscape-scale can be assessed using a range of modelling approaches. These vary in complexity from those that consider only the structural characteristics of the landscape to those that consider how the dispersal biology of the focal organism(s) may be modified by elements of the landscape. In this chapter, modelling approaches to assess connectivity at a landscape-scale are appraised to identify a suitable approach for developing connectivity maps for wetland biota across Victoria.
Structural connectivity infers dispersal from the geographical arrangement of habitats in the landscape (Calabrese and Fagan 2004). Metrics related to distance or the patterning of habitats is typically used. Distance metrics — Two types of distance metrics are used: (1) nearest-neighbour distance, and (2) neighbourhood metric. Nearest-neighbour distance is the more commonly used distance metric. It uses the shortest straight line (Euclidean) distance between a focal habitat patch and its nearest neighbour to infer dispersal, assuming that the smaller the inter-patch distance the greater the likelihood of dispersal. The potential for all other patches in the landscape to contribute dispersers is ignored (Moilanen and Nieminen 2002). In contrast, the neighbourhood metric does consider the distances between a focal habitat patch and all patches. As the likelihood of dispersal declines exponentially with increasing distance, a negative exponential function is used to score connectivity from these dispersal metrics. Spatial pattern metrics — A host of spatial pattern metrics have been developed to characterise various landscape elements that influence connectivity, including patch quality, patch number, patch area, core area, patch perimeter, contagion, perimeter–area ratio, shape index, fractal dimension, patch cohesion, extent, shape, and spatial arrangement of landscape elements (Calabrese and Fagan 2004, Schuemaker 1996). Measures of structural connectivity are appealing as they provide a rapid assessment of connectivity for large areas. However, they should be applied cautiously as species dispersal abilities are not considered and patterns of connectivity may fail to represent real patterns in nature (Jacobson and Peres-Neto 2010). Even where a relationship between spatial metrics and dispersal are established the relationship may not hold in different landscapes or for different species (Calabrese and Fagan 2004).
Potential connectivity is an indirect measure of connectivity that combines spatial information of landscape structure and the dispersal biology of the focal organism(s). The ecological realism of models is enhanced by incorporating detailed information on species-specific habitat requirements and dispersal behaviours, including:
The simplest models assess connectivity according to whether the distances between habitats patches are within the dispersal ability of the species of interest. More complex models map dispersal routes based on the permeability of the landscape, as perceived by the species. Graph theory, least cost analysis and circuit theory are three key modelling techniques used to assess potential connectivity at a large spatial scale. Two other techniques — buffer radius and incidence function — are best applied to detailed studies of connectivity at smaller spatial scales (Moilanen and Nieminen 2002) and are not discussed here.
Graph models calculate all possible pair-wise connections among habitat patches using information on dispersal distance and indices of landscape structure (Keitt et al. 1997, Bunn et al. 2000, Calabrese and Fagan 2004). In graph models the habitat patches are represented as nodes (also called vertices), and dispersal between habitat patches are represented as links (Urban and Keitt 2001, Urban et al. 2009). Habitats (nodes) are often weighted by area and sometimes by quality, which serve as proxies for population size and provide some indication of the strength of dispersal (Calabrese and Fagan 2004). In graph models, links are the shortest straight line distance between habitat nodes. Habitat nodes are connected by links if they are in the assigned dispersal distance (Keitt et al. 1997). However, species change their route in response to the traversability of the terrain and graph models do not account for this. Graph models therefore provide only a coarse representation of dispersal. Graph models are attractive because they provide a visual representation of the overall structure of habitat connections for large areas that is not possible with more complex models. They also offer the utility of assigning directionality to dispersal pathways. Basic graph models can provide ecological information about the system (Urban and Keitt 2001), including that indicated below.
Graph-based models can also be used to evaluate the effects of landscape-scale management strategies on connectivity. For example, the importance that different habitat patches have to connectivity can be evaluated by deleting a node and assessing what effect this has on connectivity. Conversely, the benefits of creating or restoring habitat patches in the landscape can be assessed by inserting a node or changing the attributes of a node (e.g. habitat quality or size) and re-assessing connectivity. A review by Calabrese and Fagan (2004) proposed that graph theory represents the best trade-off between the level of information generated from the model and its data requirements. Despite these benefits, graph models do not take into account landscape permeability, and this limits their use to situations where the mobility of the focal taxa is largely unaffected by landscape elements, or where landscapes do not contain features that constrain movement. Graph models also do not identify multiple pathways, although overlaying graph models may enable multiple pathways to be assessed.
Least cost analysis and circuit theory methods are based in graph theory, but ecological realism is increase by considering the ability of focal species to traverse the non-habitat matrix. Least cost analysis identifies dispersal pathways among habitat patches by assigning permeability scores to the intervening landscape. Dispersal distances along a single path of greatest permeability — or least cost — is measured, rather than simply the geographic (straight line) distance. Least cost analysis has been used to inform conservation planning for over a decade (McRae 2006, McRae and Beier 2007, McRae et al. 2008). Circuit theory, also called isolation by resistance (IBR), is similar to least cost analysis in that it identifies pathways through the intervening landscape that offer the lowest resistance to movement between habitat patches. It is analogous to electrical circuits as it represents connectivity as circuit diagrams (Fig 6.1). The main advantage of circuit theory is that it can model multiple dispersal pathways between habitats (McRae and Beier 2007). Circuit theory models can assess connectivity between multiple habitats but are computationally demanding when the number of habitats is large.
Neighbourhood analysis and cost distance analyses are two approaches that can be applied to assess connectivity in a GIS framework (VEAC 2010). These approaches, like circuit theory and least cost analysis, increase realism by incorporating rules around landscape permeability and the direction and scale of movement for the organism(s) of interest. In both neighbourhood analysis and cost distance analysis the landscape is represented as grids, and cells are assigned a permeability score based on the dispersal constraints for the organism of interest, with habitat cells having the highest permeability. Neighbourhood analysis and cost distance analyses represent connectivity in slightly different ways. Cost distance analysis only assesses the permeability of cells surrounding habitats that are in the dispersal range of the organism of interest. This approach provides a detailed representation of connectivity among habitats but provides no information on the permeability of the landscape beyond the specified dispersal distance. In contrast, Neighbourhood analysis assesses the permeability of the landscape surrounding each cell in the landscape, which identifies permeable corridors that are not revealed by least cost analysis. Combining outputs from both analyses is more representative of connectivity than the outputs of the individual analyses. |