FINDING THE APPALACHIAN SCENIC CORRIDOR

From its inception, the ASCS was conceived as a project which would rely on GIS as the centerpiece for all project related activities, including:

• Guiding the initial transportation corridor identification process
• Performing impact assessment analyses for transportation corridor alternatives
• Quantifying and describing corridor alternatives
• Preparing public information exhibits and demonstrations
• Generating tables and figures for reports

This paper outlines how the Esri family of GIS software products was used to address each of these activities.

Ideally, a transportation corridor between any two points should take the shortest route possible which also minimizes adverse impacts to natural resources, cultural features, and human populations. The best routes should also avoid areas of difficult terrain, crossings of large bodies of water, and other adverse physical features. Finally, the ideal route should follow those areas of existing road which are suitable for inclusion in the transportation corridor. Evaluating all these constraints simultaneously is difficult for human beings, but much easier for a computer. Phase I of the ASCS project used the ArcInfo GRID modeling technique called Least-Cost Path Analysis to help the Project Team identify potential transportation corridors. The methodology included the following steps:

• Preparing coverages on data themes for physical, ecological or cultural constraints that influence transportation corridors in north Georgia.
• Rank the features in each data theme according to corridor suitability.
• Overlay all data themes in GIS to produce a single "cost-surface" map.
• Use the cost-surface map to find least-cost path corridors between logical termini.
• Weight input data themes for different corridor types, and find new least-cost paths.

In order to generate a "cost-surface" or combined constraints map, individual maps representing each of the input criteria of interest were individually ranked on a common scale. The ranking criteria developed for each data theme were based on a scale of 1 to 5. Such a scale provided sufficient refinement for capturing shades of difference in any given data theme, yet kept the number of categories to a number which could be easily interpreted and understood. The five levels could be thought of like this:

The table illustrates that any data layer could be thought of in terms of either constraints or suitability. Constraints were those things that, when present, would be bad for transportation corridor development. The lower the constraints score, the fewer the obstacles to development. Suitability was the converse; areas with a low rank (i.e., 1 or 2) had conditions that were very suitable for transportation corridor development. The terms suitability and constraints are largely interchangeable, as long as one realizes that the rankings run from very low to very high for constraints, but from very high to very low for suitability. Table 2 lists the input themes to the least-cost path model, and shows how attributes in each theme were ranked in terms of transportation corridor constraints and/or suitability.

To evaluate the effect of alternative weighting schemes, six cost-surface coverages were generated, each representing the results of different weighting schemes applied to the input coverages. The weighting scenarios were identified as follows:

• Baseline
• Roadway
• Parkway
• Environmental
• Rail
• Engineering

The Baseline model represented the average of all input layers, each treated as if it had the same importance or weight. The other models varied the weights on the input layers so as to emphasize (or de-emphasize) a given set of constraints or suitability factors which together would suggest transportation corridors sensitive to a certain set of requirements. The weights for each alternative model appear in Table 3.

The ArcInfo least-cost path technique allowed the identification of best routes across a given cost surface for many origin points to a single destination. The Project Team identified a series of points which represented the maximum number of logical termini which might conceivably be connected for an east-west trans-Appalachian corridor. These points fell along the break-lines for each of the four study sections into which the project area had been subdivided.

The least-cost path analysis was conducted for the set of all termini on the east side of each section as origin points, leading to each of the termini on the western end as destinations. (The use of the terms "origin" and "destination" here do not having any meaning from a transportation standpoint, but only reflect the direction in which the computer evaluated the least-cost path. The designation of origin and destination could have been switched, and the results would have been the same.) Each least-cost path run for a given terminus was saved as a separate GRID coverage. After all GRID coverages were complete, the least-cost paths were converted to vector lines using the GRIDLINE command. These lines were plotted on 1:100,000 scale maps of each section (along with color-coded maps of the cost-surface coverage) for evaluation by the Project Team.

The next step in the process required the further reduction of the corridor links. Two GIS methodologies were used to aid the Project Team in evaluating the links. First, the original least-cost paths were combined in GRID to generate a coverage representing the sum of all input coverages. The resulting coverage was used to generate a color-coded plot of paths based on the number of times each path was contributed by an input coverage. Those paths that occurred frequently for multiple pairs of termini and/or from multiple alternative weighting models were deemed high value corridors. This "popularity analysis" was used by the Project Team to evaluate the desirability of functionally equivalent corridors. For example, if two sets of links connected the same set of roadways, then the links exhibiting the greater popularity scores would be preferred.

By the end of Phase I, the Project Team reduced the numerous potential corridors suggested by the least-cost path analysis to less than ten for further study in Phase II. In Phase I, the type of transportation facility (e.g., multi-lane divided highway, or two-lane scenic parkway) did not figure prominently in the corridor evaluation. Minimizing gross environmental and cultural impacts and avoiding engineering difficulties dominated the Phase I evaluation. However, during Phase II the type of transportation facility under consideration had to be factored into the analysis. For the purpose of evaluating Phase II corridors, the Project Team identified three types of routes: (1) Those suitable for "Parkway" concepts, (2) those suitable for "Roadway" concepts, and (3) those suitable for both. Those parts of the Appalachian Scenic Corridor evaluated for the Parkway model would be largely two lanes, with occasional three-lane sections for passing or climbing. The Roadway model presumed a four-lane facility, possibly with a center median. Routes for both models usually followed existing roadways, but would occasionally deviate to form a more direct route for the Roadway alternative.

Throughout all three Phases of the ASCS project, ArcInfo was used to quantify resources within the various corridors being evaluated. These data were generated using the ArcInfo INTERSECT command between coverages containing the corridor alternatives and the resource coverages. For example, the number of acres of potential listed species habitat within each Phase I corridor was calculated by overlaying the corridors on the appropriate coverage. Results were reported by corridor link, which allowed for the impacts of individual links to be compared. This sometimes led to selection of one functionally equivalent link over another as part of the initial corridor identification process. However, most quantitative analysis of corridors occurred during Phases II and III, when the corridors under study had been reduced to a more manageable number and more detailed data were available for each corridor.

Quantification results were often expressed in units of acres per linear mile of corridor, to adjust for the widely varying lengths of the corridor links. Link lengths were obtained by performing an IDENTITY between the corridors (polygon coverages) and the nominal centerlines for those corridors (line coverages). The length of the centerline within a corridor link was taken as the link length. Keep in mind that the "nominal centerlines" were identified for the purpose of evaluating wide corridors (0.5 miles wide or larger), and were not "alignments" in the sense that they precisely located the final route of any future transportation facility. That is, the nominal centerlines were identified merely as a reference point for automated generation of corridors measuring hundreds of feet across using the ArcInfo BUFFER command. Identification of final roadway alignments was outside the scope of this study.

For the purpose of quantitatively comparing alternative corridors in Phase II, the Project Team counted the resources that fell within 150 feet on either side of the Parkway corridor centerlines, and within 300 feet of the Roadway centerlines. It is important to note that this analysis was not intended to evaluate the actual potential impacts of future transportation facilities on resources. Instead, this process provided a means to count the resources encountered along a "likely" or "typical" corridor, and thereby allow for objective comparisons of alternatives. The 300 foot and 600 foot buffer widths for Parkway and Roadway, respectively, were roughly twice as wide as the nominal rights-of-way (ROW) which these alternatives would actually require. The wide buffers ensured that a conservative assessment of resources was made. If the buffers had been set equal to actual ROW requirements, important resources just outside the buffer areas might have gone uncounted.

Before beginning the Phase II GIS quantification effort, the ASCS Project Team decided on which factors to analyze. Most of these factors constituted cultural or natural resources that we wished to avoid in our identification of corridors suitable for further analysis in Phase III. Examples include historic sites, archeological sites, streams, wetlands, parks and forested areas. Interpretation of other factors, such as the amount of urban land or percentage of multilane highway, sometimes depended on the type of facility considered along a particular corridor. For example, a large amount of multilane road in a corridor link would be good for inclusion in a Roadway concept, and less desirable in a Parkway concept. The GIS factors analyzed during Phase II appear in Table 4.

Once the Project Team identified which corridor links comprised the Parkway and Roadway corridor alternatives, the resulting corridor coverages were overlaid using the INTERSECT function on each of the resource coverages listed in Table 4. The INTERSECT command worked like a digital cookie cutter. For example, when the Parkway buffer coverage was combined with the GDOT stream hydrology coverage using INTERSECT, the resulting output coverage contained only those stream reaches within the limits of the Parkway buffer. Furthermore, each stream reach automatically received the appropriate corridor link designation for its part of the corridor. The process was essentially the same for all other line, point and polygon coverages.

The FREQUENCY summarized the amount of each resource by type and link designation. To use the stream example again, each corridor link contained tens or hundreds of stream reaches. Some of those reaches were designated "Major," some "Minor" and some "Channel." Using FREQUENCY, ArcInfo automatically summed the total length of all stream reaches within each link that were attributed as "Major," "Minor" or "Channel." This frequency analysis was output to an ArcInfo INFO file, which was later exported into a Comma Separated Value (CSV) format readable by Microsoft Excel. The same procedure was used for all other linear features. For polygon features, the data were summed according to area (measured in square feet); for point data, the total count of each resource type by link was calculated. Roadway and Parkway statistics were calculated independently, and produced distinct CSV files.

The CSV files for each resource from a given alternative (i.e., Parkway or Roadway) were imported into an MS Excel spreadsheet. Some data from different FREQUENCY output files were imported into a single spreadsheet so that they could be added together, when appropriate. For example, the separate statistics for "Cropland" and "Pasture" from analysis of the Landsat land cover data were brought together in one spreadsheet so that they could be added together as a measure of "Agriculture." The spreadsheets also calculated acreage from the original square foot areal measurements output by ArcInfo.

Since otherwise comparable links might differ in length, a direct comparison of the absolute value of resource statistics was not always meaningful. For example, Parkway alternative Links A03 and A05 were functionally comparable (connecting Interstate 75 to a point roughly midway across Section 1 of the study area), but A05 was over three times as long as A03. The total number of stream reach feet in Link A05 was much larger (50,533) than in link A03 (about 22,000). However, when the number of stream reach feet in the two links was normalized by dividing each by the number of linear miles of corridor in each, it became evident that link A03 had the greater potential to affect stream resources (over 3,000 stream reach feet per mile of corridor for A03, vs. about 2,000 stream feet per corridor mile for link A05).

To perform the "per mile" normalization for all resource data, the Parkway and Roadway buffer coverages were overlaid on their respective centerline coverages, thereby assigning each portion of the centerline files to their appropriate corridor links. Using FREQUENCY, the length of each corridor link was ascertained. Those values were imported into the Roadway and Parkway spreadsheet models, and used to generate the appropriate normalized values: "Feet Per Mile" for linear resources, "Acres Per Mile" for polygon resources and "Sites Per Mile" for point data. The normalized data for roadway lanes were expressed as a percentage of total corridor length, rather than in units of "Feet Per Mile" or "Miles Per Mile." The resulting statistics were ranked in order from worst to best. The corridors at the top of the list thereby represented those with the greatest potential adverse impact; in any comparison of two otherwise functionally equivalent corridors, the one with the greatest potential impact could be identified.

A number of cultural and natural resources factors were analyzed in Phase II as part of the process of identifying the four Phase III Corridors. To further differentiate the four remaining corridors, we identified factors contributing to the cost or benefit of any future road project. For example, the number of stream crossings encountered by a corridor not only provided a measure of relative environmental impact, but allow calculation of relative costs due to the number of bridges it required. Similarly, the number of accidents per mile on existing roads within a corridor provided a means to evaluate the savings in injuries, lives, medical costs and property losses possible through improvements that reduce accident frequency or severity. The GIS factors quantitatively analyzed during Phase III appear in Table 5.

In addition to those factors, we continued to refine the cultural, historic and recreational database developed during Phases I and II with new information. Table 6 lists the data layers created or updated as part of that effort.

This project illustrated the utility of using GIS as the central focus of a transportation planning project. In the initial phase, GIS provided the means to visualize the various constraints and suitability factors that must be balanced in the selection of transportation corridors. The least-cost path analysis provided a tool to assist the Project Team in organizing and evaluating the large amount of data available, and focus on corridors for further consideration. Throughout the project, the ability of GIS to quickly quantify all manner of impacts or corridor characteristics made it invaluable to the Project Team as they sought to differentiate between corridor alternatives. Map products prepared from the same data sets used in GIS analysis were useful to the Project Team as they evaluated corridor alternatives, and were just as valuable in explaining the Team's work to the public. Finally, the interactive use of ArcView in a public informational setting showed that GIS can help to further understanding of complicated issues by people as they contemplate the work of techical experts on matters that affect their community and their way of life.

The following agencies provided spatial data or other assistance to the author which contributed enormously to the success of this project: The Southern Appalachian Assessment (Southern Man And Biosphere project), the Georgia Department of Transportation, the Georgia Department of Natural Resources, the North Georgia Regional Development Center, the Coosa Valley Regional Development Center, and the Georgia Mountains Regional Development Center. The principal members of the ASCS Project Team devoted the time and effort necessary to ensure that the full potential of GIS was realized for this study: David Cheeney, Mark Cheskey, Rusty Ligon and Tom Ziegler (HDR); Janet Harvey and Carl Spinks (Georgia DOT, Office of Planning); Dale Jaeger, Chet Thomas and Amy Kissane (The Jaeger Company); Theresa Hamby (New South Associates); Abraham Lerner (Transcore); and Douglas Bachtel (University of Georgia). Finally, I'd like to thank HDR Project Managers Margaret Ballard and Ken Anderson for their enthusiastic and continued support of GIS to transportation planning in this and other projects.

Michael J. Gilbrook
GIS Manager
HDR Engineering, Inc.
201 S. Orange Avenue, Suite 925
Orlando, FL 32801-3413
Voice (407) 872-7801
FAX (407) 872-0603
Email mgilbroo@hdrinc.com