Landscape Ecology

, Volume 25, Issue 5, pp 683–695

Spatial scale effects on conservation network design: trade-offs and omissions in regional versus local scale planning

Authors

    • Department of Environmental DesignUniversity of California, Davis
  • Steven E. Greco
    • Department of Environmental DesignUniversity of California, Davis
  • James H. Thorne
    • Information Center for the EnvironmentUniversity of California, Davis
Research Article

DOI: 10.1007/s10980-010-9447-4

Cite this article as:
Huber, P.R., Greco, S.E. & Thorne, J.H. Landscape Ecol (2010) 25: 683. doi:10.1007/s10980-010-9447-4

Abstract

Ecological patterns and processes operate at a variety of spatial scales. Those which are regional in nature may not be effectively captured through the combination of conservation plans derived at the local level, where land use planning frequently takes place. Conversely, regional conservation plans may not identify resources important for conservation of intraregional ecological variation. We compare modeled conservation networks derived at regional and local scales from the same area in order to analyze the impact of scale effects on conservation planning. Using the MARXAN reserve selection algorithm and least cost corridor analysis we identified a potential regional conservation network for the Central Valley ecoregion of California, USA, from which we extracted those portions found within five individual counties. We then conducted the same analysis for each of the five counties. An overlay of the results from the two scales shows a general pattern of large differences in the identified networks. Especially noteworthy are the trade-offs and omissions evident at both scales of analysis and the disparateness of the identified corridors that connect core reserves. The results suggest that planning efforts limited to one scale will neglect biodiversity patterns and ecological processes that are important at other scales. An intersection of results from the two scales can potentially be used to prioritize areas for conservation found to be important at several spatial scales.

Keywords

ConnectivityReserve selectionConservation planningCentral ValleyCaliforniaEcoregionMARXANScale effectsCorridor

Introduction

Ecological processes take place at a variety of spatial scales (Forman 1995). Genetic exchange can occur at a local scale while disturbance processes can be regional in nature (Baker 1992; Greco et al. 2007), and migrations may take place at continental or even inter-continental scales. Ecologically successful conservation planning should then take these hierarchical scales into explicit consideration (Poiani et al. 2000). Although spatial scale has been recognized as an important component in conservation, few studies have attempted to quantify the potential effects introduced by scale in conservation network designs.

One planning scale that has received much attention is the ecoregion (Bailey 1996). Global conservation assessments are conducted at this scale (Myers et al. 2000; Olson et al. 2001) and conservation planning at this scale can allow for inclusion of important ecological disturbance processes such as fires and flooding. It also is crucial for preservation of potential animal movement within metapopulations of many large vertebrate species, especially in increasingly fragmented landscapes (Beier and Noss 1998). Many land use planning agencies and conservation organizations have devoted resources to creating ecoregional conservation plans (e.g., Cowling et al. 2003; Thorne et al. 2006).

However, in countries such as the USA, there are few legal structures for implementation of regional conservation plans. Most land use planning authority resides at the local level, with entities such as cities and counties controlling most land use decisions within their borders (Theobald et al. 2005). This “home rule” regulatory environment generally leaves little room for state or federal agencies to implement regional conservation planning projects. Thus, given current land use policy and regulations, implementation of regional conservation plans usually requires integration of multiple local-scale conservation efforts that may not be conducted in coordination with each other.

One method for creating a regional conservation plan begins by identification of important core areas that contain regionally important habitats and ecological features (Noss et al. 1999; Margules and Pressey 2000). These features might include rare or sensitive species (Rothley 1999; Wiersma 2007), high biodiversity (Margules et al. 1988; Prendergast et al. 1993; Shriner et al. 2006), and ecological processes (Turner et al. 1999). Another conservation planning approach is the use of focal species (Lambeck 1997; Wiens et al. 2008), in which a species considered to be the most sensitive to perturbation of a particular ecosystem pattern or process is used as a surrogate (or “umbrella”) in the conservation planning process for other species sensitive to disturbance of the same ecosystem components. Core reserve areas that are identified to serve the needs of focal species are then presumed to offer similar protection for other species that fall under their “umbrella” (Lambeck 1997; Caro and O’Doherty 1999). There have been some criticisms of the focal species approach (see Lindenmayer et al. 2002); however, it remains a useful planning construct especially in areas requiring substantial restoration to achieve conservation functions (Lambeck 2002). In our study we address the limitations of a single-species umbrella approach by selecting a suite of focal species that use a spectrum of regional habitat types.

In the design of ecological networks, core areas identified by conservation planning processes are linked by corridors meant to permit the movement of animals and plants between the cores in order to maintain the viability of the core populations (Rosenberg et al. 1997; Bennett 2003). While there is still concern whether corridors provide a universally effective means of conservation (Simberloff et al. 1992; Davies and Pullin 2007), several studies have shown that they can provide ecological benefits to regional populations (Tewksbury et al. 2002; Damschen et al. 2006).

Some effects of spatial scale on the conservation planning process have been noted. Change in the grain size of raster datasets used in reserve selection has been shown to alter the resulting reserve networks (Andelman and Willig 2002; Rouget 2003), and species range size and location as defined by species distribution models is also affected (Seo et al. 2009). Additionally the size of the planning units can affect identified conservation networks (Warman et al. 2004; Pascual-Hortal and Saura 2007) as can the extent of the planning area (Erasmus et al. 1999; Pascual-Hortal and Saura 2007; Vazquez et al. 2008). However, little work has been done to date on investigating scale effects on whole ecological networks that incorporate connectivity analysis and corridor planning in addition to reserve selection. It remains to be seen how functional connectivity (sensu Noss and Daly 2006) changes as the spatial extent of analysis is changed.

This paper examines the level of spatial congruity between conservation networks designed to meet ecological needs that are identified at ecoregional and county (local) scales. Here we ask: is the whole equivalent to the sum of its parts? Our general hypothesis is that conservation networks derived at one scale will not be the same as those derived at another scale in the same location. If our hypothesis proves to be correct, the implication is that land planners and managers need to balance the ecological needs identified at multiple spatial scales in order to create effective (i.e., functional and sustainable) conservation networks.

Methods

Study area

The analysis was conducted in the Central Valley ecoregion of California (Fig. 1), as defined by Hickman (1993). This ecoregion is largely agricultural, having been converted beginning 150 years ago with the advent of the California Gold Rush. The natural areas in the region (primarily riparian forest, freshwater wetland, grassland, and oak woodland) are generally small and highly fragmented remnants embedded within this agricultural landscape matrix (Ricketts et al. 1999). These natural land cover types are currently facing urbanization pressure with the population of the region expected to approximately double in the next 40 years (PPIC 2006). This ecoregion encompasses portions of 29 counties (only one of which is entirely within the ecoregional boundaries) (Fig. 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs10980-010-9447-4/MediaObjects/10980_2010_9447_Fig1_HTML.gif
Fig. 1

The location of the Central Valley ecoregion within California. Also shown are the 29 counties portions of which comprise the ecoregion with the five used for analysis indicated in shades of gray

Data preparation for core reserve selection

We created a regularly spaced planar tessellation surface of hexagonal cells 13.3 ha in size using a geographic information system (GIS; ESRI 2005) for use as planning units. Conservation land polygons (public or private parcels managed for conservation purposes), as identified in the Public, Conservation and Trust Lands dataset (California Resources Agency 2005), were then embedded in the hexagonal planning unit surface using the ArcGIS Update tool. Polygon fragments smaller than one hectare resulting from this embedding process were merged with neighboring polygons, creating a final study area of 424,805 planning units with a one hectare minimum area (Huber 2008).

Eight conservation targets, comprised of seven focal species and one unique ecological community were selected to represent the major ecological patterns and processes of the study area and serve as potential “umbrellas”: tule elk (Cervus elaphus nannodes; chosen to represent large-scale lowland and upland connectivity); bobcat (Lynx rufus; fine-scale forested connectivity); giant garter snake (Thamnophis gigas; freshwater wetlands); pronghorn (Antilocapra americana; grasslands); western yellow-billed cuckoo (Coccyzus americanus occidentalis; riparian forest); San Joaquin kit fox (Vulpes macroitis mutica; desert scrub); Swainson’s hawk (Buteo swainsoni; agricultural/natural areas interface); and vernal pool community complexes, included as a unique ecological feature (Huber 2008).

We developed suitability models for the seven focal species, using the full Central Valley ecoregion as the modeling area for tule elk, bobcat, pronghorn, and Swainson’s hawk. For giant garter snake, kit fox, and yellow-billed cuckoo we used range maps to spatially restrict the habitat analyses as these species historically did not exist throughout the entire ecoregion. Range maps for giant garter snake and kit fox are from the California Wildlife Habitats Relationship System database (CWHR; CDFG 2005). We used the historic distribution of riparian forest in the Central Valley (GIC 2003) as the potential range map for the yellow-billed cuckoo.

The highly fragmented nature of the study area landscape would necessitate substantial habitat restoration in order to assemble adequately large or well-connected blocks of land suitable for self-sustaining populations of many species. Thus, we could not use standard habitat suitability indices solely based upon existing conditions, but rather also took into account the context and “restorability” of human-converted planning units. We used the following habitat variables (for all species unless otherwise noted):
  • Current land cover: a value of 0–1 for each major land cover type for each focal species was taken from the CWHR database (CDFG 2005). Values were derived through expert-applied suitability ratings of habitats for individual species in three life necessities—breeding, forage, and cover. The land cover dataset used was the statewide California Department of Forestry and Fire Protection land cover dataset (FRAP 2002).

  • Road density: TIGER road data (U.S. Department of Commerce 2007). Road density was calculated in km/km² at both a 3 and 5 km radius. These densities were then converted to a 0–1 scale and inverted; low density raster cells have values closer to 1 while high density raster cells have values closer to 0.

  • Urban area density: Farmland Mapping and Monitoring Program (FMMP 2004) urban data. Urban areas were given a value of 1 and non-urban areas a value of 0. The average urban area value within both a 3 and 5 km radius was calculated. These density values were then converted to a 0–1 scale and inverted, so that low density urban areas had a value of 1 and high density urban areas a value of 0.

  • Natural area density: FRAP land cover data (FRAP 2002). All native vegetation types (plus annual grasslands) were included in this category. This metric was used to lower the overall scores of those habitats that are largely surrounded by potentially incompatible land uses (e.g., agriculture or urban) and to raise the values of habitats embedded within a natural matrix, and thus less susceptible to detrimental human-caused edge effects. To calculate this value, the natural vegetation types were given a value of 1 and non-natural types a value of 0. The 3 and 5 km radii were used to calculate natural area densities which were then converted to a 0–1 scale with areas of high natural area density receiving values closer to 1.

  • Current land management status: Public and Conservation Trust Lands data (California Resources Agency 2005), a dataset comprised of all public lands as well as private lands managed for their conservation value. Lands contained in this dataset were given a value of 1 and lands outside these boundaries a value of 0.

  • Waterway density (for giant garter snake only): National Hydrography Dataset (NHD; USGS 1999) waterway data. Waterway density was calculated in km/km² at a 3 km radius. These values were then converted to a 0–1 scale with areas of high waterway density receiving a value of 1.

  • Tree/grass interface density (for Swainson’s hawk only): FRAP land cover data (FRAP 2002). The density of forest types was calculated with a maximum value of 0.5 (when forest accounts for ≥50% of the total neighborhood area). The same was done with grasslands (including field-type agricultural land cover types). Results were summed, for an overall interface (or ecotone) metric ranging from 0 to 1, with a value of one equal to 50% forest and 50% grassland, the presumed optimal habitat combination for this species.

Density surfaces were calculated using a 5 km radius for tule elk, yellow-billed cuckoo, and bobcat, while a 3 km radius was used for all the other focal species. These values were chosen to account for mobility, with wider-ranging species assigned the larger radius. For each focal species’ suitability model, the current land cover value was given half of the overall weight in order to ensure that existing habitat was given the highest priority for inclusion in the identified conservation network. The value assigned to the land cover variable represented an approximate existing habitat suitability score. The other variables were then equally weighted and summed to provide the other half of the overall weight of the habitat score. These variables represented both current spatial context and potential for restoration. The sum of the land cover score and the other variables’ scores were then converted to a 0–1 scale (with 1 being the highest value for a particular species). The vernal pool focal element was given a simple binary score of 0 (not present) or 1 (present) determined by the boundaries of the vernal pool complexes dataset (USFWS 1998). This dataset represents both the pools proper and critical surrounding uplands.

The hexagonal units used for subsequent analyses were given a habitat value for each focal element by multiplying the planning unit area by the average value of the raster cells representing focal element habitat suitability that fell within the planning unit. While the majority of planning units were identical in area, existing conservation lands planning units ranged in size (Huber 2008). Also, many of the planning units on the study area boundary or existing conservation lands boundaries were smaller than the standard size.

Core reserve selection at the regional scale

The planning units along with their values for the eight conservation targets were inputted into the MARXAN reserve selection algorithm (Ball and Possingham 2000). This algorithm evaluates the effectiveness of various combinations of planning units at achieving conservation goals. While it is not an optimization tool, MARXAN identifies “low cost” solutions (cost is defined by the user and is generally both ecological and economic in nature). The cost value in this analysis was designated simply as the area of the planning unit. Representational goals were set at 30% of the total habitat value for each focal species (Svancara et al. 2005). A unit-less “boundary modifier” of 1,500 was selected after test runs showed that this parameter value led to what we considered to be a pattern of reserve clustering that balanced the desires for large, compact reserves and spatial dispersion of reserves across the study area. We conducted 100 runs with 10 billion iterations each (see Thorne et al. 2009 for more information on MARXAN).

For the ecoregion-wide planning effort, we identified only those reserves that were repeatedly selected by MARXAN and were part of larger blocks of existing or potential habitat rather than scattered parcels of land. Thus, we designated 30 runs as the minimum number of times out of 100 that a planning unit had to be selected as part of a MARXAN “solution” for it to be included in our identified reserve network. Further, we eliminated contiguous groupings of the selected planning units that were less than 2,000 ha in size. We chose this threshold because it was both large enough to encompass the home ranges of the widest ranging focal species (tule elk, Howell et al. 2002; bobcat, Riley et al. 2003; pronghorn, Jacques et al. 2009) and small enough to reflect the currently fragmented Central Valley landscape.

Connectivity analysis at the ecoregion scale

Each of the identified core reserves was analyzed for potential use by the eight conservation targets. We selected a mean habitat value of 0.33 within a specific reserve as the minimum for that reserve to be considered suitable for a particular focal species. This value was chosen in order to include reserves in a particular species’ ecological network that contained both high quality habitat and marginal habitat that could potentially be enhanced through restoration efforts over a long time frame. It also corresponds to the value given to low quality habitat in the CWHR system (CDFG 2005). Connectivity analyses were then conducted individually for each of the five mobile terrestrial focal species only between the cores that were selected for that species.

The connectivity analysis was performed using the Least Cost Corridor (LCC) ArcGIS function (ESRI 2005). LCC analysis uses a user-defined “cost surface” to evaluate the connectivity value of raster cells located between designated end points (e.g., the selected cores) by summing the calculated cost distances from each of the two end points (Theobald 2006). The cost surface used in these analyses was the inverse of the habitat value dataset (e.g., high habitat values became low cost values) for each focal species. This analysis produced a “connectivity surface” between adjacent reserves. To identify potential corridors from these surfaces, we selected those raster cells with a value of within 2% that of the lowest cost raster cell and converted these to a polygon shapefile.

Local analysis—a county-based approach for reserve selection and connectivity

We replicated the regionally based methods to run independently in five counties (out of the 29 comprising the Central Valley) within the Central Valley: Sutter, Yolo, Stanislaus, Fresno, and Kern (listed north to south; Table 1). These counties were selected to provide for a wide range of local conditions within the ecoregion. All five counties, with the exception of Sutter, have portions of their area in the various surrounding mountainous ecoregions. Only those portions of the counties falling within the Central Valley ecoregion boundary were used.
Table 1

Extents of cores and corridors in the five analysis counties

County

Network

Cores

Corridors

Total (ha)

Total (%)

Area (ha)

(%)

Area (ha)

(%)

Sutter

Region

9,641.0

6.1

19,494.9

12.4

29,135.9

18.5

Local

18,993.3

12.1

7,361.9

4.7

26,355.2

16.7

Yolo

Region

18,508.4

9.2

67,491.7

33.4

86,000.1

42.6

Local

22,568.0

11.2

42,156.5

20.9

64,724.5

32.0

Stanislaus

Region

37,157.5

12.3

93,849.2

31.0

131,006.7

43.3

Local

28,578.5

9.4

116,055.5

38.4

144,634.0

47.8

Fresno

Region

55,528.0

7.0

116,584.7

14.6

172,112.7

21.6

Local

61,893.2

7.8

45,829.5

5.8

107,722.7

13.5

Kern

Region

81,907.7

8.9

223,048.1

24.4

304,955.8

33.3

Local

77,257.6

8.4

206,790.6

22.6

284,048.2

31.0

The “Region” network refers to the section of the regional network located within each county while “Local” refers to the network identified at the local scale. The “%” columns refer to the portion of each county’s Central Valley ecoregion area (rather than the full county extent) comprised by the conservation network components

The same focal elements were used as conservation targets in these analyses although for some counties not all of the focal species historically occurred there (e.g., the San Joaquin kit fox did not occur in Sutter County) and so were excluded from consideration in those respective counties. Because we considered the regional minimum reserve area (2,000 ha) to be too large for a local-based approach, we scaled the local reserves to a percentage of the area of each county. Specifically, we set the minimum reserve size for both the ecoregion and each of the five counties equal to 0.034% of the respective total areas. This value was derived by calculating the percentage of the total area of the ecoregional study area that the 2,000 ha minimum core reserve area covered.

The local-scale corridors were identified in the same manner as with the regional approach. However, the county boundary was used as an analysis mask so that the full corridor area had to fall within the county.

Overlap analysis

An overlap analysis was conducted between the results from the two spatial scales of analysis to identify differences between them. The county boundaries were used to extract the results from the ecoregional analysis for the five counties that were individually analyzed. The results from the ecoregional analysis were overlaid on the local-scale results and the area of the unioned sets was calculated. Finally the total area of each of the unioned sets was classified into three categories: (1) occurs in both networks, (2) only found in the regional network, or (3) only found in the local network. In addition, an overlap analysis was conducted solely on the reserves and likewise solely on the corridors for each of the five counties.

Focal element coverage

The effect of scale on representation of the focal elements in the two scales of modeled ecological networks was measured through comparison of habitat value for each focal element in the core reserves. For each county, we calculated a “habitat value area” (H) for each focal element:
$$ H = A \times E $$
(1)
Here, A refers to the total core reserve area and E represents the mean suitability value for that focal element across the total core reserve area. Within each county, a modified ratio score (V) representing total habitat value area within reserves identified at the local versus regional-scale was calculated:
$$ V = 2 \times \left[ {{\frac{{H_{L} }}{{(H_{L} + H_{R} )}}} - 0.5} \right] $$
(2)
where H is the “habitat value area” (see Eq. 1), L is local-scale, and R is regional-scale. This score then has values ranging from −1.0 to 1.0, with scores of −1.0 indicating habitat for that focal element was only found in regional-scale cores, 1.0 indicating habitat only within local-scale reserves, and 0.0 indicating equal habitat value areas between the scales.

Mean V scores were calculated by county and by focal elements. Finally, in order to measure the magnitude of this scale-driven variation of habitat value coverage, absolute mean difference (AMD) was calculated by taking the mean of the absolute value of V, leading to a value ranging from 0 to 1.

Results

Regionally based analysis

Fifty-two core reserves within the Central Valley ecoregion were delineated (Fig. 2). The spatial distribution of the core reserves was relatively uniform throughout the region. The core reserve areas ranged in size from 2,061 to 116,527 ha and covered 12.2% of the total area of the ecoregion. The connectivity analysis identified 388 species-specific corridors linking the core areas (bobcat = 120, giant garter snake = 27, kit fox = 50, tule elk = 121) (Fig. 2). These corridors totaled 18.2% of the study region area.
https://static-content.springer.com/image/art%3A10.1007%2Fs10980-010-9447-4/MediaObjects/10980_2010_9447_Fig2_HTML.gif
Fig. 2

A potential Central Valley ecoregional conservation network. The network includes cores comprised of clusters of planning units each of which was part of at least 30 run solutions and collectively measured at least 2,000 ha in area. The corridors were identified for all five focal species and include all pixels within 1.02× the value of those found on the true lowest cost path. Marxan selection results (out of 100 runs) are shown in a. Identified cores and corridors are shown in b

The total area identified as cores from the ecoregional network that fell within each of the five analysis counties (Table 1) ranged from 6.1% (Sutter County) to 12.3% (Stanislaus County) of county area located inside the ecoregional boundary. The area occupied by identified regional corridors ranged from 12.4% (Sutter County) to 33.4% (Yolo County). Footprints of the entire network ranged from 18.5% (Sutter County) to 43.3% (Stanislaus County).

Locally based analysis

The number of core area reserves identified in the local scale analyses ranged from 7 to 18 (Sutter County and Kern County, respectively) and from a minimum area of 61.8 ha (Sutter County) to a maximum area of 43,341.5 ha (Fresno County). The number of species-specific corridors identified in these analyses ranged from 27 (Sutter County) to 169 (Kern County).

The total area identified as core reserves in the local networks ranged from 7.8% (Fresno County) to 12.1% (Sutter County) of the total county area located inside the ecoregional boundary. The area occupied by local corridors ranged from 4.7% (Sutter County) to 38.4% (Stanislaus County). Footprints of local ecological networks ranged from 13.5 to 47.8% (Fresno County and Stanislaus County, respectively) of the total county area located inside the ecoregional boundary.

The mean percent of total county area covered by core reserves in the local networks was 9.8%, slightly more than in the regional network which was 8.7%. The reverse was true with the identified corridors: the local mean was 18.5% and the regional mean was 23.2%. The overall county mean was 28.2% for the local networks and 31.9% for the regional networks.

Overlap analysis between regionally- and locally-based network designs

The mean percent overlap (i.e., spatial congruence) between the two conservation network scales within the five analysis counties was 35.9% (Fig. 4). Of the remaining area of the unioned regional and local networks, a mean of 41.0% was identified only through the regional analysis and 23.2% through the local analysis. The county with the greatest overlap was Kern with a 49.9% overlap while that with the least was Stanislaus with a 23.4% overlap between the networks (Figs. 3, 4).
https://static-content.springer.com/image/art%3A10.1007%2Fs10980-010-9447-4/MediaObjects/10980_2010_9447_Fig3_HTML.gif
Fig. 3

Results of the overlap analysis for each of the five individual counties. The ecological networks from the analyses for individual counties were overlaid on the portions of the regional network found within each county. The colors depict areas of overlap (black), areas that are only identified for the individual county (medium gray), and those found only in the regional network (light gray)

https://static-content.springer.com/image/art%3A10.1007%2Fs10980-010-9447-4/MediaObjects/10980_2010_9447_Fig4_HTML.gif
Fig. 4

The percent overlap of (top) cores (middle) corridors, and (bottom) the overall network between the regional and local scales for each of the five analysis counties. The mean of the county scores is also included. “Regional %” refers to percent of the unioned sets that is just found at that scale while “Local %” refers to that at the local scale only. “Both %” is the amount of overlap between the scales for that network component

The county with the greatest overlap of cores was Fresno (72.4%) and that with the least was Sutter (41.4%). The mean core overlap for the five counties was 54.9%, with 15.3% the mean for each county identified as a regional core only and 29.8% as local core only (Fig. 4).

The county with the greatest overlap of identified corridors was Kern (44.8%) and the least was Sutter (5.1%). The mean corridor overlap was 21.4%, substantially less than the mean core overlap. The mean corridor area identified during the regional analysis only was 55.2% of the total area of the unioned corridors while the mean local corridor area was 23.5% of the total corridor area (Fig. 4).

Focal element coverage

Two of the five analysis counties (Sutter and Yolo) had higher habitat value areas in identified local core reserves for the seven occurring focal elements (Table 2). All vernal pools included in cores in these counties (as well as Kern County) were found in local-based reserves (scaled ratio score = 1.0). Sutter County focal elements had a modified ratio score and AMD of 0.44, the largest of the five counties, while Yolo County had scores of 0.26. Conversely, in Stanislaus County, higher habitat value areas were included in cores for all focal elements except for yellow-billed cuckoo. The effect was less pronounced than for Sutter and Yolo counties, however, with a modified ratio score of −0.13 and AMD of 0.16. The only other county with a negative modified ratio score (indicating more habitat value area found in regional core reserves than local) was Fresno County, although only three focal elements received negative ratio scores (with the other five focal elements having slightly positive ratio scores). Fresno County also had the lowest AMD (0.07) indicating a nearly equal distribution of habitat value areas in regional and local cores.
Table 2

Scaled habitat value area ratios for each focal element in the identified core reserves in each analysis county

County

Tule elk

Bobcat

Pronghorn

Giant garter snake

San Joaquin kit fox

Swainson’s hawk

Yellow-billed cuckoo

Vernal pools

Mean

AMD

Sutter

0.36

0.35

0.32

0.35

n/a

0.32

0.38

1.00

0.44

0.44

Yolo

0.13

0.14

0.17

0.12

n/a

0.17

0.08

1.00

0.26

0.26

Stanislaus

−0.15

−0.16

−0.21

−0.14

−0.18

−0.16

0.10

−0.15

−0.13

0.16

Fresno

0.02

0.03

0.04

−0.18

0.04

0.02

−0.09

−0.16

−0.04

0.07

Kern

−0.03

−0.04

−0.02

0.29

−0.04

0.03

0.20

1.00

0.18

0.21

Mean

0.07

0.07

0.06

0.09

−0.06

0.08

0.13

0.54

  

AMD

0.14

0.14

0.15

0.22

0.09

0.14

0.17

0.66

  

The scale ranges from −1 to +1, with positive values indicating more habitat value area in locally based core reserves and negative values indicating more habitat value area in regional cores. Absolute mean difference (AMD) is the calculated mean of the absolute scaled ratios, thus ranging from 0 to +1

Five of the eight focal elements had modified ratio scores between 0.06 and 0.08 (Table 2). Vernal pools had the highest modified ratio score (0.54), indicating that in the five analysis counties, much of the total vernal pool area was included in local core reserves rather than in regional cores. Of the eight focal elements, only the San Joaquin kit fox had a negative score (−0.06), indicating greater inclusion of kit fox habitat at the regional scale than at the local scale. AMD ranged from 0.14 to 0.17 for five of the focal elements while vernal pools had the highest AMD (0.66).

Discussion and conclusions

These results clearly demonstrate the importance of considering spatial extent in the design of reserve systems to achieve conservation goals. The two spatial scales of analysis identified substantially different conservation networks within the respective study area. On average, just over one-third of the area identified for inclusion within a conservation network at either scale was identified at both scales.

Especially significant is the lack of congruence between corridors identified at the different spatial scales. This disparity suggests several things. First, many regionally important corridors are not identified at the local scale. On average only 43.8% of regional corridors overlapped at all with locally based corridors. Regional and inter-regional connectivity could therefore be threatened if planning only occurs at the local scale. Second, corridors connecting locally important core areas can be missed if only the regional scale is taken into account in the planning process. Only 53.5% of local corridors overlapped with those identified at the regional scale. This could lead to isolation of locally important core areas.

These findings lead to the conclusion that planning at the local scale and amalgamating multiple local efforts in lieu of a formal regional planning process can lead to overlooking of important regional conservation needs (a phenomenon referred to by Groves (2003) as the “tyranny of the local”). Without referring to the regional context, there is little way to account for connectivity between areas separated by an intervening administrative unit (e.g., a county) or even between adjacent administrative units. While our local-scale analysis identified corridors leading from cores to the county boundaries, there was little indication that these corridors would link up to anything deemed ecologically important on the other side of a county’s boundary. Not only were important regional corridors not identified locally, but local analysis also led to a potentially fractured network across the region with subunits that do not integrate. We feel that resolution of these “myopic effects” cannot be addressed at the local scale alone and requires explicit regional integration.

The results of this study also suggest the converse to be true: a conservation plan that only addresses regional needs can miss areas of local ecological or cultural importance. There are several reasons why this is an important issue for conservation planning. First, if the components of a conservation network are not stratified across subunits within a region, fine-scale ecological variation may not be adequately represented or protected (Groves et al. 2002). The Central Valley in California is approximately 650 km in length; intra-taxon populations of many species may harbor important genetic diversity across that distance (Patten and Yang 1977). For example, no vernal pool habitat was included in regional-based core reserves in three of the analysis counties even though it occurs there (Table 2); a heavier emphasis was placed on conservation of vernal pools in Stanislaus and Fresno Counties. Neglect of this focal element in several counties could mean loss of ecological and genetic variation if these pool complexes are lost. Second, many ecological conservation planning efforts are intertwined with city and county efforts to provide open space for their citizens. If a purely regional approach to conservation planning is adopted, many areas might not be deemed biologically valuable enough to warrant expenditure of finite conservation resources. Finally, areas that are disproportionately targeted for inclusion in a regional conservation network might place an economic burden on the inhabitants (King and Anderson 2004). If one county in the region is a focus of conservation activity that reduces or eliminates property tax revenue as well as potential economic activity on that land, it will potentially be seen as unfairly assuming the financial burden of the conservation network for the larger region.

This study also reveals that even if there is considerable overlap of conservation network designs at county and regional scales, the designation as either core or corridor might differ. Results from Sutter County (Fig. 4), for example, indicate that much of the regional conservation network within this county consists of corridors linking cores in adjacent counties, and many of those corridor linkages happen to coincide with areas identified as cores at the local scale. Moreover, land management might differ for cores and corridors (Soulé and Terborgh 1999); thus with land being identified as important core reserve at one scale and corridor at another scale, managers would have two different mandates for these areas within a combined network. However, an alternative way to potentially view this phenomenon is that areas identified as cores at the local scale and corridors at the regional scale could serve as “stepping stones” in a regional linkage network.

While the analysis in this paper indicates that changes in spatial scale can have dramatic effects on design of a conservation network, some caveats should be noted. First, we only looked at two possible scales of analysis, local and ecoregional. However, given the results of this study, there are likely to be cross-ecoregional boundary effects as well. Additionally, we used only one of many potential techniques for identification of a conservation network. We focused on focal species habitat needs (including restorable areas); we did not include approaches such as representation or irreplaceability (sensu Margules and Pressey 2000). These other techniques, either as stand-alone analyses or in conjunction with a focal species analysis, might yield conservation networks more resilient to change at different spatial scales.

Finally, caution should be taken with the spatial footprints of the ecological networks identified here. The focus of the analysis in this paper is on scale effects rather than the network design itself. While we strived to adhere to sound conservation planning principles as found in the scientific literature (e.g., Soulé and Terborgh 1999; Beier et al. 2008) as well as to the in situ realities of the Central Valley region, it was beyond the scope of this effort to fully investigate the sensitivity of our analysis to choices made in terms of cut-off points and variable weightings. For example, analyses should be conducted to better understand an effective cut-off point for managing a core area for a particular focal species. While we assumed that a minimum mean value of 0.33 would indicate that a core consists of good and moderate habitat, this value may or may not be sufficient to ensure future viability in that core.

This study presents an approach by which scale differences can be detected and resolved. Areas that appear in the identified conservation networks at both scales, especially those that retain their component designation (i.e., core or corridor), could serve as the foundation for a network that incorporates the conservation needs of multiple scales. Once these areas have been accepted as components of a network, decisions on inclusion of other areas to complement them can be made on an individual basis, but with a greater understanding of the ecological role they might play in the overall network. This approach is one means by which the spatial hierarchical framework called for by Noss et al. (1997) could be implemented.

A key question that will need to be addressed on a case-by-case basis by participants in the conservation planning process is: what level of spatial stratification (i.e., scale selection) of conservation effort is appropriate in the given context? Successfully answering this question can lead to a more robust plan that addresses a variety of ecological processes and biological concerns. The ecological and legal/policy realities of the current planning environment in the USA require that local implementation of conservation efforts be conducted within a regional context for greater effectiveness (Simberloff et al. 1999). In this study we demonstrated an emergent property of multi-scale planning and that the whole is, in most cases, different than the sum of its parts.

Acknowledgments

The authors would like to thank the Arenz Foundation for their support of this project. We would also like to thank Rob Thayer for his helpful comments on earlier versions of this manuscript.

Copyright information

© Springer Science+Business Media B.V. 2010