This modeling approach combined land-use and land-cover configurations with wildlife population dynamics. First, I chose a region that has predicted urbanization until 2060. I then chose three species that urbanization in that region is likely to affect. I produced eight scenarios of open space and urban development that present distinct landscape patterns (Penteado 2013) using computer software Envision to produce 20 rule-based replicates of each scenario, resulting in varying habitat quantities. A land-use land cover map for each scenario mean replicate was converted to habitat suitability maps for each of the three species (Schumaker et al. 2004, Baker and Landers, 2004, Hulse et al., 2002). I used those suitability maps and species’ life history parameters with individual-based model HexSim to develop dispersal models and evaluate the effects of the various landscape arrangements on individual dispersal and resulting populations. The following sections describe these steps.
The goal was to produce simulations that were complex enough to capture the influence of landscape patterns on the ability of animals to move across the landscape to establish territories and breeding habitats, but simple enough to be incorporated in open space planning.
Study area
I applied this framework to two areas designated for future urban expansion adjacent to Damascus, OR, in the south-eastern portion of the Portland metropolitan region (Fig. 1a and b). Their areas sum 1879 ha. An 800 m buffer surrounding those areas was added to provide connections among them and to adjacent habitats (Fig. 1c). The total area used in the simulations sums to 4592 ha. The study area presents a highly fragmented landscape (ca. 2010). There was a significant alteration of pre-settlement habitats once composed of forests, woodlands, and savannas. Low density rural residential areas and agricultural fields prevail among patches of forest, oak savanna, and grassland; a few small, isolated wetlands occur in the area; riparian vegetation is at various degrees of degradation along the North Fork Deep Creek (Fig. 1c).
Wildlife species
This study targets three wildlife species that require various habitat types that may be affected by urbanization. The Red-legged frog (Rana aurora aurora) is associated with wetlands for breeding and moist forests for seasonal migration. Red-legged frogs disperse to relatively large areas and require close association with moist forests, stream banks, and wetlands (COSEWIC 2004). They typically breed in vegetated shallows of wetlands, ponds, ditches, springs, marshes, margins of large lakes, and slow-moving portions of rivers. They could also breed in ephemeral ponds, moist house yards and neighborhood parks where building density is low, as well as in small natural or modified catchment areas used for storage of stormwater run-off (Chelgren et al. 2006; COSEWIC 2004; Davidson et al. 2001; Lannoo 2005; O’Neil 2001). Habitat fragmentation is of particular concern in view of the species’ seasonal migrations between forested areas and wetland breeding sites (COSEWIC 2004).
Western meadowlarks (Sturnella neglecta) breed and feed in relatively large expanses of grasslands and prairies, but flocks sometimes feed on corn, wheat, and other grains (Morrison 1993; Oregon Department of Fish and Wildlife 2006). Declines of grassland bird populations result from loss (urbanization), degradation (land management practices, disruption of natural disturbance regimes), and fragmentation (smaller isolated patches) of habitat (Johnson and Igl 2001; Oregon Department of Fish and Wildlife 2006).
Douglas squirrels (Tamiasciurus douglasii) associate with conifer forests ranging from west of the Cascade Mountains to the coast, from southern British Columbia, Washington, Oregon, to northern California. In general, old-growth stands are preferred over young and mature stands, although studies have shown larger abundance in second-growth or mature stands (Ransome and Sullivan 2004). They feed on seeds, fungi, and occasionally bird eggs and nestlings; food supply determines population fluctuations (Gonzales et al. 2008; Sullivan and Sullivan 1982). Douglas squirrel is highly territorial and solitary, except during mating. Home range is less than 0.6 ha. Migration may occur if food supply diminishes (O’Neil 2001).
By selecting a suite of target species, planning measures to support them may also influence viability of other species with similar requirements (Rubino and Hess 2003). For example, the Red-legged frog may share habitats with Northwestern salamanders, Long-toed salamanders, Pacific chorus frog, and Rough-skinned newts (Lannoo 2005). The Western meadowlark may coexist with other grassland birds such as Western bluebird, Oregon vesper sparrow, Horned lark, Grasshopper sparrow, and Common nighthawk (Oregon Department of Fish and Wildlife 2006). Douglas squirrels share habitats with other tree squirrels such as the Northern flying squirrel and the Townsend chipmunk, and may indicate the presence of their predators such as the Northern spotted owl, goshawk, and weasel (Duncan 2004).
Alternative future scenarios
Future scenarios depart from a ca. 2010 representation of the study area’s existing conditions. Eight future scenarios for the year 2060 combine four open space patterns (no specific open space pattern, corridors, patches, and network) and two urban development patterns (compact and dispersed) (Penteado 2013). Planning rules using principles of landscape ecology for corridors, patches and networks, and compact and dispersed urbanization patterns determined the landscape arrangement present in the eight scenarios.
All scenarios assume at least a set of minimum habitat conservation strategies: areas within a 60 m-wide buffer around streams, mature and old growth forests, wetlands, grasslands and oak savannas were protected from development. In those areas, modeling incorporated automatic processes to simulate vegetation succession. The different open space patterns used in scenarios contrast and test landscape patterns intended to support species movements via 1) increased corridors to connect habitat patches; or 2) increased patch size and distribution both to increase total habitat area and to serve as stepping stones for movement; or 3) a combination of increased habitat patch sizes and area with corridor connections; or 4) neither increased patches or corridors.
Greenway scenarios emphasize corridors and strategies for protecting and restoring riparian vegetation. Streams create a framework for promoting an armature of open space that provide habitat and connectivity. Park System scenarios adopt parks as a means to create larger habitat patches and stepping-stones. These scenarios test the ability of the chosen species to move through a fragmented landscape where there are fewer connecting habitat corridors. Network scenarios combine habitat patches, stepping-stones and corridors to protect and connect habitats for the chosen species and consequently protect biodiversity (Opdam et al. 2006).
Compact development scenarios depict urbanization strategies for built land uses that concentrate development around existing transportation corridors, in areas of lower ecological impact. Urban development in these scenarios has higher proportions of high-density residential and mixed uses (residential and employment) to minimize loss of open space and maximize ecological function to the year 2060. Dispersed development scenarios reproduce existing trends in urban development (large-parcel, single-family, lower densities), which occur, in the simulations, in any developable areas except those where habitat conservation is a priority.
The combinations of open space and development produce, then, the following scenarios: the first two, Compact Development (CD); Dispersed Development (DD), had no open space concept adopted; Greenway and Compact Development (GCD) and Greenway and Dispersed Development (GDD) adopted corridors; Park System and Compact Development (PCD) and Park System and Dispersed Development (PDD) focused on the distribution of habitat patches; and Network and Compact Development (NCD) and Network and Dispersed Development (NDD) combined corridors and patches.
The software Envision simulates a predefined human population growth and vegetation succession (Penteado 2013). Assumptions about open space and urban development patterns were translated into policies, then into rules that drove scenario simulations (Table 1). Combinations of policies produced the eight scenarios.
Table 1 Examples of policies that translate spatial concepts to rules used in the simulations Dispersal model
I used computer software HexSim (version 2.5) to assess wildlife population viability from a dispersal perspective, which assumes organisms are in search of suitable territories to meet their life history needs. My aim was to build simple but scientifically defensible models that evaluate population viability in the endpoint landscapes (2060) of each scenario for the three chosen species.
HexSim is a spatially-explicit, individual-based computer model designed for simulating terrestrial wildlife population dynamics and interactions (Schumaker 2011, 2018). This model combines spatial landscape data with organism response to various land cover types to examine population viability (Stronen et al. 2012). HexSim couples species’ habitat needs to their survival, reproduction and movement rates. HexSim evaluates the effects that spatial patterns may have on wildlife populations by testing the ability of individuals to disperse in the landscape.
HexSim uses species-habitat associations, area requirements, estimates of demographic parameters and movement characteristics, survival, reproduction, and movement information (Schumaker et al. 2004) (Table 2). Species population viability in HexSim is strongly based on the ability of individuals to move through the landscape for both foraging/feeding and for dispersal to breeding locations. HexSim produced spatial data (HexMaps) and simulation results expressed in census tables (measures of population size through time) that contain population size data by replicate and time step. For Schumaker and Brookes (2018), the model adds more biological nuance to connectivity assessments.
Table 2 Species parameters used in the simulations. Reproduction considers individuals that survive the 1st year (Red-legged frog: 5% survive to metamorphosis; Western meadowlark: 50% fledge; and Douglas squirrel: 25% survive first year) to improve processing time. Report logging period starts after populations reach steady state For the dispersal model representations of the species in the modeled alternative future landscapes, I adopted a landscape classification composed of four elements: breeding habitats, movements and foraging habitats, agricultural matrix, and urban matrix. The representation attempted to echo both species life histories and land mosaics components – patch, corridor, and matrix – in a form sufficiently simplified to enhance its applicability within the time and resource constraints of a typical metropolitan open space planning process.
Landscape representations of scenarios in a geographic information system contained habitat scores, ranging from zero to ten, that reflect habitat quality for each species (Baker and Landers 2004; Schumaker et al. 2004). I adopted those scores to produce suitability maps for each species (Online supplementary material). Hence, each scenario generated three suitability maps, one for each species that I then converted into bitmap representations (see supplementary online material for suitability maps for ca. 2010 and all scenarios). These maps originated hexagonal representations (HexMap) that HexSim uses to simulate life-cycle events. Each hexagon is 30 m wide. The hexagonal grid facilitates movements to adjacent hexagons in multiple directions. HexMaps contained a simplified representation of the landscape; four land cover categories represented the landscape: breeding habitats, suitable non-breeding habitats (for foraging and dispersal), urban matrix (which includes all roads), and rural matrix. Urban matrix hexagons received higher mortality rates to impose a higher stress on moving individuals.
Twenty HexSim simulation replicates for ca. 2010 and for each of the eight 2060 combinations of open space and urban development patterns were conducted for 50 (Red-legged frog), 100 (Douglas squirrel) and 200-year (Western meadowlark) simulation periods (time steps in the model). During modelling calibration, it was necessary to use different time frames for each species to allow populations to reach a steady state. Simulations started with populations in breeding sites. I used different numbers of individuals for each species. Because there was a small amount of wetlands in the area, I used a starting population of 300 Red-legged frogs to make sure most wetlands were populated. I used the same strategy for the Western meadowlark but with a larger initial population (1000 individuals). Douglas squirrel habitats were abundant in the ca. 2010 landscape. Its initial population was smaller (100) in order to observe their ability to move across the landscape and colonize habitats in the ca. 2060 future scenario landscapes.
Evaluation
I measured population viability by looking at population sizes that resulted from the capacity of the landscape to facilitate or impede species dispersal. I then explored wildlife habitat effects of urban open spaces in the 2060 scenarios, by contrasting them with the same qualities in the ca. 2010 landscape. I tracked two categories of population, breeding individuals and floaters (individuals that disperse in the landscape in search of breeding habitats), and used population size mean estimates across the multiple replicate simulations to compare across scenarios (Carroll et al. 2003; McRae et al. 2008; Stronen et al. 2012). Increases and/or decreases of breeding populations indicate the ability of those landscapes to sustain populations of the chosen species as a function of habitat arrangement and can be compared across scenarios. Comparing resulting populations (census) for each species for each scenario shows which spatial concepts were more effective in providing conditions for dispersal. By looking at breeders and floaters, I could also look at the influence of different types of habitats – habitats that are used for breeding and habitats that are used for movements respectively. I used a two-way ANOVA to test the interaction between open space and urban development patterns and a Tukey test to perform multiple comparisons of means with a 95% family-wise confidence level. Both tests used statistical software R version 2.14.1 (The R Foundation for Statistical Computing 2011).