Peter John LaPlaca

Nature Park Site Analysis in Fairfax County Virginia
Using the ArcView Spatial Analyst Extension


Abstract

Fairfax County, Virginia is experiencing rapid population growth due to its proximity to Washington D.C. As more and more communities develop in the county, it will be important to provide open space for leisure and recreational activities. This paper explores using the ArcView Spatial Analyst Extension to locate a new nature park in Fairfax County. The Spatial Analyst Extension provides the GIS analyst with powerful raster GIS processing that, until recently, was only available with the ArcInfo GRID module.

Using what is known as suitability modeling , cost grids were created based on the physical and human made parameters that people find most attractive in a nature park. Suitability modeling transforms raw data measurements such as elevation into numeric values (1-10) that can be combined mathematically using map algebra expressions. The result was a series of grids containing the same range of values, resulting in an apples-to-apples comparison as opposed to an apples-to-oranges comparison. A final cost grid was derived that identified several areas for a new park.

Introduction

The purpose of this project, an outgrowth of some Ph.D. coursework, was two-fold: to locate a new Nature park in Fairfax County using raster GIS modeling, and to determine the robustness and ease-of-use of grid processing using the Spatial Analyst Extension in comparison to using ArcInfo GRID. All the work was done using ArcView with the Spatial Analyst Extension under the Windows NT operating system.

Raster GIS is a software tool that can display, query, analyze, and output geospatial information. The structure of a raster GIS is based on a grid of some assigned resolution. Each grid square or "cell" contains a numeric value. These values (and their associated attributes) are stored in a database for easy query and retrieval. The values may also represent raw reflectance values, as in the case of satellite imagery, values that represent certain attributes, such as all values of 1 are equal to "coniferous forest," or values that represent actual real world physical phenomenon, such as wind speed or temperature.

The primary use of raster GIS technology is to model discrete (such as vegetation type, etc.) and/or continuous (such as elevation, wind, etc.) data on the surface and in the atmosphere. To do this, raster GIS supports the structure known as map algebra. Map algebra is a high level language used to characterize cartographic spatial analysis using gridded data. This high level language can be used to construct weighted "cost surfaces" for various surface conditions and atmospheric properties. A model may contain many sub models or cost grids. These can be combined and manipulated to determine the optimal location for a nature park.

Since the county has a preponderance of multi-purpose parks that serve many uses, it was decided to model a "Nature Park". The park is envisioned to be a place where people can come to enjoy a natural, quiet setting, away from the near-urban environment of Northern Virginia life. People may come to such a park to fish, canoe, bird and animal watch, walk on nature trails, or visit a nature center.

Study Site

The center of Fairfax County is located approximately fifteen miles west of Washington D.C. The terrain in the county is somewhat flat with some rolling land in the western half of the county. The county is experiencing rapid growth, and can be considered an urban/near-urban environment.

The primary land covers are urban and residential areas surrounding the interstate 495 beltway. Urban or near-urban areas in the county include Fairfax City, Herndon, Reston, Chantilly, Springfield, and Tysons Corner.

There are several water bodies in the county, notably Lake Anne, Thoreau, Audubon, Fairfax, Burke, and Accotink. The county is partially bounded on the south by the Occoquan reservoir and its watershed and on the north by the Potomac River. The western half of the county is less densely populated, made up of residential, forest, and farmland areas.

Modified Delphi Process - Establishing Utility Measures

In order to make the potential park site more realistic and useful to people, a "Nature Park Survey Form" was circulated to a sample of thirty five people. These people were asked to rate the physical and/or human parameters that are most useful in a nature park. The Delphi Process is a method used to assess the usefulness of data. This is accomplished by determining utility and/or setting weights. The results from a modified Delphi survey of thirty-five people (see Appendix A) were tabulated and are presented below.

Near Work Near Home Near an Interstate Near an existing park Near a historic site Near the metro/bus route Near a police station Near a hospital Near a school Boating Fishing Woods Water / River / Stream Water / Lake / Stream Wetlands/

Marsh

Animals
1 2 10 5 2 6 5 8 3 1 8 1 8 5 8 3 7
2 1 9 1 4 9 2 3 3 8 8 6 10 9 9 6 10
3 1 9 1 6 5 1 4 1 6 8 7 10 10 9 9 8
4 6 6 1 1 3 3 3 4 4 8 8 10 10 10 3 9
5 6 10 4 4 4 9 7 3 9 10 5 10 10 10 9 10
6 1 4 3 6 1 4 2 4 6 7 7 9 10 8 9 10
7 1 8 1 6 7 1 7 7 9 10 10 10 10 8 3 10
8 3 7 1 2 1 2 5 2 1 5 5 10 6 9 4 8
9 1 6 2 1 4 2 2 3 4 6 4 8 9 9 1 9
10 4 6 3 5 6 1 4 2 1 10 10 10 10 10 10 10
11 10 10 1 7 7 1 2 5 3 9 9 10 9 9 9 8
12 3 9 1 1 7 1 3 3 1 10 10 10 10 10 7 10
13 1 10 8 8 9 1 4 5 1 2 7 9 9 9 9 10
14 2 8 2 2 2 2 6 2 8 7 7 7 7 7 2 2
15 6 9 8 1 3 6 6 6 8 6 6 10 10 10 5 9
16 3 8 3 3 6 1 2 2 2 6 9 10 9 10 7 6
17 7 9 4 1 1 7 6 2 1 4 4 5 3 3 3 5
18 1 8 4 7 7 6 6 6 7 10 8 10 10 10 10 10
19 6 10 1 6 4 3 1 1 9 6 6 10 10 10 8 10
20 6 9 6 6 9 6 5 5 5 8 8 9 9 9 9 9
21 1 6 1 6 2 2 1 1 8 9 9 10 10 10 10 10
22 2 9 3 3 6 3 7 4 9 5 5 10 10 10 8 10
23 1 10 6 1 1 1 1 6 8 8 8 8 8 8 6 10
24 6 6 1 1 9 7 10 10 10 1 1 10 10 10 10 10
25 1 10 1 1 7 8 6 4 6 10 10 10 8 6 4 10
26 3 10 1 4 8 6 1 6 8 10 10 10 10 10 10 10
27 2 7 9 5 9 9 9 7 8 6 6 10 7 7 5 9
28 1 4 3 1 5 1 1 1 1 6 1 9 8 8 8 8
29 3 7 1 1 1 1 1 1 1 8 10 10 10 10 10 10
30 1 6 1 1 7 3 3 6 1 9 9 10 10 10 9 10
31 6 10 4 6 6 10 1 6 6 3 3 8 6 6 3 3
32 6 10 1 6 1 1 1 1 1 6 6 6 6 6 6 6
33 1 6 1 1 3 8 9 7 1 3 9 9 9 9 3 7
34 4 4 1 8 6 4 4 4 4 9 9 10 10 10 9 10
35 1 10 2 6 6 8 6 7 4 9 3 9 8 9 1 8
Tot 110 280 96 130 178 136 147 140 170 250 236 324 305 306 228 301
Avg 3.14 8.00 2.74 3.71 5.09 3.89 4.20 4.00 4.86 7.14 6.74 9.26 8.71 8.74 6.51 8.60

As one would expect, people seem to agree that woods, animals (habitat), and water are useful and attractive attributes that influence selecting a nature park site. This project focused on aesthetic characteristics to determine a suitable site. Obviously other attributes like land cost, construction cost, and other economic factors are important and would be considered in an actual study.

Data

All of the datasets, with the exception of the USGS DEM elevation dataset, were initially vector ArcInfo coverages. Most of the data were derived from county maps at a scale of 1:4000. The elevation data is a USGS DEM at a scale of 1:250000. All of the data were converted to GRID format, projected to State Plane, with units in feet.

Park Site Analysis

Using the results from the nature park survey, cost surfaces, or preference grids were created to help determine a suitable park site. A total of thirteen cost grids were combined to create the final cost-surface. Several requests in the Spatial Analyst were used. Some of these were Spatial.Distance (same as Euclidean distance in grid), Spatial.CostDistance, Reclassify, and Slice. The utility scale of used was 1=worst to 10=best. Below is a listing of the cost-surface grids along with a brief description of how they were created.

1. Railroad

Because of noise and safety considerations, the park should not be within 300 ft. of a railroad. The railroad cost grid was created by running the Spatial.Distance request. Reclass was run and making a one mile distance NO DATA, thereby excluding railroads from the cost surface calculation. The rest of the grid was classified as a 10, or best.

2. Solid Waste

The park should be at least two miles away from any solid waste site. This would cut down on any noise or smell. The solid waste cost grid was created by running Spatial.Distance, and then the grid was reclassified with a 1 mile area around the solid waste sites as NO DATA so they would not be considered. Everything else was coded as a 10.

3. Existing Parks

The "near an existing park" parameter only scored a 2.74 in the survey. Since most people in the survey seem to agree that they do not want a new park too close to an existing one, Spatial.Distance was run, followed by reclassing the existing parks to NO DATA, and classing a half-mile NO DATA buffer around them. This excluded existing parks and a small half-mile area around them from being considered. The remaining area was given a cell value of 10.

4. Police

From the public facility coverage, Police stations in the county were selected. These were converted to a grid, and Spatial.Distance was run. After this, the grid was reclassified by distance zones based partly on survey results. "Near a Police Station" averaged a score of 4.2 on the survey. Some people felt it was important that the park be near a police station. The police station grid was therefore reclassified as 10 = one mile distance from stations, and 9 = two miles from station, and 6 for the remaining area.

5. Historic Sites (4 protected points)

For historic sites, the four point locations were converted to a grid and Spatial.Distance was run. Because two of the points were "very significant" and two were of "minor significance," the historic sites grid was reclassified to create two NO DATA areas two miles and one mile respectively from these sites. Areas within the 2 and 1 mile area were assigned NO DATA values so that no parks could be built too close to these historic sites. Areas outside the 2 and 1 mile cells were coded a 10.

6. Historic Sites (Already Established)

Existing historic sites were selected from a public facility coverage. "Near a Historic site" scored a 5.09 on the Delphi survey. People seem to feel that it is somewhat important to have a new nature park near an historic site. For that reason, Spatial.Distance was run against existing historic sites, then reclassified one mile and 2 mile areas as 10 and 9 respectively. All other grid cells were coded as 8.

7. Industry (excluded)

Reclassed all industrial areas to NO DATA. Everything else to 10.

We do not want a nature park in an industrial area at all.

8. Street Center Line

Ran Spatial.Distance on the street grid. Then reclassed NO DATA out to 500 ft to account for housing. Cells greater than 500 feet were coded 10.

9. Airports and Government Facilities

Reclassed this grid to the following: Airports, Federal government, and Lorton Prison areas were coded NO DATA. One does not want to build parks here. Everything else got coded 10.

10. Public Facilities

Ran Spatial.Distance on all public facilities cells (points); schools, shopping centers, etc. Reclassed NO DATA out to 300 ft. Everything else coded a 10.

11. Universities

Ran Spatial.Distance on universities selected from the public facilities grid (George Mason and Northern Virginia Community College). Reclassed a one and 5 mile area around universities. Coded 1 mile area 10, 5 mile 9, all other cells 7.

12. Landuse/Landcover

Landuse/landcover was one of the more interesting and important grids to reclassify. Here's how attributes were coded. Survey results whenever possible (for example, woods scored a 9.26 on the survey, so all forest type were coded high).

13. Slope

Slope was created from the USGS DEM and reclassified into the following:
Slope 1 = 9, 1-2 = 10, 3 = 7, 3-8 = 4, and >8 = 1

Final Park Site Cost Grid

After cost grids were created for these thirteen layers, a final park cost grid, divided by thirteen, was created using the following expression in the Map Calculator (see Final Park map)

([Reclass of Distance to Sldwaste] + [Reclass of Exclude Current Parks] + [Railroad Reclass] + [Police Reclass] + [Historic Sites (4 given)] + [Historic Sites already established] + [Industry Excluded] + [Reclass of Distance to street center line] + [Reclass of Airports and Govt, Facilities] + [Reclass of Distance to Pubfacilities (300ft.)] + [Reclass of Universities] + [Reclass of Landuse/Landcover] + [Reclass of Slope of FCDEM]/13

A map of the final suitable nature park site is shown below.

As an additional experiment, an alternative cost model was also produced which weighted the preferences by percentage of certain parameters from the nature park survey form. That cost grid was divided by fifteen, using the following expression in the Map Calculator. The sum of the weights must add up to 100%.

([Reclass of Distance to Sldwaste] + [Reclass of Exclude Current Parks] + [Railroad Reclass] + [Police * .10] + [Historic Sites (4 given)] + [Historic Sites already established * .20] + [Industry Excluded] + [Reclass of Distance to street center line] + [Reclass of Airports and Govt, Facilities] + [Reclass of Distance to Pubfacilities] + [Reclass of Universities] + [Landuse/Landcover (Woods) * .30] + [Landuse/Landcover (Streams) * .15] + [Slope of FCDEM * .10] + [Landuse/Landcover (Lakes) * .15]/15

Conclusion

It is clear from this preliminary project work, that it is possible to use the ArcView Spatial Analyst Extension as a powerful raster analytical tool to determine suitable sites for a nature park.

Like any model, better data and methods yields better results. In an actual study, I would have liked to have good population information, at the district or even subdivision level. Also, some of the datasets were out-dated. The land-use cover comes to mind. Using aerial photo or satellite imagery to update the landuse/landcover in the county would ensure a more accurate model. Also, to fully exploit the Delphi process, a facilitator and several expert(s) would need to be used, instead of me making judgment calls on the utility of certain types of data.

Overall the approach, however, was well grounded and produced good results. It is possible and desirable to use the ArcView Spatial Analyst for park site determination. The power of raster GIS has come full-circle with the addition of raster GIS processing in the ArcView desktop environment.


Appendix A

FROM: Pete LaPlaca
TO: ISG Personnel
SUBJECT: Park Site Survey Form

Dear ISG Colleagues:

As part of my Ph.D. course work, I am using the ArcView 3 Spatial Analyst Extension Geographic Information Systems(GIS) to determine a site for a new "Nature Park" in Fairfax County. For this study, it is important to determine real world information as to what people may find attractive in such a park. For example, you may consider water as an important part of the park, or lots of trees, or nature trails, etc.

Using what is known as suitability modeling (Delphi Process), we can create a series of cost grids based on the physical and human made parameters that people find most attractive in a Nature Park. Suitability modeling transforms raw data measurements such as elevation into numeric values (1-10) that can be combined mathematically using map algebra expressions. The result is a series of grids containing the same range of values, resulting in an apples-to-apples comparison as opposed to an apples-to-oranges comparison. By doing this for example, one can compare a slope grid with a proximity-to-schools grid in the same geographic space.

Your Mission (should you choose to accept it)

Please take a couple of minutes to rate the following parameters from 1 to 10, 1=least attractive, 10 = most attractive, based on your idea of what the ideal Nature Park should be. Thank you for your help.

For example:

            Parameter                         Importance
            Water/Lake or Pond         9   (you feel it's pretty important that your nature park have a lake or pond)
            Water/River or Stream     4   (you feel that a stream/river running through the park is not really important)

Your name (optional)_______________________________

ParameterImportance (1=not important to 10= very important)
Near work
Near home
Near an interstate
Near an existing park
Near a historic site
Near the metro/bus route
Near a police station
Near a hospital
Near a school
Boating
Fishing
Woods
Water/River/Stream
Water/Lake/Stream
Wetlands/Marsh
Animals(existing habitat - squirrels, deer, birds)



Peter John LaPlaca
Department Research Analyst
TASC, Inc.
12100 Sunset Hills Road
Reston VA 20190-3233
Tel. (703) 834-5000
Fax. (703) 318-7900
pjlaplaca@tasc.com