Olympic Changes in Atlanta

Olympic Venues, Social Index & Property Values

    Olympic venues in Fulton County were geocoded to a map of Fulton County. The venues were obtained through Google Earth and Wikipedia . Census data was obtained for the year 1990 and 2000 from the US Census Bureau for Fulton County, Atlanta. The data obtained was for census block groups, since this study examines the immediate areas around the Olympic venues.

1) First, we created a Social Index (SI) for 1989 (1990 census) and for 1999 (2000 census).  Our SI is somewhat similar to the one that David Ley used.  We added the percentage of people aged 25 years and older who hold an advanced degree (above a bachelor’s degree) to the percentage of households earning over $50,000/year, and divided by two to get a SI percentage. One problem we had was that $50,000 does not provide the same buying power between 1989 and 1999, due to inflation. However, because we were using the number of households making $50, 000 or more, not the actual dollar value, we could not simply just adjust for inflation. To try to account for inflation, for the latter year we used households that earned $75,000 or more per year. The actual value of $50,000 in 1999 should have been close to $70,000, but the closest census data break was at $75,000 so this is what we used.

We also looked at the median owner occupied property values. The 1999 dollar values were adjusted to 1989 levels to make the comparison more accurate.

2) Once we downloaded and adjusted all of the socio-economic we created a DBF file, with fields for the SI and property values.

3) Downloaded the shapefiles for Fulton County, cities in the county, 1990 and 2000 census block group shapefiles, and the roads shapefile. For some reason the shapefiles did not project properly, so we had to reproject all of the shapefiles. From the Fulton counties shapefile, we selected for just Atlanta, and saved as a layer so that we could use just Atlanta for part of the analysis.

4) Created an address locator file, and geocoded the venues. We ended up geocoding only 5 venues, because we wanted to look only at venues that were built or retrofiited for the '96 Games and because many of the Olympivvenues were actually located outside of Fulton County (some even outside of the state). All of which were in downtown Atlanta. Each venue was saved as a shapefile.

5) Joined the SI and property values dbf file to the census block group shapefile.

6) The boundaries of the census block groups had changed between the two census’s.  Not only had some been broken into two or three blocks, but some had been joined together as well. Some boundaries had actually changed too, not just broken apart or merged. This means it would have been very difficult to create a map of the changes in SI and property values. In order to adress this, first we used the clip tool to create layers with all the data for Atlanta proper only. This gave use both Fulton County and Atlanta proper layers with the SI and property values joined.  We then used the spatial analyst tool>convert feature to raster so that we could rasterize the Fulton county and Atlanta layers. The rasterizations were based on both the SI and the property value fields.

7) Created maps for both the SI and property values for both Fulton County and Atlanta, to visually compare.

8) To create maps of the change in the SI and property values, we used the spatial analyst> raster calculator. We added all the values for both SI and property values for both Fulton County and Atlanta, and minused the 1990 census vales from the 2000 census values.  For the SI index, this gave us maps of percent change between the two years. For the property values, we had to add an extra step. Because the range of vales was so high, we could not look at the attribute table, and therefore could not use the raster calculator to get a map of the change in property values. To remedy this, for both 1990 census and 2000 census property value field, we used the raster calculator and divided the total by 1000 using this formula: [raster name] =int([raster name]/1000). We did this for both the Atlanta and Fulton county raster layers. We could then minus the 1999 census property values from the 2000 census property to create maps of the change in median property value for both Atlanta and Fulton County.

8) We wanted to look at the changes in SI and property values at several distances from the venues, so we created 1000, 2000, 3000, 4000, and 5000 meter buffers (Arc Toolbox>proximity>buffer) around three venues (we only used three because the other two are so close to one of the others they are virtually on top of each other). For each distance, we used the union tool to join all the buffers. This allowed us to have one shapefile for each distance from the venues.

9) In order to used these buffer/union shapefiles to clip from the Atlanta SI and property values maps, we had to convert these maps back to vector using spatial analyst tool>convert raster to feature.

10) We then used the clip tool to clip the Atlanta SI and property values vector maps to each of the five distance shapefiles.

11) Used Hawth’s tools to calculate the area of each polygon within each clipped distance map (1000, 2000, 3000, 4000, 5000 meter).  We then used Hawth’s tools to calculate the total area of all the polygons within each clipped buffer distance map.  Then added the areas of each polygon within a certain range (10 percent range for the SI index) and divided by the total area of each buffer distance map to create pie charts of the percentage change of each range of values of the total area of each buffer distance from the venues. We also used the same method for property values (added all polygons areas in $50,000 property value ranges together in each distance buffer, and divided by the total area of each distance buffer).

12) Used Hawth’s tools to calculate the centroids of the SI and median property values polygons for Atlanta proper. We then created a spatial join, joining the Social Index change and property value change polygons to the centroids of each polygon. Then we used Hawth’s tools to calculate the distance from each centroid to the three venues.  We then plotted distance to venues as the independent variable against the SI and property value of the polygon containing each centroid as the dependent variable to determine if there was a correlation, but there was no real correlation.

13) Used Hawth Tools and the spatially joined centroid/Social Index change polygons to create a Kernel density map of the social index change.

14) Used Arctoolbox>analyzing patterns>clustering tool to create a hotspot cluster map of the social index change.

Comparing Land Use Data from 1980s to 2003

    Land use data compiled from the 1970s and 1980s was obtained and compared to 2003’s land use data.  To compare the two land use data layers, we wanted to see how much land use has changed in the city of Atlanta in Fulton County, and whether the Olympics created any significant changes.  Buffers of 1 km, 3 km and 5 km were created around the Olympic venues to analyse the areas aroud the venues.  After the data was obtained for the buffers, a chi squared test was used to see if the observed land use changes were similar to the expected land use changes.

Creating a new land use category and land use code.

    A new land use column called NEWLU_NAME was created in the attributes table for both the 1980's and 2003's land use data. This was created to have a comparable legend between the data sets. The legend entries were already very similar to begin with. The legend categories for 2003 was modified from the 1980's data. The end result was 12 new land use categories and land use codes added as fields to the attributes table for both land use years and the buffers.

AGRI (agriculture)	1
COM_SER (commercial and services)	2
FOREST	3
IND/COM (industrial and commercial)	4
INDUSTRIAL	5
QUARRIES	6
RESIDENTIAL	7
TCU (transportation, communication and utilities)	8
TRANSITIONAL	9
URBAN_OTHER (other urban or built up areas)	10
WATER	11
WETLANDS	12

Clipping land use maps to the shape of the city of Atlanta

    A shape file was created for the area in the city of Atlanta that is in Fulton County using US Census Tiger files from 2000.  The land use data were clipped to the city of Atlanta shape file using ArcToolbox, Analysis Tools, Extract, Clip.
    Three different buffers were created to examine the areas near the Olympic venues: 1 km, 3 km and 5 km. These buffers were created using ArcToolbox, Analysis Tools, Proximity, Buffers and buffered around each of the venues. Then, the buffers were joined together using the Union tool in ArcToolbox. To ensure that the buffers were one continuous polygon, a new shapefile was created and a polygon was drawn around the buffered areas and clipped to the buffers using the Clip tool from ArcToolbox.

Rasterizing the  land use layers

    The buffers and clippings of Atlanta were rasterized using Spatial Analyst, Convert, Features to Raster.  The shapefiles for Atlanta were rasterized based on the new land use codes.  In order to extract the buffer areas from the city of Atlanta, Spatial Analyst's Raster Calculator was used with the expression: buffer (1 km, 3km or 5km) + rasterized city of Atlanta layer (1980 or 2003).  The final product was 2 rasterized city of Atlanta layers, one for 1980 and one for 2003, and 6 rasterized buffered areas around the Olymic venues.  Three were from 1980 and the other three were from 2003, with the buffers having a distance of 1 km, 3 km and 5 km.

Reprojecting  1980's Land Use Data

    The 1980’s land use data set was in latitudinal and longitudinal coordinates.  The data was reprojected into a NAD 83, transverse Mercator coordinate system by going into Properties of Layer, Coordinate System and importing the coordinate system NAD 83, transverse Mercator using 2003’s projections.  Originally, an Error Matrix would compare the two land use maps but a problem occurred during the comparison.  The 1980's file was set to a different projection system.  Even after projecting the 1980’s file to a Transverse Mercator projection, the data set didn’t match the 2003 data set.  The two data sets were offset from each other, making it difficult to compare using an Error Matrix.  The map shows the differences in the data sets.

Chi Squared (χ2) Test

    A chi squared test examined the significance of Olympic venues to land use change for the buffered areas.  The chi squared test formula is as follows:

Expected values of land use were based on the city of Atlanta land use change from the 1980s to 2003, and the rate was multiplied to the buffered areas for 1980.  Land use from 2003 buffered areas were the observed values.  The values obtained from the city of Atlanta as a whole was used in the calculation for the percent change in land use between the two years for each category of land use.  The expected and observed values were compared see if the expected change and the observed changed were significantly different.  If there was a significant difference between the observed and expected, we will know that the Olympic venues created a significant change to the areas around it.