John Snow used his map of 1854 deaths by cholera to effectively convince London’s decision makers that the Broad Street water pump was responsible for the distribution of the disease. As a result, Snow is often credited with pioneering the use of information mapping to study the relationship between the built environment, epidemiology, and health outcomes more generally. The spatial pattern of the outbreak appeared unquestionably centered around the pump, but he lacked the benefit of spatial statistics and “clustering” analyses to further quiet skeptics. Using data digitized from his published map, you will test the clustering of his collected cholera death incident data. The questions you will answer are
Keep in mind what we know of the historical case: Cholera is water-borne, and water collection and consumption is a daily activity. Infrastructurally, the Broad Street pump belonged to a separate water system than its surrounding pumps. Further, remember what the original map includes (and excludes): Building locations where cholera deaths are recorded within the map; all other building locations are not.
For a brief background on re-analyzing Snow’s map while thinking through what we call “a neighborhood” (as well as analyzing other historical thematic maps with GIS), consider
Meisterlin, L and Baics, G. “Old Maps, New Tricks: Digital Archaeology in the 19th-Century City,” Urban Omnibus, June 17, 2015.
Complete the exercise below, answering the embedded questions.
Create one thematic map that revisits and re-argues John Snow’s original claim, intended for online viewing by a general, interested-public, audience. For this, your map composition image should measure 1600 pixels wide, be landscape-oriented, with either 3:2 or 16:9 aspect ratio. The map composition must include (graphically or otherwise) your responses to the three questions listed in the scenario.
Note that the data provided to execute the assignment do not include background or reference layers. These will have to be gathered separately for the composition of your map. (You might consider downloading a georectified image of the original Snow map or other base layers. Consider whether current base layers of London are appropriate, or the argument implied if they are used.)
To answer the questions listed above, we will use some common, basic techniques of describing spatial patterns statistically. We will first compute the geographic center (comparing the geographic mean, median, and central feature) of the building locations where John Snow recorded deaths by cholera in September 1854. We will then measure the distance from each geographic center to the Broad Street water pump.
We will also determine the standard distribution of affected locations around the mean center and calculate the percentage of recorded cholera deaths within one standard distribution of the mean. This should reveal whether cholera-related deaths were more or less concentrated around the center than would be expected in the case of a normal spatial distribution.
Lastly, we will test whether the locations of cholera-related deaths (and the number of deaths themselves) clustered within the Soho neighborhood of London with two tests. As a simple global measure of clustering, we will use the average nearest neighbor test to determine whether the building locations are globally clustered within the neighborhood. We will then use the Getis-Ord Gi test* for local clustering to determine whether and where high numbers of deaths clustered locally within the neighborhood.
<aside> 💾 Download and unzip the exercise data package, saving it to your working directory.
Exercise data packages are accessible via Google Drive with a Columbia University UNI login. If you would like a copy of the full course materials, see here.
</aside>
We will use data point-level data of John Snow’s 1854 cholera study in the Soho neighborhood of London.
Houser, Rhonda. 2013. John Snow Cholera Map as Rectified Raster Data, Water Pump Location Data, Deaths by Building Data, 2011 and 2013. University of Kansas Libraries. https://kuscholarworks.ku.edu/handle/1808/10772
The data package provided for this exercise includes several datasets, including their metadata as a txt file. (You will need the metadata file to decode the attribute fields.) A PDF version of the University of Kansas Libraries website from which they were downloaded is also provided in the \docs\ folder.