Development and Pilot Application of the
California Urban and Biodiversity Analysis (CURBA) Model
by John D. Landis, Juan Pablo Monzon, Michael Reilly, and Chris Cogan(1)
Introduction
The U.S. has made tremendous progress over the last 25 years in improving its air and water quality. Between 1986 and 1995, for example, national average carbon monoxide emissions decreased 37 percent, even as total vehicle miles of travel (VMT) increased 31 percent (EPA, 1996). Hundreds of rivers and water bodies classified as "unfishable" in 1975 are now classified as fishable.
Far less progress has been made in the areas of land and habitat conservation. Most habitat loss during the last 25 years has occurred as a result of urban growth(2). The nation's metropolitan areas are currently consuming land at a much faster rate than they are adding population. Whereas the nation's metropolitan population increased 28 percent between 1970 and 1990, its metropolitan land area increased 82 percent (Bureau of the Census, 1990). Nowhere are the problems of metropolitan growth and habitat loss more serious than in California. More than half of the species listed as "endangered" under the U.S. Endangered Species Act make their homes in California.
Urban growth diminishes habitat quality in three ways. Foremost, it consumes habitat, replacing natural, presumably diverse habitats with less diverse and less natural urban habitats, or, in some cases, with barren habitats. Second, it reduces habitat integrity by promoting fragmentation. Third, it generates vegetation and species-damaging spillover effects such as runoff and air pollution. The contribution of each of these factors to overall habitat loss varies by location.
In most parts of the U.S., the task of regulating urban development--and thus of protecting habitat--falls to local government(3). It is a task beset with difficulties. The first such problem is that of jurisdictional mismatch. Because municipal boundaries are determined politically rather than functionally, individual jurisdictions are rarely able to undertake effective habitat conservation programs. This is especially a problem in large, fast-growing, and politically-fragmented metropolitan areas. Second, even where they are inclined to do so, the types of development controls available to local governments--usually some form of zoning--are poorly suited to protecting sensitive environmental areas. When and where they do exist, local environmental protection initiatives tend to be organized along functional lines (e.g., protecting farmlands, hillsides, or riparian areas) rather than along ecological lines.
Third, planning institutions of all types--not just local governments--typically lack the analytical capability to undertake long-term strategic planning. In the case of habitat planning this involves understanding where, when, and under what circumstances future urban growth is likely to occur. It also involves understanding the effects of different patterns and forms of urban development on habitat quality and species viability. The final--and in some ways most significant difficulty--is the fact that we don't yet know what forms and types and mixes constitute stable ecologies. It is one thing to regulate urban development in the hope of preserving habitat; it is quite another to conserve and promote diverse and stable ecologies.
The California Urban and Biodiversity Analysis (CURBA) Model was developed as a first step in addressing some of these issues. The CURBA Model was designed to help bridge the gap between urban land use planners--who are principally concerned with directing urban growth--and conservationists and wildlife ecologists--who are concerned with promoting environmental and ecological quality. The CURBA Model integrates three sets of data sources and modeling approaches which have heretofore been separate:
1. A statistical model of urban growth incorporating spatial and non-spatial components.
2. Procedures for simulating the effects of alternative development and conservation policies on the amount and pattern of urban growth.
3. Detailed and spatially-explicit map and data layers regarding habitat types, biodiversity, and other natural factors.
The CURBA Model was designed and developed at the University of California, Berkeley by a team of planning and environmental researchers. It runs in ArcView 3.0, using the Spatial Analyst and Dialog Developer extensions. So far, CURBA Model datasets and equations have been developed for nine California counties, including El Dorado, Monterey, Nevada, Placer, Sacramento, San Joaquin, Santa Cruz, Sonoma, and Stanislaus. Datasets (and models) for additional counties are under development.
This paper explains the logic, calibration, and use of the CURBA Model. We begin by looking at the structure of the model, how it measures and reports habitat quality, and how it integrates vastly different spatial data sources. Next, we present the results of a pilot study of the use of the CURBA Model, focusing on issues of urban expansion and habitat loss in Santa Cruz County. We conclude by reviewing the model's features, strengths, and limitations.
The Logic of the CURBA Model
The CURBA Model is a distant cousin of the second generation of the California Urban Futures Model (see Landis and Zhang 1998) Like CUF II, the CURBA Model consists of two major components: (1) An Urban Growth Model, which includes procedures for calibrating countywide equations describing past urbanization patterns, and for using those equations to construct future development scores; and (2) A Policy Simulation and Evaluation Model, consisting of procedures for simulating how alternative development policies might affect the future urbanization patterns and the impacts of those patterns on habitat integrity (Figure 1). The Urban Growth Model component uses spatial data but is calibrated outside of a GIS environment. The Policy Simulation Model is embedded in ArcView and makes use of ArcView Spatial Analyst as well as specially-written Avenue scripts. The CURBA Model's basic unit of analysis and minimum mapping unit is the one-hectare (100 meter by 100 meter) grid cell.
Data Sources
The CURBA Model makes use of a wide variety of spatial and non-spatial data sources, including:
1. The location and type of farmland and urban development: Information on the location and type of farmland in California is collected and distributed as part of the ongoing work of the California Farmland Mapping and Monitoring Project (FMMP). Begun in 1986, the FMMP database covers all or parts of most California counties and is updated every two years. Areas (polygons) are classified into eight categories (urban, prime agricultural lands, agricultural lands of importance to the state economy, agricultural lands of importance to a local economy, unique agricultural lands, grazing lands, wetlands, and other areas) according to current land use and land cover, soil quality, and cultivation potential. Prior to its use in the CURBA Model, FMMP data is converted from its native polygon form.
FMMP data layers are used in both components of the CURBA Model. They are used as the dependent variable in the Urban Growth Model to identify multi-year changes in the location and extent of urbanization. They are used in the Policy Simulation and Evaluation Model as potential constraints to future urbanization. FMMP data are grided into one hectare cells (100 meter by 100 meter) prior to their use in the CURBA Model.
2. California GAP Analysis GIS Layers: Developed by Scott, et.al (1993) at the University of Idaho, GAP Analysis is a GIS-based procedure for identifying "gaps" in biodiversity protection. Gap Analysis consists of three primary GIS layers: (i) the distribution of actual vegetation types, as delineated from satellite imagery; (ii) the distribution of public vs. Private land ownership, as assembled from various land information systems; and (iii) the distribution of terrestrial vertebrate species, as predicted from the distribution of vegetation.
The CURBA Model makes use of second-generation GAP Analysis, as prepared at the University of California, Santa Barbara. Because of the California's size and complexity, the UCSB GAP classification system incorporates both Jepson eco-regions ( Hickman 1993) and Holland vegetation classification zones. Image interpretation is being guided by vector overlays of existing vegetation maps, land use maps and forest inventory data. Upland types are being mapped with a minimum mapping unit of 100 hectares. Major wetland areas are mapped using a 40 hectare minimum mapping unit. The California Wildlife-Habitat Relationships System (WHR), in conjunction with digital species range maps, is applied to the vegetation map to predict the current distribution of potential habitat for each native terrestrial vertebrate species (570 species).
GAP Analysis data was obtained from UCSB in polygon form, and then grided into one-hectare cells. Note that this "re-griding" suggests a level of classification accuracy that is not present in the original data.
3. Slope and elevation data: USGS Digital Elevation Model (DEM) data are used to generate one-hectare slope and elevation grid cells. These data are used in both the Urban Growth and Policy Simulation Models.
4. Locations and types of roads, hydrographic features, and jurisdictional boundaries. U.S. Census TIGER files are used to identify major roads, hydrographic features, and city and place boundaries. ArcInfo was used to generate one-hectare distance grids to major highways, rivers and streams, and city and sphere-of-influence boundaries. These grids as well as the original line features are used in both the Urban Growth and Policy Simulation Models.
5. Wetlands and floodzones: Digital maps of wetland areas and floodzones were obtained from the National Wetlands Inventory and the Federal Emergency Management Agency, respectively, and then grided. The floodzone data is used in both the Urban Growth and Policy Simulation Models; the wetlands data is used solely in the Policy Simulation Model.
6. Jurisdictional spheres-of-influence (S-O-I). Sphere-of-influence boundaries(4) were obtained in paper and/or digital form from Local Agency Formation Commissions, and then digitized or imported. ArcInfo was used to generate one-hectare S-O-I distance grids, which are used in both the Urban Growth and Policy Simulation Models.
7. Various socio-economic data. Population and employment counts by jurisdiction were obtained from the California Department of Finance and the California Employment Development Department, respectively. Historical population and employment estimates are used in the Urban Growth Model; population projections are used in the Policy Simulation and Evaluation Model.
CURBA Model datasets are organized and accessed by county. Future versions of the CURBA Model will allow for city or jurisdictional-level analysis.
The Urban Growth Model
The Urban Growth Models consists of one or more logit equations comparing observed changes in urbanized land to a variety of spatial and non-spatial factors(5). These equations take the following general form:
Prob [undeveloped grid-cell i is urbanized between 1986 and 1994] = f { grid-cell proximity to highways facilities, slope and other natural constraints to development, proximity to jurisdictional boundaries, local growth policies, recent population and job growth}
Separate logit equations are estimated for each county. Hectare-scale urbanization is identified from the California Farmland Mapping and Monitoring Project database. The ability of the logit models to explain historical urbanization changes varies by county from a low of 86% for San Joaquin County, to a high of 96% for El Dorado County. Table 1 shows the results of two logit models of urbanization in Santa Cruz County.
Once the various logit models have been estimated and checked, the resulting coefficients are used to calculate future urbanization probabilities for all remaining undeveloped grid-cells. These probabilities range from a high of 1 (indicating development is certain) to 0 (indicating the impossibility of future development). The calculated urbanization probability grid is then exported in ArcView format.
The Policy Simulation and Evaluation Model
The Policy Simulation and Evaluation Model consists of a series of ArcView commands and scripts and is designed to make it easy for users to simulate and then evaluate multiple local growth policies. The first step in constructing and evaluating a growth policy scenario is to import and display the calculated urbanization grid.
The second step is to enter the increment of population growth to be allocated to suitable grid-cells, and the minimum allocation density, in persons per hectare(6). Population growth projections are obtained from the California Department of Finance. Gross development densities are estimated by dividing current county population levels by the number of one-hectare grid-cells identified as urbanized. Unless otherwise specified, projected population growth can be allocated to suitable grid-cells anywhere within a subject county.
Policy scenarios are constructed by selecting particular development constraints or parameters (Figure 2). Users can identify whether particular grid-cells are to be precluded from development regardless of their development probability score on the basis of their slope; whether they are in a designated wetland or floodzone; their agricultural class; their proximity to a river, stream, highway or road; and their proximity to particular jurisdictional boundaries, including city limits and sphere-of-influence lines. The effect of precluding a grid-cell from development is to shift the population growth it might otherwise have been allocated to another, lower-scoring grid-cell. (Policy scenarios can also be constructed by by specifying alternative allocation densities.) Once a particular policy scenario has been fully specified, the CURBA Model displays a summary map of developable and undevelopable grid-cells.
The next step is for the CURBA Model to allocate projected population growth to the remaining developable sites. Development is allocated to sites in order of their calculated development probability, subject to specified policy constraints. The allocation process proceeds until all required sites have been developed, or until no more sites are available, or until some user-specified minimum development probability threshold has been reached. Once the allocation process has been completed, the CURBA Model reports the average allocation density and displays a map showing the resulting growth allocation.
The resulting growth allocation is then compared with various habitat designations, as present in the Gap Analysis habitat layers(s). Users can backtrack at any time to modify a policy scenario or specify a new one. The CURBA Model keeps careful track of all policy scenarios and growth allocations
Measuring Habitat Change and Fragmentation
After generating alternative development scenarios, the CURBA Model allows users to analyze the impacts of projected development on habitat fragmentation and thus, indirectly, on habitat quality. In addition to Total Habitat Area, the CURBA Model calculates the following before-and-after habitat fragmentation measures:
1. Percent of Landscape: This is a specific habitat's share of total undeveloped (or landscape) area. It is obtained by dividing total habitat area by total landscape area. Higher values mean that the landscape is composed of fewer distinct habitat types, and, all else being equal, suggest a lower level of biodiversity.
2. Number of Patches: This measure is a count of the number of distinct (non-adjacent) areas, or patches of a particular habitat type. For a given total habitat area, a larger patch count indicates greater habitat fragmentation, and thus reduced habitat quality.
3. Maximum Patch Size (in hectares): This measure is the area of the largest patch of a specific habitat type. All else being equal, larger patches are preferable to smaller ones.
4. Minimum Patch Size (in hectares): This measure is the area of the smallest patch of a specific habitat type. Patches that are too small may lack sufficient food sources to support particular species populations.
5. Mean Patch Size (in hectares): This is the typical, or average patch size for a particular habitat type. It is obtained by dividing total habitat area by the number of patches. Larger mean patch sizes are almost always preferable to smaller ones.
6. Patch Size Variance and Standard Deviation (in hectares): These measure the distribution of habitat patch size. A small variance and standard deviation indicates that distribution of patch sizes clusters around the mean. A large variance and standard deviation indicates a wide variety of patch sizes.
7. Patch Density: This is the number of habitat patches of a particular type per 100 hectares of landscape. As an indicator of habitat quality, lower patch densities are preferable to higher ones.
8. Largest Patch Index: This is the area of the largest patch of a particular type divided by the total landscape area.. An index value close to 1 (or 100%) indicates that most of the landscape is composed of a single habitat patch. Depending on the type of habitat, this may be a positive indicator of habitat quality, but a negative indictor of biodiversity.
9. Total Edge: This is the total (outside) perimeter of patches of a particular habitat type. Higher amounts of edge permit easier movement across habitats types.
10. Average Edge-Area Ratio: This measure is the ratio of total patch edge (or perimeter) to total patch area. Higher edge-area ratios are typically associated with greater patch fragmentation, or with long-and narrow patch shapes.
11. Edge Density: This measure is the ratio of total patch to landscape area.
CURBA Model users can choose to analyze all habitat types, or just selected ones. They can also analyze multiple habitat types according to common vertebrate species.
The extent to which particular habitat fragmentation values are associated with habitat quality and biodiversity varies by area, habitat type, and species. Across the board, much more research is needed into the relationships between habitat fragmentation and the ability of habitats to support stable and diverse species populations.
Santa Cruz Pilot Study
In its first application, the CURBA Model was used to test the effects of three growth policy scenarios on projected urban development patterns and habitat fragmentation in Santa Cruz County for the year 2010. According to the California Department of Finance, the population of Santa Cruz County is projected to increase by approximately 50,000 persons between 1995 and 2010. At the County's current average density of 20 persons per hectare, an additional 2,500 hectares (or approximately 6,250 acres) will be required to accommodate this level of population growth.
Where this growth goes and how it impacts Santa Cruz habitats will depend, in part, on how Santa Cruz County's four incorporated governments(7) choose to regulate private development. For the sake of simplicity, we assumed all five local governments would act in concert on a common set of policies. Three distinct scenarios were tested using the CURBA Model (Table 2):
1. Scenario SC1, entitled, No Constraints, permits urban development to occur just about anywhere in Santa Cruz County, except on wetlands. Urban development is permitted on all types of farmlands, within floodzones, adjacent to rivers and streams, on hillsides of any slope, and outside existing sphere-of-influence boundaries.
2. Scenario SC2, entitled Farmland Protection, assumes the adoption of zoning and other regulatory policies which would preclude the development of prime and unique agricultural lands, as well as farmlands classified as being of importance to the state and local economy. Scenario SC2 would also prohibit development on wetlands. Other hazard areas and environmental resources such as floodzones, riparian zones, and hillsides would be unprotected.
3. Scenario SC3, entitled Environmental Protection, would impose numerous limits on new development throughout Santa Cruz County. Development would be prohibited from occurring on wetlands, within FEMA designated floodzones, on sites with slopes greater than 10%, and within 100 meters of a river stream. Development would also be limited to sites within 500 meters of existing sphere-of-influence boundaries. To further reduce land consumption, Scenario SC3 assumes the adoption of a development density floor of 25 persons per hectare--a level 20% higher than the current countywide average density.
Scenario Results
Map 1 presents baseline information for all three scenarios. The top panel, which is based on the results of the Urban Growth Model, shows the calculated probability that each undeveloped site will be urbanized: dark red sites are the most likely to be urbanized; light blue sites the least likely. The bottom panel of Map 1 shows Santa Cruz County's major habitat zones, according to the Holland classification system. This data was obtained from the GAP Analysis program.
Maps 2 through 4 present the results of the various scenarios. The top panel of each map shows which sites are to be considered developable and undevelopable given the constraints imposed under each scenario (development is allowed in the yellow areas, but precluded from the red ones). The bottom panel presents the growth allocation results for each scenario; existing development is in dark grey; projected new development is in red.
Under Scenario SC1: No Constraints, almost every undeveloped site in Santa Cruz County is considered developable (top panel). As the bottom panel shows, however, projected new development will tend to favor sites at the edges of existing cities, particularly Watsonville. These locations are flat, and are well-served by existing infrastructure, especially regional highways. They are also less likely than more distant sites to arouse political opposition to sprawl.
The effect of adopting policies designed to protect farmland (Map 3: Scenario SC2: Farmland Protection) is to place most of the county's coastal and southern areas off limits to development. The areas east of Watsonville, in particular, which comprise some of California's best farmland, would be protected from development. The effect of these constraints (compared to Scenario SC1) would be to shift more new development northward to the outskirts of Santa Cruz and Scotts Valley, and to the unincorporated areas of Ben Lomand.
Because Santa Cruz County is so hilly, and contains hundreds of miles of stream bed, the effects of limiting development on hillsides and in riparian areas (as well as in prime and unique agricultural areas) is to place most of the county off-limits to development. This is the result shown in the top panel of Map 4, which summarizes the results for Scenario SC3: Environmental Protection. The effect of these constraints is to shift projected new development to those few remaining areas judged not to be the lest environmentally sensitive. This includes areas surrounding the city of Scotts Valley, areas to the northwest of Santa Cruz, and a few "infill" sites north of Watsonville. Thus ironically, one of the primary effects of adopting policies designed to protect the environment would be to shift much of the county's prospective growth to Scott's Valley, a city known for its small-town, environmentally-friendly character.
We note that these are scenarios, not forecasts. The extent to which the development patterns we have identified under the various scenarios would ultimately depend not on which conservation programs and regulations are adopted, but on how those regulations are administered.
Habitat Loss and Fragmentation
Regardless of which of the three scenarios are pursued, the habitat types likely to be most affected by projected urban growth are Agriculture and Upland Redwood Forest (Table 3). Under Scenario SC1: No Constraints, projected urban growth would consume 902 hectares of Agricultural habitat and 405 hectares of Upland Redwood Forest habitat. Under Scenario SC2: Farmland Protection, the Agricultural habitat losses would decline to 447 hectares, while Upland Redwood Forest habitat losses would rise to 620 hectares. Agricultural habitat losses would decline somewhat further under Scenario SC3: Environmental Protection, while the loss of Upland Redwood Forest would increase to 1,232 hectares. On a percentage basis the loss of Agricultural habitat would range from a high of 4.4% under Scenario SC1, to a low of 1.8% under Scenario SC3. The maximum percentage loss of Upland Redwood Forest, 2.5%, would occur under Scenario SC3.
Two other habitat types, Non-native Grassland, and Mixed Evergreen Forest, would be substantially diminished under Scenario SC3: Environmental Protection. Non-native Grasslands would decline by 145 hectares, or 21.6%, while Mixed Evergreen Forest habitat would decline by 282 hectares, or 3.7%.
More significant than the issue of habitat loss is the issue of habitat fragmentation. Table 4 presents various fragmentation change measures for Agricultural and Upland Redwood Forest habitats for the three scenarios.
With respect to Agricultural habitats, it is Scenario SC2: Farmland Protection, which, surprisingly, results in the most additional fragmentation: compared to their initial 1994 level the number of individual habitat patches increases the most, while average patch size declines the most. Scenario SC2 also produces the highest patch density. Across the board, Scenario SC3: Environmental Protection results in a somewhat lower level of Agricultural habitat fragmentation than Scenario SC2. The reason for this surprising result is that Agricultural habitats are much more extensive than just agricultural lands. Thus, the preservation of agricultural lands under Scenario SC2 does not result in a comparable conservation of Agricultural habitat quality(8).
The results for Upland Redwood Forest habitat are also surprising. Of the three scenarios, it is the one that is the least environmentally friendly, Scenario SC1: No Constraints, that consumes the least additional habitat and results in the least additional fragmentation. The most injurious scenario, ironically, is Scenario SC3: Environmental Protection. Scenario SC3 shifts growth inland from coastal hillsides, thereby resulting in greater Upland Redwood Forest habitat loss and fragmentation. Scenario SC1, conversely, preserves Redwood habitat at the expense of Agricultural habitat.
The various habitat results all point to an interesting and important policy conclusion: the adoption and implementation policies designed to protect and conserve environmental features such as riparian corridors, hillsides, and/or farmland, will not automatically--and not in all places--result in the appropriate habitat conservation.
Conclusions and Caveats
The California Urban and Biodiversity Analysis Model represents a huge step forward in the ability of policy-makers and planners to project and evaluate the possible effects of alternative urban growth patterns and policies on natural habitat quality and biodiversity. The CURBA Model achieves significant advances on three fronts. First, it allows planners, policy-makers, interest groups, and residents to better understand the forces and factors behind recent urbanization trends and patterns. Second, it allows them to easily project future urban growth patterns, and to investigate the sensitivity of projected urban growth patterns to alternative regulatory and environmental policies. Lastly, by making use of here-to-fore difficult-to-use GAP Analysis data, it allows policy-makers, urban and environmental planners, wildlife ecologists, natural scientists, and everyone else concerned with the future of the natural environment to constructively evaluate the effects of projected urban growth on habitat integrity and quality.
The CURBA Model also demonstrates the incredible amount of spatial data and useful analytical power it is now possible to put on a desktop. The Policy Simulation and Evaluation component of the CURBA Model runs entirely in ArcView, a powerful and flexible mapping program which, as this article makes clear, can also serve as a robust, beyond-the-state-of-the-art analytical and simulation tool. A typical run of the CURBA Model makes use of a dozen grid layers, each of which commonly includes a million-plus hectare grid cells. Yet running the CURBA Model--including generating reports-- typically takes less than 5 minutes per scenario.
Properly using the CURBA Model, however, requires understanding its limitations. First and foremost, the model results are only as good as the quality of the underlying data. To the extent that the data is mis-classified, or that map feature boundaries which are supposed to line-up do not (particularly between different map layers), the model is likely to produce erroneous and biased results. This is especially true when analyzing inter-scenario differences in habitat fragmentation.
The probability scores generated by the Urban Growth Model component are a second possible source of bias. To the extent that the model equations do a poor job explaining historical urban growth patterns, or to the extent that the factors driving future development patterns differ from those of the past, the CURBA Model may direct future development to some unlikely locations.
Third, the forward effects of regulatory constraints on development are notoriously difficult to predict. Simply removing inappropriate sites from consideration for development greatly oversimplifies how real-world land and development markets react to regulatory constraints. Nor is the model in its current form able to project the likely effects of new investments (such as roads) on future development patterns.
Fourth, the model treats all urban development as homogeneous, and does not distinguish between different land uses, or allow for the possibility of redevelopment. Nor is the model in its current form able to simulate a variety of development densities.
Lastly, users should remember that although related, habitat quality and biodiversity are not the same thing. Having large amounts of contiguous habitat is a necessary condition for species health and biodiversity, but it is not sufficient. Other factors, including ones which are completely outside the realm of control of humankind are sometimes at work.
References
J. C. Hickman, editor; 1993. The Jepson Manual of Higher Plants of California. Berkeley: University of California Press, 1993.
Landis, John D. and Ming Zhang. 1998. The Second Generation of the California Urban Futures Model. forthcoming in two issues of Environment and Planning B.
Landis, John D., Juan Pablo Monzon, Michael Reilly, and Chris Cogan. 1998. The California Urban and Biodiversity Analysis Model: Theory and Pilot Implementation. UC Berkeley: Institute of Urban and Regional Development. forthcoming.
Scott, J.M., F. Davis, B. Csuti, R. Noss, B. Butterfield, C. Groves, H. Anderson,S. Caicco, F. D'erchia, T.C. Edwards Jr., U. Ulliman, and R.G. Wright. 1993. Gap Analysis: A Geographic Approach to Protection of Biological Diversity. U.S. National Biological Service: Wildlife. Monograph No.123.
U.S. Bureau of the Census, 1990. 1990 Census of Population and Housing, CPH-S-1-2, Supplementary Reports: Urbanized Area of the Unites States and Puerto Rico, Section 1 of 2.Washington, D.C.: U.S. Government Printing Office.
U.S. Environmental Protection Agency. 1996. National Air Quality and Emissions Trends Report, 1995. Washington, D.C.: U.S. Government Printing Office. Pp. 11-12.
Notes
1. The authors are respectively, associate professor of City and Regional Planning at the University of California Berkeley; planner and GIS analyst for the city of Vallejo; Ph.D. student in the Department of City and Regional Planning at UC Berkeley; and Ph.D. student in the Department of resource Ecology at the University of California-Santa Cruz. We also wish to recognize the contributions of Bruce Goldstein, Ph.D. student in the Department of City and Regional Planning at UC Berkeley. Development of the CURBA Model was funded by the National Biological Service of the USGS.
2. There are no reliable estimates of habitat loss during any period. Most of the nation's habitat loss prior to World War II , however, is believe to have occurred as a result of agricultural cultivation, and resource extraction and harvesting.
3. Federal habitat protection is limited to certain federal lands (e.g., national parks) and to species covered under the Endangered Species Act. Recent federal initiatives to more comprehensively preserve habitat through the designation of Habitat Conservation Planning Area rest on the full and willing participation of local governments, state agencies, and private landowners.
4. Spheres-of-influence demarcate local area planning boundaries. Set at the county level by Local Agency Formation Commissions, or LAFCOs, they are intended to indicate planned "build-out" boundaries.
5. The assumptions behind the use of the logit estimator to model changes in land use are explained in greater detail in Land is and Zhang, 1998.
6. The CURBA Model allocates urban development based on projected population growth and average population density. The Model does not differentiate between different land use types. Commercial, industrial, and other urban land uses are all subsumed in the average population density estimate.
7. Santa Cruz County, and the cities of Santa Cruz, Watsonville, and Scotts Valley.
8. This result may also be the result of discrepancies between the GAP and FMMP layers.