Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)

Legacy Department

Environmental Design and Planning


Chanse, Victoria

Committee Member

Baldwin , Robert

Committee Member

Tonkyn , David

Committee Member

Lauria , Mickey


Ecological theories including island biogeography, intermediate disturbance, metapopulation and metacommunity all suggest that habitat patches of larger size and those comprised of substantial configurations of interior or core habitat possess the greatest potential for long-term species viability. As a direct means of mitigating edge encroachment and fragmentation's other adverse effects, there is a growing consensus among conservation planners that assembling larger, more cohesive tracts with substantial core area is of ecological value in conservation planning. Larger and more cohesive patches are believed to sustain larger and more viable local populations, enhance overall biodiversity, incorporate a wider array of natural disturbance regimes, and maintain more vulnerable, specialist species for the long term. Therefore, it is important that size and cohesion metrics be incorporated in patch and reserve modeling and design.
This research developed a spatially explicit patch modeling approach designed to incorporate these metrics. This new modeling tool is entitled the Cohesive-Patch Aggregation and Network (C-PAN) model. It was created using ArcMap 9.3 and the Spatial Modeler extension. The model was first tested at a pilot scale (the State of South Carolina) and then up-scaled to evaluate a much larger area (the Northern Appalachian/Acadian Ecoregion). The C-PAN approach is most appropriate for use on species requiring substantial core area and those sensitive to edge characteristics. It is also intended to serve as an alternative approach to heavily parameterized patch modeling methods when species-specific parameterization data are not available.
There exist a number of potential benefits associated with C-PAN usage. The C-PAN model searches landscapes for highly cohesive patches with substantial core area within an existing GIS framework. The aggregation and overlay processes used by the model also appeared to be an improvement over highly parameterized approaches which utilize region-growing components for generating patches. Additionally, the Landscape Cohesion Index (LCI) that is generated as part of the patch generation process proved beneficial for measuring fragmentation metrics across multiple sites and landscapes. This may be the first patch modeling approach to use landscape cohesion scores as a means of seeding patches based on their core area composition from the onset of the modeling process. The LCI allows users to delineate patches based on the statistical uniqueness of their core composition. This frees the user from selecting potentially unknown parameter settings when using other more complex approaches. Instead, it allows patches to be delineated and ranked based on how cohesive they are within the landscape. Both of these features may prove attractive to users as they ultimately make the tool more readily accessible to less technical practitioners.
The C-PAN model was then used to generate a unique set of patches in the Northern Appalachian/Acadian Ecoregion. C-PAN was then compared to two ArcGIS (v9.3) based commonly used patch generation tools. The tools, Corridor Designer (v1) and FunConn (v1) were used for this analysis because they represent two highly utilized approaches which are most similar to the C-PAN model in both modeling mechanics and process. The patch outputs from the three tools were then compared and evaluated. This analysis was aimed at addressing a void within the literature of comparing the results of multiple patch modeling approaches. This analysis also served as a means of validating the C-PAN approach by comparing patch outputs of the three approaches.
C-PAN performed well when compared to the existing patch modeling tools of Corridor Design and FunConn. For all of the spatial and target capture metrics measured, C-PAN ranked first or second among all approaches. The results indicated that the C-PAN patch modeling approach performed as well, and better, in the patch metrics evaluated here (patch area, edge/area ratios, average nearest neighbor, average Human Footprint (HF) score, Last of the Wild (LOW) capture, and patch commission. At relatively high patch selectiveness, the outputs of C-PAN and Corridor design were the most similar in size and distribution across the ecoregion-scale study area.
Furthermore, of the three patch delineation tools, C-PAN appears to provide users with greater site discrimination capabilities than Corridor Design or FunConn. This resulted in providing users with a more selective set of discrete patches than the FunConn approach. Both C-PAN and Corridor Design were effective in delineating highly homogenous patches. These results indicate that the C-PAN patch modeling approach outperforms Corridor Designer and FunConn when measures of patch cohesion and core area are of importance.
A graph theory based connectivity analysis was then conducted in order to identify and compare linkages between patches from the three patch modeling scenarios. The landscape networks modeled for each of the three scenarios indicated that while local connectivity in portions of the ecoregion may exist, widespread connectivity across the ecoregion as a whole was less likely. This was apparent in the C-PAN and Corridor Design patch scenarios, as multiple connections were delineated across the majority of the study area. Alternatively, no connections were delineated linking portions of the large graphs located within the central portion of the ecoregion with smaller and more linear graphs located in the periphery of the region. This was attributable to natural bottlenecks and relatively high Human Footprint (HF) values in those potential linkage areas. The landscape network derived as part of the FunConn patch scenario indicated even further diminished connectivity within portions of the ecoregion.
The C-PAN patch network scenario was comprised of the greatest number of patches. This ultimately resulted in the delineation of multiple and potentially functional redundancies in the landscape network. Increasing the number of patches also improved distance metrics within the minimum spanning tree for this scenario. More patches served as intermediate stepping stones which resulted in shorter linkage and edge lengths and smaller average area corridor requirements. The FunConn patch landscape network however connected significantly fewer patches. This resulted in the longest linkage and edge distances and the largest average corridors within the ecoregion. This represents an apparent tradeoff between the number of potentially beneficial redundant connections and total landscape network corridor area. While more connections may contribute to increased landscape connectivity and landscape function, the increased area requirement make it more costly to implement. On the other hand, fewer connections may be less costly from an implementation standpoint, but may also reduce landscape connectivity and ecological function.
The landscape networks were then used to test a simplifying assumption often used in conservation planning: that coarse-scale corridors may provide overlapping or 'umbrella' effects for other scenarios. This was accomplished by conducting an analysis of corridor overlap among these three scenarios. This work is among the first corridor gap analyses to be conducted at the ecoregion-scale. The corridor gap analysis indicated that 5% of the corridor area for all 3 scenarios was spatially coincident, 34% was coincident over 2 scenarios, while the majority of corridor area (59%) was non-redundant.
These results are intriguing for two reasons. First, this gap analysis proved to be a useful tool in identifying potential priority conservation areas. Areas held in common may prove to be no-regret areas for conservation action as they provide overlapping coverage across multiple conservation scenarios. Second, the significant coverage gaps among corridors from these three scenarios indicates that selecting 'what' to connect at the ecoregion-scale has significant implications for selected corridors. As there was so little modeled corridor area in common among scenarios, there is little reason to believe alternate corridors would be functionally equivalent. This indicates that connecting any one set of habitat nodes would not likely serve as a corridor umbrella for all other scenarios.
The ecoregion-scale connectivity analysis conducted here was also useful in flagging areas for conservation prioritization based on their connectivity role within an ecoregion-scale context. Connectivity analysis at this scale may also prove useful for evaluating connectivity at local scales. Any one of the subgraphs found within these modeled landscape networks could help inform local scale conservation efforts. Similarly, local scale connectivity and conservation actions could be added to the ecoregion-scale landscape network. As with many things, a successful landscape network is made up of the sum of its locally implemented parts.
Of additional interest, the large size and area requirements of ecoregion-scale corridors may prove to be potential mechanisms by which landscape scale gradients and processes can be included within present day networks of protected lands. While this research did not explore this explicitly, ecoregion-scale corridors may prove to be a provocative means by which natural disturbance regimes, environmental gradients, and shifting species ranges may be captured in conservation networks by virtue of their large size. As such, it may be worth considering ecoregion-scale corridors as implementable conservation components that may facilitate planning for persistence in the face of global climate change.