Four classes of change Buffer on 
1. Addition Line From 
2. Deletion 2008 
3. Modification 
4. No Change 
Additions 
made in 2008 
  
  
     
No >. 
" S. ge. 
Buffer on change - 
Line From 
200 Modification 
    
Lines from both dates 
contained in both buffers 
Deletions 
made in 2008 
Figure 2 The determination of change in line geometry (e.g., 
roads) between two dates. 
3. TEST DATA 
More than ten OSM data sets from cities and towns within 
South Africa (see Figure 3) were acquired for a period spanning 
from 2006 to 2011. Two data sets per year were used (for April 
and October) for year 2007 to 2011 and one data set from 2006, 
which gives 10 epochs. Eleven test areas have been determined 
as follows in: four large cities (Cape Town, Johannesburg, 
Durban, and Kimberley), three towns (Korsten, Middleburg and 
Stuttereheim), three suburbs (Brackenfell, Universitas and 
Gondeni) and one village (Makhwezini)), 
From examinations of the thematic attributes of the data sets a 
category of spatial features has been defined (e.g., streets, 
residential, recreational, religious, business, educational, etc,). 
The test areas have been categorised into three groups namely 
commercial (Cape Town, Johannesburg, Durban and Korsten), 
residential (Brackenfell, Kimberley, Middelburg and 
Universitas) and low urban density areas (Gondeni, 
Makhwezini and Stutterehim). Aerial imagery was used to 
determine visually that each site was representative of the 
chosen categories. Each test area covered a square area of 
approximately 6.5 km. 
A timeline has been built for each test. The timelines show the 
category and quantity of features that have been collected for a 
given period of time (epoch). For example, epoch 1 is the period 
between June 2006 and Jan 2007, epoch 2 is from Jan 2007 to 
June 2007, etc. The timelines from the different data sets are 
then compared to determine (i) the features that are recorded at 
different stages of the development of a volunteer geographic 
data set (ii) the rate of development of a volunteer geographic 
data set (iii) and the variation in data collection for different 
regions, e.g. rural versus urban, industrial versus residential, 
etc; 
  
SOUTH AFRICA 
OpenStreetMap Test Areas * 
Limpopo 
Gauteng » 
North West 
Mpumalanga 
* Free State . 
Northern Cape > 
Eastern Cape Legend 
® 
i, Commercial Test Areas 
,9Westem Cape @ Residential Test Areas 
« Rural/Low Urban Density Test Areas 
  
  
  
Figure 3 Location of areas in South Africa from which the test 
areas are drawn. 
4. RESULTS AND ANALYSIS 
The tests were performed in ESRI ArcGIS environment using 
the standard operations: buffer, clip, intersections, erase and 
statistics. Several python scripts were created with the help of 
the ESRI workbench and further adapted to process the data 
sets. 
4.1 Points — Location Data 
Additions 
Figure 4 portrays the additions only in the commercial and 
residential categories for the 10 epochs. There was no point data 
to analyse for low urban density areas. The vertical axis 
provides the total count of additions between epochs and the 
horizontal axis shows the epochs. Each line represents a 
different test area as follows: red-Cape Town, green-Durban, 
purple-Johannesburg and turquoise-Korsten in figure 4a. In 
figure 4b:  red-Brackenfell, green-Kimberley, purple- 
Middleburg and turquoise-Universitas. (Please note that the 
colour represents the same test areas in all figures.) As can be 
seen from the figure, there is a vast difference in the total 
additions made for the two categories. The number of additions 
in commercial areas had varying increases, per test area. The 
Cape Town data set was extensively updated during 2007 and 
2008 (epoch 1-2 and 3-4). Thereafter additions to Cape Town 
stabalised but remained low. Durban only had activity during 
2010 (epoch 7-8). Johannesburg started with a gradual increase 
of additions between early 2008 and late 2009 (epoch 4-6). 
Between epoch 7 and 8 (early 2010 to late 2010), Johannesburg 
had a sudden increase in activity. Korsten had no activity up 
until 2011 (epoch 9-10), but it was very low. After 2010 the 
number of additions increased more rapidly, which might be an 
indication that the FIFA World Cup of 2010 did influence the 
use and update of OSM. 
Additions to Residential areas have been very low in general. 
Brackenfell had interrupted periods of activity that is for 2009 
(epoch 5-6) and then only in 2011 (epoch 9) again. Kimberley 
and Middleburg had no activity for the entire study period. 
Universitas had minimal additions between late 2010 and early 
2011 (epoch 8-9). It is also apparent that, contributions in 
commercial areas started at an early stage and the data volumes 
are much higher, whereas contributions for residential areas 
were more progressive with lower data volumes. 
520