Modeling Future Growth in the Washington, DC-Baltimore Region 1986-2030

The Washington, DC – Baltimore Metropolitan Region (Figure 1) covers ten percent of the Chesapeake Bay watershed and includes over forty percent of its total population.  The decisions we make in this region about lifestyle, land use, transportation, and resource conservation affect the water quality and restoration of the Chesapeake Bay.

Figure 1: Study area

The main goal of this project was to create a predictive modeling system capable of depicting, quantitatively and graphically, the growth impacts of various land use or land management policies and trends in the Washington, DC-Baltimore, DC area.  This study area provides an opportunity to model exurban sprawl in a rapidly changing region.  Current trends in policy and public debate in the region are focused on growth control measures broadly referred to as “Smart Growth,” or policies aimed at curbing the detrimental effects of sprawl development.  This investigation can help facilitate this debate since it simulates the impacts of alternative land use controls and mitigation strategies on several dimensions of environmental quality and ecosystem health.  Given its success with regional scale simulation, its ability to incorporate different levels of land protection through an “excluded” layer, and the relative ease of implementation and computation, the model developed by Clarke, Hoppen et al. (1997), known as SLEUTH , was adopted for this project.  Three alternative growth scenarios were modeled: (1) current trends, (2) managed growth with minimum protection placed on resource lands, such as riparian areas, wetlands, forests, and agricultural areas, and (3) managed growth with maximum protection.  The model was focused on a forty-five year time period.  It was calibrated based on analyses of growth and land use data between 1986 and 2000, using a time series of Landsat Thematic Mapper (TM) imagery, and growth was projected up to 2030.

The SLEUTH Model

SLEUTH belongs to the cellular automata class of models, so the study area is represented as a regular grid of cells and each cell has only two states: urbanized or non-urbanized.  The current version of SLEUTH is not capable of modeling density of development within a pixel.  Whether or not a cell will become urbanized is determined by four growth rules, each of which attempts to simulate a particular aspect of the development process.  In their seminal application of the SLEUTH model in the San Francisco Bay area, Clarke, Hoppen et al. (1997) stress the utility of the model in simulating historic change, the description of which can aid in the explanation of growth processes at a regional scale, and in predicting future urbanization patterns.  The model was successful in simulating change between 1900-1990 for the San Francisco area.  Recent applications of the SLEUTH model are taking place in Santa Barbara, CA, the Middle Rio Grande River basin in New Mexico, Denver, CO metropolitan area, and Chester County, PA.

Implementation of the model occurs in two general phases: calibration, where historic growth patterns are simulated, and prediction, where the historic patterns of growth are projected into the future.  For calibration, the model requires inputs of historic urban extent for at least four time periods, a historic transportation network for at least two time periods, slope, and a data layer of non-developable land, or excluded layer.  The excluded layer contains probabilities of exclusion; areas that should be completely excluded from development, such as water, are given a value 100, while areas open to development are given lower probabilities of exclusion.  Based on these inputs, the model is calibrated and the results are used for predicting future urbanized extent.

SLEUTH simulates four types of growth, which are applied sequentially during each growth cycle :

1. Spontaneous new growth, which simulates the random urbanization of land,
2. New spreading centers, which simulates the development of new urban areas,
3. Edge growth, which stems from existing urban centers,
4. Road influenced growth, which simulates the influence of the transportation network on development patterns.

1. Slope, the resistance of an area to development due to terrain,
2. Dispersion, the random urbanization of single pixels,
3. Breed, the likelihood that spontaneous growth will spawn new urban centers,
4. Spread, old or new urban centers spawning additional growth, and
5. Road gravity, the ability of roads to attract new growth.

Methods: Preparation of Historic Data

For calibration, SLEUTH requires at least four time steps for urban extent.  Using new techniques to map impervious surfaces from Landsat TM imagery, continuous maps of impervious surfaces for 1986, 1990, 1996 and 2000 were produced.  The original data were at a resolution of 30 meters, which produced an array that exceeded the available computational resources.  The data were therefore re-sampled to a resolution of 45 meters to decrease the size of the array while maintaining the spatial extent of the study area.  Because SLEUTH requires a binary representation of urban extent, these continuous data were transformed into binary maps of development extent using a threshold of 10%.  Click here to see images of urban growth between 1986 and 2000.

Two time steps for transportation were prepared.  Roads layers for 1986 and 1996 (Figure 2) were developed using the primary road network defined in the 1:100,000 scale US Geological Survey (USGS) digital line graphs (DLGs).  A USGS 7.5 minute digital elevation model was used to create an input layer for slope (Figure 3).  SLEUTH also requires an excluded layer for calibration (Figure 4).  For the calibration phase, the excluded layer consisted of water, which was 100% excluded from development, as well as federal, state, and local parks, which were 80% excluded from development.  This 80% level of exclusion was used since limited development within many of the parks had occurred in the historic time period.  All input files were rasterized at a 45-meter resolution to the extent of the study area and checked for overlay accuracy.

Figure 2: Roads layers used in calibration

Figure 3: Slope

Figure 4: Excluded layer used in calibration

The goal of calibration is to derive a set of values for the growth parameters that can effectively simulate growth during the historic time period, in this case 1986-2000.  This is achieved in the SLEUTH modeling environment through a brute force Monte Carlo method, where the user indicates a certain range of values and the model iterates using every possible combination of parameters.  For each set of parameters, simulated growth is compared to actual growth by several least squares regression statistics, such as the number of urban pixels, urban cluster edge pixels, the number and size of urban clusters, and spatial match.

Methods: Prediction

The set of coefficients derived during calibration are used to predict future patterns of urbanization.  For prediction, SLEUTH requires the following inputs: urban extent for initialization, an initial transportation network (subsequent future networks can be incorporated on user specified dates), an excluded layer, slope, and a hillshade, or background, image.

Three future scenarios were modeled: current trends, managed growth with moderate protection and managed growth with maximum protection.  The excluded layer served as the primary instrument to differentiate between the three policy scenarios, but different future transportation networks were also created and incorporated into the model in 2010.  In addition, the input image of urban extent was altered to include future planned developments in the current trends scenario.

The current trends scenario reflects policies that are currently in place.  The excluded layer (Figure 4) serves to model protection policies applied to different lands.  All parks and easements were fully protected from development.  Partial protection was given to large, contiguous wetlands and riparian buffer strips along major streams.  Preservation of land adjacent to tidal waters was also included.  Keeping with current policies, a slightly higher protection was applied to the tidal buffer in Maryland than in Virginia.  In Maryland, land outside the state-designated Priority Funding Areas was given minimal protection.  Major new planned roads and road widenings (Figure 5) identified by the CBF were incorporated. Areas of development (Figure 6) that are planned or that were in early stages of development in 2000 were also included in this scenario.  The approximate location, size and density of these developments were identified by the CBF and then random points were distributed within these areas at varying densities.  These points were rasterized and incorporated into the 2000 image that initialized the prediction.

Figure 4: Excluded layer used in current trends scenario

Figure 5: Roads used in each predictive scenario

Figure 6: Areas of new development for current trends scenario

The managed growth (moderate protection) scenario reflects a stronger commitment to focused growth and resource protection.  In the excluded layer (Figure 7), higher levels of protection were assigned to wetlands, riparian buffer strips, and the tidal buffer.  This protection was extended to include all wetlands that are larger than 0.5 acres, a more extensive stream network and a wider tidal buffer.  New “smart growth areas” (SGAs) were developed for both Maryland and Virginia to limit growth outside of established urban centers.  Protection of forest and agriculture was also incorporated into this scenario.  The transportation network and the input image for urban extent also reflect a commitment to focused growth; no new roads were included in the future transportation network and no new major planned developments were added.

Figure 7: Excluded layer for managed growth scenario (moderate protection)

The third scenario, managed growth with maximum protection, reflects a more extreme set of policies targeted toward limited growth and natural resource protection.  The data elements for the excluded layer (Figure 8 ) were similar to those included in the previous scenario, but the levels of protection were increased.  In addition, riparian areas were augmented to include a larger buffer and the headwater streams.  Like the previous scenario, no new roads were included in the transportation network, and no new major planned developments were added.

Figure 8: Excluded layer for managed growth scenario (maximum protection)

Results: Trends in Urban Development 1986-2000

Growth rates were highest between 1986 and 1990 and between 1996 and 2000 (Figure 9).  Click here to see images of urban growth between 1986 and 2000.  It is interesting to note the differences in the spatial pattern of the development that occurred during these times.  For example, new growth that appears in 1990 seems to be concentrated around existing urban and suburban centers.  Although some infill development also occurs between 1996 and 2000, the majority of new development during this time seems to be low-density residential development.  These patterns are evident in traditional rural counties, such as Frederick County, MD and Loudoun County, VA, as well as in previously undeveloped areas within more urbanized counties, such as Fairfax County, VA and Montgomery County, MD.

Results: Predictions and Impact Assessment

The SLEUTH model produces annual images that show the probability of any given cell becoming urbanized.  To see images of how the region could change over time, follow the links below.  The highly dispersed development patterns for the current trends scenario are striking compared to the managed growth scenarios, while the managed growth with maximum protection scenario shows highly constrained growth over the whole region, with most growth occurring in and around existing urban centers.

1. Current trends
2. Managed growth (moderate protection)
3. Managed growth (maximum protection)

Figure 10: Current Trends Impact Assessment

Figure 11: Managed Growth (Moderate Protection) Impact Assessment

Figure 12: Managed Growth (Maximum Protection) Impact Assessment

Conclusions

The SLEUTH model was found to be a useful tool in visualizing alternate future scenarios.  Existing growth management and conservation policies were incorporated into the current trends scenario, which also reflects a continuation of the rapid growth rates that were observed between 1986 and 2000.  Although the predicted spatial pattern of development may be variable, the magnitude of land use change is comparable to the changes that have occurred in the recent past.  The results for this scenario are particularly salient to public discussion since they demonstrate the potential losses in resource lands that would occur if the observed rates of land use change were to continue into the future.  While the results for any specific area may not be precise, the results for the region represent a reasonable estimate of the amount of growth that could occur.

The two scenarios that depict the impact of various managed growth policies are also useful in showing the potential reductions in loss, but should be carefully considered in light of the capabilities of the SLEUTH model.  Although the SLEUTH model is dynamic to the extent that it is self-modifying, the current version is not able to redirect growth pressure.  This makes the potential impact of stringent resource conservation policies difficult to simulate.  In a similar application of the California Urban Futures model, for example, Landis (1995) found that strong protections placed on natural resource lands drove sprawl development into outlying areas.  Since SLEUTH lacks the capability to redistribute growth pressure, a discovery of this kind of unanticipated impact would not occur.  This is a potential limitation on the knowledge that can be gained from SLEUTH in its present form.

Additional Information

• For more information about this project, see the Chesapeake Bay Foundation website at http://www.savethebay.org/land/landuse/maps/future_growth.html, or contact Dr. Scott Goetz or Claire Jantz at the Mid-Atlantic RESAC.
• For more information about the SLEUTH model, see the Project Gigalopolis at http://www.ncgia.ucsb.edu/projects/gig/.

Funding for this project was provided by the Chesapeake Bay Foundation (CBF) and the NASA Land Cover Land Use Change Program.  Steve Libbey and Lee Epstein at the CBF were instrumental in the development of the scenarios.  Collaborators at the U.S. Geological Survey should be acknowledged for their expertise and support: Janet Tilley, Jeannette Candau, Mark Feller and Dave Hester.  Preliminary calibration runs were performed using computing resources provided by Janet Tilley at the U.S. Geological Survey offices in Reston, VA, and the final calibration was run at the Rocky Mountain Mapping Center in Denver, CO under the supervision of Mark Feller.

References

Clarke, K. C., S. Hoppen, et al. (1997). "A Self-modifying Cellular Automaton Model of Historical Urbanization in the San Francisco Bay Area." Environment and Planning B: Planning and Design 24: 247-261.

Jantz, C. A., S. J. Goetz, and M. A. Shelley. Using the SLEUTH urban growth model to simulate the land use impacts of policy scenarios in the Baltimore-Washington metropolitan region. Environment and Planning (in press).

Landis, J. (1995). "Imagining Land Use Futures: Applying the California Urban Futures Model." Journal of the American Planning Association 61(4): 438-457.

Smith, A. J., S. J. Goetz, et al. (forthcoming). "The Application of Subpixel Impervious Surface Algorithms in Urban Planning." Remote Sensing of the Environment.

US Geological Survey (2000). National Land Cover Data, Earth Resources Observation System (EROS) Data Center Land Cover Characterization Program.