Slide 1
Slide 1 text
Mapping Data:
Beyond the
Choropleth
Oliver O’Brien
Senior Research Associate
UCL Dept of Geography
Twitter: @oobr
Research blog: http://oobrien.com/
ADRC-E Training Course: Introduction to Data Visualisation
16-17 May 2016, Farr Institute, UCL
Slide 2
Slide 2 text
Contents
• Technology Summary for Web Mapping
• Choropleth Maps: The Good and the Bad
• Moving Beyond the Choropleth
• Example: CDRC Maps
• Example: named – KDE “heatmap”
• Case example: Country of Birth Map
– Concerns of the data scientist & digital cartographer
2
Slide 3
Slide 3 text
1. Technology Summary for Web Mapping
3
Slide 4
Slide 4 text
Managed Wrappers
• MapBox Studio
• CartoDB
• ESRI ArcGIS Online
• Tableau
• Google Fusion Tables
• Google Maps Embed API
• Google Static Maps API
4
Example: MapBox Studio style editor
Slide 5
Slide 5 text
JavaScript APIs
• OpenLayers
• Leaflet
• D3
• Google Maps JS API
5
http://earth.nullschool.net/
http://osm.org/
Slide 6
Slide 6 text
Programming/Scripting
– Typically to produce raster imagery which then can be
combined with vector data in a Javascript API (or other)
map
– R
– Java
– Mapnik
6
http://twitter.mappinglondon.co.uk/
Slide 7
Slide 7 text
WMS/WFS
• “Non-slippy” webmap servers
– MapServer (C++)
– GeoServer (Java)
7
http://worldnames.publicprofiler.org/
Slide 8
Slide 8 text
2. Choropleth Maps: The Good and the Bad
x Treat unpopulated and
populated areas
similarly
x Can be hard to see the
“story” – the interesting
results
x Prone to M.A.U.P.
8
ü Easy to make
ü Computationally quick to make
ü Retain a geographic familiarity if done carefully
ü Good for comparing areas quickly
http://maps.cdrc.ac.uk/
Slide 9
Slide 9 text
A Basic Choropleth Map
Map of household income in the US by census tract. Source: Campus Activism blog.
Slide 10
Slide 10 text
10
London Mayoral Election 2016 – Result. Source: BBC News. http://www.bbc.co.uk/news/uk-politics-36303157
A Basic Choropleth Map
Slide 11
Slide 11 text
Adding Geographical Contextual Features
US Census: ACS Survey results
Slide 12
Slide 12 text
Adding Geographical Contextual Features
Map of Housing Affordability, 2014. Source: The Guardian.
Slide 13
Slide 13 text
Adding Geographical Contextual Features
Map of Philadelphia housing prices per square foot using 2014 property assessment data. Source: Campus Activism blog.
Slide 14
Slide 14 text
Adding Geographical Contextual Features
Change in Socio-Economic Status, 2011-2011. Source: Neal Hudson at Savills
Slide 15
Slide 15 text
3. Moving Beyond the Choropleth
• Adapting Choropleths
– Limiting Display to Populated Places
• Dot Maps
• Cartograms
• Grid Maps
15
Slide 16
Slide 16 text
Source: http://jamesjgleeson.wordpress.com/
Slide 17
Slide 17 text
Adapted Choropleth
• Limiting Data
Display to
Populated
Places
• Adding
Contextual
Information
Source: Neal Hudson at Savills
Slide 18
Slide 18 text
Dot Density
• Fairer for population
fluctuations
– Although layering of dots
is crucial
• Hard to read data in
high-density areas
• Assumption of random
distribution across areas
– Unless areas are
restricted to buildings
Ethnicity across the New York City metropolitan area. 1 dot = 20 people. Source: Eric Fischer.
Slide 19
Slide 19 text
19
Map of ethnicity supergroups based on ONS Census (2001) data. 1 dot = 50 people. Source: @geographyjim
Dot Density
Slide 20
Slide 20 text
EthniCity from London: The Information Capital (James Cheshire, Oliver Uberti) http://theinformationcapital.com/
Dot Density
Slide 21
Slide 21 text
Single Dot in Area Centroid
21
Wards - London’s Political Colour
http://vis.oobrien.com/london/
Map: OSM CC-By & OS OGL
Slide 22
Slide 22 text
22
http://vis.oobrien.com/tube/#metric=wardwords
Data: ONS & OS, OGL.
Slide 23
Slide 23 text
Cartograms
• Fairer display of data
• Harder to relate geography unless carefully done
Maps from Worldmapper: http://www.worldpopulationatlas.org - © SASI Research Group, University of Sheffield
Slide 24
Slide 24 text
Cartograms
Created by Ben Hennig – http://viewsoftheworld.net/
Slide 25
Slide 25 text
Square Cartograms
• Aftertheflood
Squares
(Boroughs)
25
Map/Concept: Aftertheflood.co (L). GLA version (R).
Slide 26
Slide 26 text
Half Way Between Choropleth & Cartogram
• Rentonomy (Postcode Prefixes)
26
Source: Rentonomy. http://www.rentonomy.com/london-rental-map
Slide 27
Slide 27 text
27
Gridded Choropleth
Map data: ONS OGL. Source: Duncan Smith, UCL CASA. http://luminocity3d.org/
Slide 28
Slide 28 text
28
Gridded Choropleth
Map data: ONS OGL. Source: Duncan Smith, UCL CASA. http://luminocity3d.org/
Slide 29
Slide 29 text
29
Gridded Choropleth
Map data: ONS OGL. Source: Duncan Smith, UCL CASA. http://luminocity3d.org/
Slide 30
Slide 30 text
4. Example: CDRC Maps
• CDRC needs maps of population/consumer data
which are quick to interpret and effective
• The technology is simple
– We want to maintain the simplicity of creating
choropleth mapping
• The key innovations are to:
– Put some contextual information above the choropleth
– Constrain the choropleth display to areas of population
30
http://maps.cdrc.ac.uk/
Slide 31
Slide 31 text
31
Map data: Ordnance Survey and ONS OGL. Source: http://maps.cdrc.ac.uk/
Slide 32
Slide 32 text
32
Map data: Ordnance Survey and ONS OGL. Source: http://maps.cdrc.ac.uk/
Slide 33
Slide 33 text
33
Map data: Ordnance Survey and ONS OGL. Source: http://maps.cdrc.ac.uk/
Slide 34
Slide 34 text
Map data: Ordnance Survey and ONS OGL. Source: http://maps.cdrc.ac.uk/
Slide 35
Slide 35 text
Map Layers
• Also Postcode Pin Layer (Vector)
• KML Drag-and-Drop Display Layer (Vector)
Slide 36
Slide 36 text
Tabular Data > Choropleth > Real-World Map
Slide 37
Slide 37 text
UI Evolution: “New Booth” > DataShine > CDRC Maps
37
Map data: OS & ONS OGL. Sources: http://vis.oobrien.com/booth/ + http://datashine.org.uk/ + http://maps.cdrc.ac.uk/
Slide 38
Slide 38 text
Geodemographics
& Indices
on CDRC Maps
• OAC, LOAC, TOAC
• COWZ-EW, IUC
• IMD 2010 & 2015,
2010-15 Change,
Components Diff
• SIMD 2012
• Retail – Rental Value,
Value Change by
Sector
Map data: Ordnance Survey and ONS OGL. Source: http://maps.cdrc.ac.uk/
Slide 39
Slide 39 text
Map data: Ordnance Survey and ONS OGL. Source: http://maps.cdrc.ac.uk/
Slide 40
Slide 40 text
Software (for CDRC Maps)
• Simple (static image “tiles”) – no “map server”
• Apache web server
• Mapnik (pre-renders the images)
– & python-mapnik (to slice them up)
• OpenLayers (3)
• JQuery and JQueryUI (for the non-map UI)
• PostgreSQL database
– PostGIS spatial extensions
Slide 41
Slide 41 text
5.
Examples:
named
41
Source: Adapted from named. http://named.publicprofiler.org/
Slide 42
Slide 42 text
Source: named. http://named.publicprofiler.org/
Slide 43
Slide 43 text
Replacing the “Worldnames” Choropleths
Source: http://worldnames.publicprofiler.org/
Slide 44
Slide 44 text
KDE Mapping: “Sinclair”
Slide 45
Slide 45 text
KDE Mapping
Graphs: Wikipedia. Background Map: OpenStreetMap.
Slide 46
Slide 46 text
• “Where your name is
unusually popular”
• “Where we think you might
have met”
The Website
Source: named. http://named.publicprofiler.org/
Slide 47
Slide 47 text
The Response…
Sources: Daily Mail and Mirror websites.
Slide 48
Slide 48 text
Software (for named)
• Java service which retrieves the data, grids it,
creates a corresponding KDE grid and converts it
to a PNG
• Apache web server
• PHP to glue the two together
• OpenLayers (3)
• JQuery and JQueryUI
• PostgreSQL database
– No need for a spatial database as just points
Slide 49
Slide 49 text
6. Case Example: Country of Birth Map
• “Top Metric” maps are pseudo-geodemographic
maps
– showing a single value for an area that represents the
most significant part of the population there
49
Slide 50
Slide 50 text
6. Case Example: Country of Birth Map
• Need three kinds of skills
– Data Scientist
• to manage the data and discover the story
– Demographic Geographer
• to make it a representative map
– Digital Cartographer
• to communicate the story effectively
50
Slide 51
Slide 51 text
• Top Metric Maps require care and curation to produce a
map with:
– fair groupings – wary of aggregation bias
– a sensible threshold – maximise signal-to-noise ratio but don’t
lose the story
– appropriate removal of spatially overwhelming majorities
– appropriate colours – use hues to show categorization and
hierarchies (HSL)
– curated emphasis with colours - emphasise/fade certain data to
tell the story of the data effectively - use brighter hues for more
unusual results, and more modest ones for results that would
otherwise dominate, while retaining balance
Country of Birth Map
Slide 52
Slide 52 text
Country of Birth Map: Data Scientist
Slide 53
Slide 53 text
• An all-UK map of
Census 2011 data,
combining the equivalent
(but subtly varying) tables
from the 3 National Statistics
bodies – ONS, NRS &
NISRA.
• English excluded from all
UK, & natives from their
country
– Internal national land borders
included to show these rule
transitions
• 8% threshold
– Balance between
“exaggeration” and showing
an interesting story
Country of Birth Map: Demographic
Geographer
Map data: Ordnance Survey and ONS OGL. Source: http://maps.cdrc.ac.uk/
Slide 54
Slide 54 text
Country of Birth Map: Digital Cartographer
Map data: Ordnance Survey and ONS OGL. Source: http://maps.cdrc.ac.uk/
Slide 55
Slide 55 text
Country of Birth Map: Digital Cartographer
Map data: Ordnance Survey and ONS OGL. Source: http://maps.cdrc.ac.uk/
Slide 56
Slide 56 text
Country of Birth Map
Map data: Ordnance Survey and ONS OGL. Source: http://maps.cdrc.ac.uk/
Slide 57
Slide 57 text
Paper on CDRC Maps Mapping
• O’Brien O, Cheshire J (2015)
Interactive mapping for large, open demographic
data sets using familiar geographical features
– Journal of Maps (T&F)
– Published online
– Online, PDF download
– Open access (CC-By)
DOI: 10.1080/17445647.2015.1060183
57
Slide 58
Slide 58 text
Links
http://maps.cdrc.ac.uk/
http://maps.cdrc.ac.uk/#/metrics/countryofbirth/
http://named.publicprofiler.org/
58
Map data: Ordnance Survey and ONS OGL. Source: http://maps.cdrc.ac.uk/
Slide 59
Slide 59 text
Thanks!
• Research blog:
http://oobrien.com/
• Twitter: @oobr
59
Source: http://vis.oobrien.com/tube/