• Open Access

Predictions of ecological and social impacts of alternative residential development policies to inform decision making in a rural landscape

Authors


  • Editor
    James Aronson

Caren S. Goldberg, Fish and Wildlife Resources, P.O. Box 441136, Moscow, ID 83844, USA. Tel: 208 885-7742; fax: 208 885-9080. E-mail: cgoldberg@vandals.uidaho.edu

Abstract

Anthropogenic landscape change has had a disproportionately large effect on temperate grassland systems. We used simulations of landscape change to compare the impacts of three commonly applied alternative residential development policies (protecting productive lands, growth boundaries, targeted protection of conservation lands) on ecological and socially important resources in a rural grassland landscape of northern Idaho, United States. Our simulations showed that development patterns were least socially acceptable and most detrimental to ecological resources under current land-use policies. Protecting productive agricultural lands led to the highest level of endangerment for remnant Palouse Prairie communities. Urban growth boundary policies produced the most socially acceptable development patterns and supported habitat for a wide range of species. Targeted conservation actions protected key habitat areas but did little to protect habitat for wide-ranging species. Detailed analyses such as these provide planners with the information required to assess and mitigate the consequences of policy decisions.

Introduction

Private land ownership and anthropogenic landscape change are not uniformly distributed with respect to biophysical characteristics and ecoregion (Hansen et al. 2005; Leu et al. 2008); of all major biomes, temperate grasslands have the highest ratio of converted land to protected area (Hoekstra et al. 2005). Conversion to agriculture in the early 20th century was a major force behind this change, which has left less than 2% of most native prairie systems intact in North America (Noss et al. 1995). Since the peak of U.S. agricultural land-use extent in the 1950s, exurban development has been increasingly expanding into these agricultural lands (Theobald 2001). This new development can affect remnant species directly through loss of habitat and indirectly through increased predation, vehicle traffic, and exposure to invasive species (Odell & Knight 2001; Hansen et al. 2005).

Regulatory mechanisms for protecting native species on private lands are few and controversial; for example, the United States Endangered Species Act has raised a plethora of issues regarding private landowner rights (Meltz 1994). Legally, this act protects animals but not plants from collection on private lands (Endangered Species Act 1973, 16 U.S.C. §1531–1544). The lack of endemism in North American grasslands due to the evolutionarily recent emergence of this biome (Axelrod 1985) means that few prairie species are included in these legal protections. At a local level, regulatory landscape planning approaches to control the pattern of residential growth have recently been undertaken by many governmental entities for a variety of socioeconomic reasons, including maintaining rural values and open space, which may also benefit native species (Bengston et al. 2004). As an alternative bottom-up approach, land trusts and governmental agencies use conservation easements to protect habitat for native species, along with other values, on private lands (Merenlender et al. 2004). While this approach avoids regulatory actions that may be unpopular, it also requires landowner buy-in to protect individual parcels and can be very expensive.

The Palouse Prairie ecoregion was almost entirely converted to dryland agriculture during the 20th century, with less than 1% native habitat remaining (Black et al. 1998). Remnant patches of this system are located close to expanding population centers (Donovan et al. 2009) and are at risk from both direct conversion and the introduction of invasive species (Lichthardt & Moseley 1997). This area has no animal species currently listed as threatened or endangered under the Endangered Species Act and no state laws are in place to protect listed plants on private lands. Growth planning has become an issue of local concern, with dissatisfaction relating to the conversion of productive lands to rural residential development as well as the endangerment of remaining prairie patches by new road and home construction (Lichthardt & Moseley 1997; Donovan 2007).

Rural landscapes contain complex social-ecological systems, where the indirect consequences of alternative landscape planning options may not be apparent. Building on previous work that predicted landscape change using a survey-based model in a portion of the Palouse Prairie landscape (Pocewicz et al. 2008; Nielsen-Pincus et al. 2010), we simulated the ecological and social effects of four relevant alternative policy scenarios: (1) current policy, (2) protecting productive lands, (3) establishing growth boundaries around existing incorporated towns, and (4) the targeted protection of high-priority conservation lands. The goal of our analysis was to identify conflicts and synergies among social and ecological outcomes under commonly applied land-use planning approaches, specifically top-down zoning versus bottom-up land protection strategies. Ecological outcomes included those for socially valued species, as conservation actions that align with local values are more likely to be successful (Wondolleck & Yaffee 2000). Our approach was similar to previous studies modeling ecological impacts of alternative development scenarios on wildlife (e.g., White et al. 1997; Theobald and Hobbs 2002; Schumaker et al. 2004; Wilhere et al. 2007; Beardsley et al. 2009), but differed in that we modeled habitat for vertebrate species in a spatial multivariate framework (see Appendix 1), used spatial predictions of landscape change modeled from social survey data (Pocewicz et al. 2008), and took an interdisciplinary approach examining both social and ecological implications (but see Baker et al. 2004; Polasky et al. 2008). This analysis of trade-offs between landscape planning strategies can be used to inform local governments, planners, the public, land trusts, and other organizations of the potential consequences and mitigation options of alternate residential development policies, as well as to expand the discussion of the broader social and conservation implications of alternative land-use strategies.

Methods

Study area

Our study area consisted of two predominately privately owned, rural counties in northern Idaho, United States (Latah and Benewah Counties, 4,794 km2; Fig. 1). Population centers in the study area are primarily located in agricultural areas, with the largest city (Moscow, ID, ∼22,000 people) situated close to some of the largest prairie remnants. Highly productive soils make this area one of the largest producers of wheat and lentils globally; however, these agricultural lands are experiencing a high rate of conversion to residential development (Pocewicz et al. 2008). Population growth here was 15% between 1990 and 2000, exceeding the national average (Shumway & Otterstrom 2001), and Latah County was in the highest quartile for population growth 2000–2009 (U.S. Census Bureau 2010). Predictions of land-use change indicate that new rural residential development in this area could increase by over 25% in the near future (Pocewicz et al. 2008). Forests here are experiencing increased clear-cutting and thinning, partially due to wildfire concerns.

Figure 1.

Map of the study area, with parcels identified as conservation priorities in the conservation scenario. Areas mapped in white consist of industrial and publicly owned land mainly in forested land cover.

Landscape simulations

To model land-use change, we used responses to a mail survey that asked landowners to probabilistically report their land-use and management plans over the coming 10 years. These reported probabilities were extrapolated across all private, nonindustrial parcels using generalized linear models parameterized with spatially explicit variables including topography, land cover, and ownership patterns (Pocewicz et al. 2008; Appendix 2). Modeled land cover transitions included conversions among agriculture, grassland, low- and high-density forest, and shrub (including early seral forest), with any land cover able to transition to residential development. Using the stochasticity embedded in our land-use change model, we simulated 100 realizations of future land cover, referred to here as the baseline change scenario.

We developed three policy scenarios for residential development as extensions to the baseline change scenario and compared the impacts to ecological and social values across these four scenarios. We realized the alternative scenarios by spatially rearranging development probabilities for nonindustrial private parcels while holding the cumulative probability of development constant. Two top-down development policies (protecting productive agricultural lands and implementing growth boundaries) were presented by Nielsen-Pincus et al. (2010), who evaluated a variety of social impacts. The productive lands scenario moved development probabilities from lands currently in agricultural production to the nearest similarly sized parcel not in agricultural production to approximate the results of policy that attempts to maintain traditional agricultural activities by restricting development from high-productivity soils. The growth boundary scenario used a distance decay function (Nielsen-Pincus et al. 2010) to weight probabilities of development close to incorporated towns higher than those far from towns. Here, we add a conservation scenario, representing permanent protection from development (i.e., purchase of development rights) for parcels with high conservation value.

To construct the conservation scenario, we first identified parcels of high conservation value. We calculated a normalized relative biodiversity index (nRBI) to represent the relative contribution of each private ownership parcel to species habitat across the study area (Schill & Raber 2009; Fig. 2). We developed the habitat maps used to calculate nRBI using two approaches. For plants, we used existing data to locate areas of remnant Palouse Prairie patches (Donovan et al. 2009) and known sites for sensitive plants (Idaho Conservation Data Center, http://fishandgame.idaho.gov/cdc). For animals, we used the list of Species of Special Concern in the Idaho Comprehensive Wildlife Conservation Strategy (Idaho Department of Fish and Game 2005). From this full list, we chose species that depended primarily on grassland habitats found in the nonindustrial privately owned part of the study area. We developed deductive habitat models incorporating minimum patch size, habitat type, and distance from development based on published literature (Table 1) and mapped resulting habitat for each species. We also included winter range for ungulates in the nRBI calculation because protection for this resource was listed explicitly as a goal in the Latah County Comprehensive Plan (adopted by resolution 1994, amended 2004).

Figure 2.

Description of the nRBI index used to construct the conservation scenario, based on Schill & Raber (2009). This index quantifies the area-weighted contribution of each parcel to regional conservation targets by ranking the amount of habitat for target species or communities contained in each parcel compared to that expected for a parcel of that size.

Table 1.  Habitat models for species of concern and socially valued species. Winter range for C. elaphus and Odocoileus hemionus are included in the county comprehensive plan, and so they were used in constructing the conservation scenario. P. sierra are the most well-known frogs in the area due to their loud call. C. californica are valued for hunting and wildlife viewing. Snow depth model used for ungulates is from Manning (2010).
SpeciesMotivationUsed in conservation scenarioGrassland associationHabitat modelCitations
Northern alligator lizard (E. coerulea)Species of concernYesEdge species that uses grassland for thermal requirements and foraging.Grassland, agricultural, and shrub areas ≥50% bordered by forestLais 1976
Short-eared Owl (A. flammeus)Species of concernYesGrassland species adverse to developmentBlocks of grassland or shrub ≥28 ha, ≥90 m from developmentHolt and Leasure 1993; Heckert et al. 1999; Banner and Shaller 2001
Swainson's Hawk (B. swainsoni)Species of concernYesWide-ranging species requiring open land cover, including grasslands, for huntingGrassland and agricultural areas of ≥600 contiguous ha, ≤850 m elevationBechard et al. 1990
Rocky Mountain elk (C. elaphus)Socially valued speciesYesMigratory species requiring low-elevation areas (grassland and former grassland) for winter foragingArea with max snow depth <70 cm in 10 of years 1983–2003, ≥400 m from developmentSweeney and Sweeney 1984; Vogel 1989
Mule deer (Odocoileus hemionus)Socially valued speciesYesMigratory species requiring low-elevation areas (grassland and former grassland) for winter foragingArea with max snow depth <51 cm in 10 of years 1989–2003, ≥400 m from developmentLeopold et al. 1951; Vogel 1989
Sierran Treefrog (P. sierra)Socially valued speciesNoLow-elevation species requiring wetlands in open areas (grassland and former grassland) for breedingAlgorithmic model of isolated wetlands near agriculture and low-density forestsGoldberg and Waits 2009
California Quail (C. californica)Socially valued speciesNoLow-elevation species that uses shrub and grassland for foraging, nesting, and overwinteringShrub and grassland patches >45 m2, >160 m from developmentBarratt 1997; Bolger et al. 1991; Odell and Knight 2001; Rottenborn 1999

To target protection of ecological values, we selected rural parcels larger than 10 acres (4 ha) that had the highest 10% of nRBI values. No new development was allowed on those parcels, and we shifted their probability of development to the nearest similarly sized parcel (10, 10–40, 40–100, or 100 acres). We then used this probabilistic surface, representing the targeted protection of high-priority conservation lands and the two top-down regulatory approaches from Nielsen-Pincus et al. (2010) with baseline projections of change for other land covers (from Pocewicz et al. 2008) to produce 100 stochastic realizations of the future landscape under each scenario.

Ecological impacts

To determine the ecological implications of each land-use planning scenario on the three vertebrate species of concern, we calculated the change in area of habitat under each scenario on nonindustrial private lands. For ungulates, we calculated the change in amount of winter range in the study area. For socially valued species, we measured change in probability of occupancy for Pseudacris sierra (a treefrog) in privately owned wetlands and change in breeding habitat for Callipepla californica (an introduced game bird; Table 1). For sensitive plants and Palouse Prairie patches, we calculated the number of new housing units located within a distance that would make them vulnerable to disturbance from invasive species (105 m; Larson et al. 2001). We used analysis of variance (ANOVA) followed by Tukey-Kramer honest significant diffrence (HSD) tests to determine if different policy decisions led to different outcomes for each species or ecological community.

Social impacts

To compare the social impacts of the conservation scenario with the other scenarios, we calculated four socially important outcomes: (1) wildfire risk to new homes, (2) number of new homes on a declining groundwater resource, (3) number of new homes on productive agricultural lands, and (4) social acceptability of new homes (Appendix 3; Nielsen-Pincus et al. 2010). We compared these measures from the conservation scenario with results from the baseline change, productive lands, and growth boundary scenarios (Nielsen-Pincus et al. 2010).

Social-ecological integration

We integrated the results from ecological and social outcomes to compare commonly applied land-use planning approaches, including top-down zoning and bottom-up land protection strategies. We evaluated the ecological and social conflicts and synergies among policy scenarios using the nine ecological and four social outcomes.

Results

Ecological impacts

Endangerment to plant communities was predicted to increase and habitat for sensitive bird species and ungulates to decrease under all scenarios (Fig. 3). Palouse Prairie, winter range for socially important ungulates, and breeding habitat for Asio flammeus were most sensitive to the simulated land-use planning alternatives. We did not detect a statistically significant change in breeding habitat for Buteo swainsoni or habitat for Elgaria coerulea under simulated changes in development patterns.

Figure 3.

Predicted changes in habitat for sensitive and socially valued species and social implications of four land-use planning scenarios. Error bars are 95% confidence intervals, p-values are from ANOVAs, and letters represent significant differences in outcomes among scenarios from Tukey-Kramer HSD tests. Outcomes for plants are green, vertebrates are blue, and social outcomes are brown.

Protecting productive lands led to an increase in houses threatening prairie patches and a slight decrease in the loss of winter range for Cervus elaphus in the study area and led to the same outcomes for other species as the baseline change scenario. While the conservation scenario protected sensitive plants and habitat for A. flammeus the best, the growth boundary scenario provided the most benefits to prairie remnants, ungulate winter range, and C. californica breeding habitat. However, the urban growth boundary also led to a reduction in predicted habitat for P. sierra and reduced breeding habitat for B. swainsoni in some realizations (Fig. 3).

Social impacts

The conservation scenario led to the highest number of new houses on agricultural lands and was not different from the baseline scenario for wildfire risk to new houses, new demand for declining groundwater resources, and the social acceptability of new development. The most socially acceptable scenario was the growth boundary scenario, which also led to the lowest fire hazard for new homes and the second lowest number of new homes built on agricultural lands. However, the growth boundary scenario was also associated with an increase in new homes using a declining groundwater resource (Fig. 3).

Social-ecological integration

The growth boundary scenario provided the most beneficial results overall for both the ecological and the social outcomes (Fig. 4). The conservation scenario produced favorable results for some (but not all) of the ecological outcomes and performed poorly with respect to social outcomes. The baseline change scenario had the most or second-most detrimental value for all of the statistically significant ecological and social outcomes. The productive lands scenario was as detrimental as the baseline change scenario for the ecological outcomes and was most beneficial for two of the social outcomes but was most detrimental for the other two social outcomes.

Figure 4.

Visualization of scenario outcomes using a spider plot. Points further toward the edge of graph indicate the most benefit or least detriment to each social-ecological outcome. Intermediate values were assigned based on how close each measure was to the best or worse outcome. Greater polygon area generally indicates more benefit for multiple outcomes, but is not quantitatively comparable among scenarios because it is dependent on the ordering of outcomes around the perimeter. P. sierra is not represented here due to lack of likely biological significance among scenario outcomes.

Discussion

The goal of this project was to provide insight into social and ecological trade-offs and synergies across a set of realistic land-use planning options. Predicting the effects of potential land-use policies is a complex process that can illuminate the tradeoffs local communities must evaluate in making planning choices and empower the decision-making process (Hulse et al. 2004, 2009). In our northern Idaho Palouse Prairie system, as in many rural landscapes today, changes predicted to occur under current conditions result in landscape patterns that are both socially undesirable and reduce habitat for sensitive and socially valued species. The top-down approach of implementing a growth boundary and the bottom-up approach of acquiring targeted conservation easements both benefited most of the evaluated species, but while some species benefited more from concentrated development, others had more habitat remaining after targeted conservation actions.

We focused this study at the county level because this is the scale at which land-use decisions are made in many U.S. landscapes (Theobald et al. 2005), including our study area. Conservation priorities may differ if analyzed in a regional context (Huber et al. 2010). We measured habitat presence in this study and did not include measures of connectivity; more detailed spatial and ecological information would be necessary to determine probability of species persistence.

Protecting species from scattered development

The growth boundary scenario resulted in many benefits to both social and ecological outcomes. This scenario provided the development pattern that the general population reported to be most socially acceptable, provided the lowest fire hazard for new homes and protected more productive land than the conservation or baseline scenarios. The amount of habitat remaining for almost all species under this scenario was equal to or greater than that in the baseline or productive lands scenarios. The major drawback of the growth boundary scenario was the likely need for a new source of municipal water.

Concentrating new growth near existing towns had the largest relative benefit for two conservation targets predicted to be most negatively affected by new development: Palouse Prairie and ungulate winter range. For Palouse Prairie, the parcel-by-parcel conservation approach left parts of patches vulnerable to development and edge effects, as prairie patches were often spread across parcels with different conservation values. Because these patches are already extremely small and vulnerable to invasive species (Lichthardt & Moseley 1997; Donovan et al. 2009), the growth boundary approach resulted in a relatively reduced threat to this community. Similarly, winter range for ungulates is concentrated in lower elevation areas with more shallow snow, where the likelihood of development is high (Vogel 1989; Leu et al. 2008). Under the conservation scenario, new development was not allowed in some of these parcels, but the predicted effect of remaining scattered development on ungulates was large, as has been found in other areas (Frair et al. 2008).

The sacrifice zone

While the growth boundary scenario benefited many species, the bottom-up approach of the conservation scenario resulted in more remaining habitat for others. Growth boundaries reduce development impact on the larger landscape by creating higher density sacrifice zones to keep larger blocks intact. For two species associated with agricultural lands, the growth boundary scenario resulted in less habitat than even the baseline scenario in some realizations. While these differences were likely not biologically significant in this analysis, it is important to consider that population centers are not located randomly with respect to habitat (Hansen et al. 2005), and habitat located near towns (agricultural in this case) will likely be lost under this development pattern. Targeted protection of high-priority conservation properties (through either fee title or easement acquisition) could help in cases where very small amounts of habitat need to be protected (e.g., individual locations of sensitive plants) or when habitat is located near existing towns (e.g., A. flammeus). For species operating at a broad scale (e.g., B. swainsoni), this approach would need to include consideration of landscape-level connectivity to be beneficial.

Integrated decision support for informed landscape planning

For this study, we simulated policy scenarios representing distinct value systems acting at a local scale. We found that although protecting farmland is important to the culture and economy of this community (Donovan 2007), the development patterns produced when zoning protects only agricultural parcels is not socially acceptable (Nielsen-Pincus et al. 2010) and would be damaging to both sensitive and socially valued species. The most socially acceptable development pattern resulted from the implementation of a growth boundary, which preserved a large amount of productive land and benefited most species; however, species with habitat located near established population centers and species occupying specialized niches would require additional protection for persistence under this option. Targeted purchase of parcels with high conservation value was predicted to successfully protect habitat for many species, but this success was limited by the edge effects of neighboring parcels and the inability to protect large contiguous tracts. Decision makers rely on a portfolio of policies to achieve complex and interacting objectives; our results illustrate that multiple strategies are required to ensure habitat for diverse groups of species, and the importance of scale in determining the most appropriate strategy for balancing species conservation and residential development.

Appendix 1

Explanation of spatial multivariate habitat modeling approach

Most studies that model the impacts of development scenarios on wildlife represent wildlife habitat either as categorical maps of habitat versus nonhabitat (e.g., Schumaker et al. 2004; Beardsley et al. 2009) or as a single land cover type in a land cover map (e.g., White et al. 1997; Wilhere et al. 2007). We endeavored to incorporate additional complexity in representing wildlife-habitat relationships by using models that accounted for patch size and shape requirements, multiple land cover types (e.g., for edge species), and species-specific development sensitivity. In addition, we incorporated uncertainty in landscape change predictions by modeling development probabilistically, measuring the outcome for each species or community based on 100 realizations of each scenario.

Appendix 2

Landscape change prediction method (Pocewicz et al. 2008)

The original land cover map was classified from Landsat 7 ETM+ data acquired in August 2002 and updated with a 2005 GIS layer of built structures obtained for Latah County and National Agriculture Imagery Program imagery from 2004. The resulting map had a spatial accuracy of 5.6 m and a classification accuracy of 87%. We conducted a mail survey of a random selection of landowners in the study area, stratified by parcel size, where landowners reported their probability of changing between major land uses on their property over the next 10 years. For example, landowners were asked to rank from 0% to 100% how likely they were to plant trees on their agricultural land in this timeframe. We used these reported values to construct predictive models for transition probabilities of land cover types in nonindustrial parcels in the study area. We then applied these models to the landscape and simulated 100 stochastic realizations of landscape change in a 10-year time step using this survey-based stochastic transition model. For parcels in each realization where development was determined to take place, the model randomly chose a point location for each new housing unit within that parcel in a build-out realization, with the number of new houses equal to the number allowed under existing subdivision ordinances for each county. Each new housing unit was expanded to a 0.6-ha block and merged over the existing land use to account for changes to vegetation immediately surrounding new homes and their infrastructure. More details about these methods can be found in Pocewicz et al. (2008).

Appendix 3

Social outcome calculates (Nielsen-Pincus et al. 2010)

1. Wildfire risk to new homes. We predicted wildfire risk to new homes using a map of wildfire hazard modeled using FlamMap3 (Finney et al. 2006) and the Fire Behavior Assessment Tool ArcMap interface (National Interagency Fuels Technology Team, http://www.niftt.gov) with climate, topographic, and canopy cover data. Each pixel was given the maximum value (range 0–5) between rate of fire spread and flame length. For each new housing unit, fire hazard was calculated as the proportion of medium-high values (classes 2–5) within a 240-m radius.

2. Number of new homes on a declining groundwater resource. We calculated the number of new homes predicted to be built on the Palouse Basin groundwater area, a declining resource that provides residential water to the largest city in the study area.

3. Number of new homes on productive agricultural lands. Productive agricultural lands were defined as parcels with >75% of land currently in agriculture and having soils defined as suitable for agricultural use by the United States Department of Agriculture's Natural Resources Conservation Service (United States Department of Agriculture 2008). We calculated the number of new houses on these productive parcels.

4. Social acceptability of new homes. We calculated the social acceptability of each simulated new home using data from a social attribute mapping exercise for this study area (Nielsen-Pincus 2011). In this exercise, randomly selected residents placed stickers on a map indicating where they felt it would be acceptable for new development to occur. These points were combined into a social acceptability surface and data from the location of each simulated housing unit was used to calculate the social acceptability.

Literature cited

Finney, M., Brittain, S., Seli, R. (2006) FlamMap Version 3.0. Joint Fire Sciences Program, Rocky Mountain Research Station. US Bureau of Land Management, and Systems for Environmental Management, Missoula, Montana.

Nielsen-Pincus, M. (2011) Mapping a values typology in three counties of the interior Northwest, USA: scale geographic associations among values, and the use of intensity weights. Soc Nat Resour24, 535–552.

United States Department of Agriculture (2008) Latah and Benewah Counties, Idaho soil survey data. Available from: http://www.soildatamart.nrcs.usda.gov. Accessed 15 December 2008.

Acknowledgments

This research was funded by the USDA McIntire-Stennis Program and National Science Foundation IGERT grant 0014304. We thank three anonymous reviewers for their comments on a previous version of this manuscript. This is contribution 1062 of the University of Idaho Forest, Wildlife and Range Experiment Station.

Ancillary