Making sense of spatial data

Maps are becoming more common place but the skills and knowledge to work with spatial data isn’t as obvious as it first appears. This public lecture explored the benefits of using maps and spatial analysis to explore problems. This talk was given at the National Library on 16 June as part of the cartography series jointly presented by the National Library and Victoria University of Wellington.

Mairéad de Roiste lectures in Geographic Information Science, cartography and all things map related at Victoria University in Wellington. She's passionate about geospatial education and currently heads the national masters in GIS programme across Victoria University, AUT and the University of Canterbury. Her research interests focus on usability, pedagogy and spatial econometric modelling.

Listen to the talk

Stream the talk (65 min):

Download the recording (mp3, 76MB)

This post is an adapted version of Mairéad’s presentation.

The case for maps

What are you looking at?

Google maps view of the Wellington region.Google maps view of the Wellington region.

A map of Wellington? Think about what you are doing when you look at this map. Do you recognise the location?

What cognitive functions are you using to interpret the image in front of you? And think of how you extend those functions by relating them to the ‘real’ world.

Google Earth view of the Wellington region.Google Earth view of the Wellington region.

Is this satellite imagery easier to interpret?

Navigation maps can be more abstract – but they’re still useful. Take the London Underground map.

The well-known London Underground map, with distorted geography that makes the tube lines more legible.The London Underground. Via VisitLondon.com.

This map is highly abstracted, with distances and angles distorted, but it’s useful to millions of commuters.

London Underground lines displayed in a more geographically accurate way. The lines are less legible, but the shape of London is more familiar.More geographically-accurate map of the London Underground system. Via Balstone.co.uk.

A more geographically accurate version is less useful for navigation within the underground system. Without the abstract map’s other features it’s harder to navigate.

The London Underground overlaid on a more classic map of London.Geographically-accurate map of the London Underground produced by TFL. Via The Telegraph.

This version of the tube map is more geographically accurate – distances are preserved and some contextual information is provided.

Maps of New Zealand's State Highways, in the style of the London Underground map.State Highway maps by Andrew Douglas Clifford. Via TheGoodPrint.co.nz.

The same map techniques have been used in New Zealand by Andrew Douglas Clifford to map New Zealand’s State Highways.

This map doesn’t aid navigation but has received media attention and is available for purchase. The map supports the idea of maps as art. Making a map aesthetically appealing can make an audience spend more time on it.

Map of the world from 1607, showing the Eastern and Western hemispheres. The South is largely taken up with 'Terra Australis Incognita'.David Rumsey Historical Map Collection: 1607 Gerard Mercator and Iodocus Hondius. Via DavidRumsey.com.

Historical maps are not just navigation – they’re a way of knowing about the world.

Map of the world from 1665, showing the Eastern and Western hemispheres. More of the South is accurately shown. The map is illustated and bordered with extensive decoration, including ships at sea, and heavenly hosts.David Rumsey Historical Map Collection: 1665 Joan Blaeu. Via DavidRumsey.com.

The extra details on these maps, like the ships in the oceans and the surrounding illustrations, convey a sense of place that augments the cartography.

In a digital age, the animation of data can make our understanding of patterns easier:

Screenshot showing a current wind map of the USA.Wind map. Via Hint.fm.

This map, for example, displays wind for New Zealand:

Screenshot showing a current wind map of New Zealand. It's much more colourful.Wind map. Via Windyty.com.

Why use a map?

There are different reasons to use a map. Some are obvious, like simple navigation. The skills required for visualisation are far more complex:

Navigation Measurement Visualisation
Search Search Search
Identify and locate own position on a map Identify Identify
Orient map Count Describe
Search for optimum route on map Compare Compare
Search for landmarks en route Contrast Contrast
Recognise landmarks on route Estimate Discriminate
Search for destination Interpolate Delimit
Identify destination Measure Verify
Verify Generalise
Prefer
Like

– via Clarke, 2003.

What is spatial literacy? According to Goodchild, it’s the ability to

...capture and communicate knowledge in the form of a map, understand and recognize the world as viewed from above, recognize and interpret patterns, know that geography is more than just a list of places on the Earth’s surface, see the value of geography as a basis for organizing and discovering information, and comprehend such basic concepts as scale and spatial resolution. (2006)

Interpretation

Maps are ways of presenting information, and when they're read they're interpreted.

Ethnicity maps based on census data from 2000 created by Eric Fischer show the racial segregation (or otherwise) of different US cities. Each dot represents 25 residents.

Maps showing the ethnic segregation of various cities at the 2000 census. Washington has a stark East/West divide; Detroit is even more strongly segregated. San Francisco has concentrated areas that mix significantly more.Ethnicity maps of (top to bottom) Washington DC, Detroit, and San Francisco. Made using data from the 2000 census.

Read more about what these maps reveal

Interpretation of maps found 17 lost pyramids in Egypt in 2011.

Image of several pyramids, along with comparative satellite photos showing areas blocked out in colour by computer analysis.House imagery and classification via Interimage.

Interpretation also finds some more surprising results.

Satellite image of mountainous terrain, overlaid in white MS paint lines to 'reveal' an image of a man with a beard, and possibly some sort of hat.Via BigThink.com.

What is spatial thinking?

It includes

  • Spatial habits of mind – know where, when, how, and why to think spatially
  • Spatial concepts
  • Spatial thinking skills – which relies on tools, technologies, and concepts
  • Critical spatial thinking – the evaluation of quality

Data quality

New data collection methods are affecting how we should interact with and analyse datasets.

For example, look at the VUW analysis of the Great Kererū Count.

Screenshots from the Great Kererū Count analysis project, showing concentrations of Kererū in Wellington's town belt.Great Kererū Count analysis project.

The consequences of bad data can be severe, as when the Limerick County Council demolished a family home due to a plot of land being mislabelled. Or when the police raid the wrong house due to an error in GPS data. Or being the default ‘centre of the USA’ led to trouble for farmowners in Kansas.

Google map showing the location designated 'the geographic centre of the USA'.The geographic centre of the USA on Google Maps.

The address in question was a ‘cleaned’ version of the geographic centre of the United States. The issue was one of IP mapping, which defaults to the nearest available IP address, but that data was then used for other kinds of mapping, making it look like this farm was the centre of an enormous amount of activity.

Power of analysis

By combining datasets, spatial analysis can create new information. In this case researchers at the University of Canterbury investigated the prevalence of fast food and convenience store outlets near schools. They combined three key datasets and also took into account distance from the schools.

Our own network analysis of Greater Wellington Region commuter data calculated the travel time to work with 4 different types of transport (walking, cycling, driving and public transport). (See T C Daglish, M de Róiste, Y Saglam, and R Law, 2015, "Commuting and Residential Decisions in the Greater Wellington Region". Working paper.)

Map of Wellington city and nearby suburbs, indicating one person's addresses, workplace, and route of travel.Commuter map of Wellington. Via Daglish et al (2015).

We calculated the actual commute mode and route as well as alternative locations. By looking at alternatives, we leveraged ‘what if’ modelling to see the effect of possible changes to the transportation network.

Privacy

While combining datasets is useful and very powerful – it can also lead to some unintended consequences. Please rob me was a site set up to raise awareness about oversharing of personal location information.

Homepage of Please rob me, a site for raising awareness about oversharing data.Please rob me.

The addition of a geographic location can make data more powerful but can also throw up some ethical issues. Family watchdog is a US website which allows users to display the location of sexual offenders around a particular geographical location.

Homepage of Family watchdog, a site for tracking locations of sexual offenders.Family watchdog.

On this site, I explored an area around the Hollywood Roosevelt hotel in Los Angeles, on Hollywood Boulevard. House icons display the location of the hotel, and the markers are different types of sexual offenders on the sexual offenders register.

Mapped locations of sexual offenders around Hollywood Boulevard, Los Angeles. Shows around 450, noticably clustered.Sexual offenders mapped around Hollywood Boulevard, Los Angeles.

Each of the markers can be clicked and information about that offender displayed – including their photograph, name, address and details of the convicted crime.

Developing critical spatial thinking

When information is presented, the choices made in that presentation can affect what is communicated.

Here, the first graph makes it look like house prices have more than doubled in one year:

Two charts of the average house price in 1998 and 1999. One uses a misleading y axis to make it seem as though prices have more than doubled. Another, more accurate one, shows the change is relatively marginal.Via the BBC.

Graphs can be misleading. Here, the axes are very different so give the impression that abortions are significantly overtaking cancer screening by Planned Parenthood in the US.

Graph 'charting' the provision of abortions versus cancer screenings and prevention services by Planned Parenthood in the US. The two kinds of service are presented as straight lines crossing each other, intending to suggest PP is now performing radically more abortions than screenings. However, the two lines use different y axes, and screenings are still triple the number of abortions provided.Via Politifact.

These two maps of population growth, using the same dataset, look very different and tell different stories:

Two maps of the USA, showing visualisations of the same population dataset. However, they use different colouration and thresholds, making the actual distinctions in population growth clearer.Via The Washington Post.

The next three maps were produced by Masters in GIS students on the GISC 403 Cartography and Geovisualization course in 2015. Each set of maps aims to tell competing stories for the same topic.

Two maps of the world, with red indicators of military spending. The map that reports military spending as a percentage of GDP suggests most spending is happening in the Middle East and Africa, while the map that uses absolute values suggest that the USA is the most significant spender, distantly followed by China, Russia, and Europe.Competing narrative maps by Hayley Hume-Merry.

The focus of these maps was on different data – one took into account the size of the economy and the other just reported total numbers.

Two maps of New Zealand, showing different approaches to narrative. The first is framed as showing labour force participation, with a strong South Island in dark green. The second is framed as showing unemployment, with a weak North Island in deep red.Competing narrative maps by Tom Norman.

The first map uses ‘calm’ colours to produce a particular pattern of unemployment while the second map uses different but related data and does not give the actual values.

Two maps of Wellington harbour. One showing swimming points in terms of their pollution threat, the other showing those points in terms of how safe they are.Competing narrative maps by Matthew Velde.

Both colour and data choice for the same locations tell different stories.

Sometimes straight geography misleads. This 2008 US presidential election map has geographically accurate areas, and seems to show that the result was quite close:

Geographically accurate map of the USA, showing states colour coded by their result in the 2008 presidential election. The result, when viewed by area, seems close.Via DashboardInsight.com.

This is a cartogram which has distorted the geographically accurate areas to take account of actual population. The map shows that more of the population voted for the ‘blue’ candidate.

Skewed map, adjusted to match the population of each state. Shows that a significantly larger proportion of voters voted blue.Via DashboardInsight.com.

Colour and our interpretation of colour plays and important role in how we evaluate maps. Blue used to be associated with the republicans and now we expect the opposite. This image is from CBS in 1980:

TV grab from CBS in 1980, showing results from that year's election. The colours are the reverse of what we now expect, with the Democratic candidate showing as red and the Republican showing as blue.Via SmithsonianMag.com.

The attraction of available data

Just because the data is available doesn’t mean we should map it. We need to ask questions – is the data appropriate? Backing up visual interpretation with statistics is important to ensure we find real patterns

And we need to understand that though the map might correlate with something, it doesn’t mean that it causes it.

Cartoon showing two figures talking. One says 'I used to think correlation implied causation. Then I took a statistics class. Now I don't.' The second replies 'Sounds like the class helped', to which the first says 'Well, maybe.'Via xkcd.com.

Testing our assumptions

Research needs to be done to better understand how we can create better maps as well as how we can develop critical spatial literacy:

  • What map designs are effective?
  • How much of an effective map design is dependent on data?
  • What other factors are involved?
  • How can we measure critical spatial thinking?

There is a skill shortage in geospatial work in New Zealand, according to organisations we surveyed. However, government is increasingly recognising this, and change is happening.

Take home points

Advances in mapping tools, technologies, and concepts are significant, and positively affect our interaction with the world.

However, we should critically examine the maps we use.

Backing up visual interpretation with analysis is important.

Research on how we use maps and spatial literacy is evolving, but there’s still a long way to go.

Further reading

S W Bednarz and K Kemp, 2011. "Understanding and nurturing spatial literacy", Procedia - Social and Behavioral Sciences, 21, 18–23.

T C Daglish, M de Róiste, Y Saglam, and R Law, 2015. "Commuting and Residential Decisions in the Greater Wellington Region", working paper.

M de Róiste, 2009. Geographic information in eLocal government: evaluating online mapping applications in Irish local authorities"Irish Geography, 42(2), 165-183.

M de Róiste, 2014. Filling the gap: The geospatial skills shortage in New ZealandNew Zealand Geographer, 70(3), 179-189.

M de Róiste, M Gahegan, P S Morrison, M Ralphs, and P Bucknall, 2010. Geovisualisation and policy: exploring the linksOfficial Statistics Research Series, 5.

M F Goodchild, 2006. The Fourth R? Rethinking GIS Education.

National Research Council, 2006. Learning to Think Spatially: GIS as a Support System in the K12 Curriculum, Washington, The National Academies Press.

By Mairéad de Roiste

Mairéad lectures in Geographic Information Science at Victoria University in Wellington.

Post a Comment

(will not be published) * indicates required field

Be the first to comment.