outline cartographic visualization · • one-dimensional – lines • two-dimensional – areas...
TRANSCRIPT
1
Cartographic Visualization
Sarah E BattersbyUC-Santa Barbara
Outline
• Type of geographic representation• Levels of measurement• Mapping techniques• Data classification
What not to do…
Pick colors that confuse the color-blind or are, well, just ugly – the “art” in cartography
Present inaccurate data
Put too much information on the map
Ooh, numbers…and colors
2
Another one of those European countries changing their name??
Wasn’t that just Switzerland / Czech Republic?
Cartographic process
MeasurementSimplificationProjectionTranslation / scaling
Cartographic modelingDesign
PerceptionCognition“Making good maps”
Data vs. phenomena
• Phenomena – the characteristics of the real world
• Data – how we represent the phenomena– Simplify / Generalize– Sample
3
Geographic data
• The things that we put on the maps• Types of geographic data
– Differentiated by spatial dimensions• Zero-dimensional – points• One-dimensional – lines• Two-dimensional – areas• 2.5-dimensional • Three-dimensional
Point Phenomena• Points have no spatial-extent: zero-dimensional
• Point symbols have spatial extent
• Examples of geographic point data?– Weather stations, oil wells, eagle nesting sites– Location described by coordinates
• (X,Y)• (X,Y,Z)
• Cartographically “flexible” – put them where they look good (to some extent)
Points and flexibility
Camp Sarah…In the middle of the river
Camp Sarah…Shifted out of the waterOn the correct side of the river
Linear Phenomena
• Lines have 1 dimension: length, but no width
• Lines have drawing width
• Examples of geographic lines?– Borders between countries– Flight lines
• Described by a series of coordinate
4
Areal Data
• 2-dimensional in spatial extent: length and width
• Examples of geographic areas?– Lakes, political areas…
• Described using a series of coordinates that close a region
How does map scale influence spatial dimension?
• Map Scale and cities?
• Map Scale and rivers?
Characteristics of phenomena
• Discrete v. Continuous• Abrupt v. Smooth
Discrete Phenomena
• Occur at distinct location– With space in between– Examples: people living in a city – as
points, space between– Weather stations– Restaurants
Continuous Phenomena
• Continuous phenomena are defined everywhere:– Examples: elevation, air pressure,
temperature, land cover
5
How do they change?
• Phenomenon can change two main ways over space…– Abruptly or Smoothly
– Abrupt: changes abruptly (e.g., sales tax by county)
– Smooth: change has no boundary (e.g., temperature)
Individual, isolated occurrences
Continuous, no space between values
Distinctive and substantial breaks
Continuous variationNo breaks
Levels of Measurement
• When we measure geographic phenomena, we “collect data”
• Levels of measurement describe data characteristics…
• There are 4 you need to know1. Nominal2. Ordinal3. Interval4. Ratio
Levels of Measurement: NOMINAL
• What does Nominal Mean?– relating to or constituting or bearing or
giving a name • Nominal Measurements are NAME-
based measurements…• Categories and Groups• You Cannot order Nominal data
Nominal data exampleOrdinal Data
• Involves categories plus an “ordering”• A distinct order exists… numerical
values are undefined• Low-Medium-High
6
Ordinal data examplesInterval Data
• Ordering plus an indication of a numerical difference
• We know the number of units between each level…but we do not know what “0” is
• Differences make sense, ratios of data do not– 10° F to 30° F is a difference of 20° F– 30° F is not 3x hotter than 10° F
Ratio Data
• Just like interval data BUT ZERO IS MEANINGFUL
• Non-arbitrary zero point
Ratio Data Examples?
• Degrees Kelvin: zero is absolute zero– 40 is twice as warm as 20 (in terms of
kinetic energy)
• Percentages are also ratio data– 40% is twice as much as 20%
So, when we look at data we ask…
1. Is it discrete or continuous?2. Is it abrupt or smooth?3. Is it nominal, ordinal, interval, or ratio?
• Then we can make our map correctly…
7
Discrete or Continuous? Abrupt or Smooth?
Nominal, Ordinal, Interval, Ratio? From Spatial Arrangement of Phenomena to Making Maps!
• Basic principles guide cartography• From phenomena, we model data• From the data characteristics we can
select the right type of map to represent out data
Types of thematic maps
• Four basic types of thematic maps– Choropleth– Proportional symbols– Isopleth– Dot density map
Proportional Symbols
8
When to use proportional symbols
• Phenomena– Discrete with abrupt variation
• Data– Pretty much any magnitude data (not
ratio!)
• Variation of data– Relatively large to provide visual interest
Proportional symbol issues• Symbol overload (don’t use too many!)• Selecting size of symbols
– Make sure that they are distinguishable from neighbors
• Color – Make sure there is a good contrast from
background• Number of variables
– Can show numerous variables (e.g., graduated pie chart)
When to use isopleth maps
• Phenomena– Continuous with smooth variation
• Data– Values measured as sample locations
(e.g., weather stations)
Isopleth map Dot Density Map
9
When to use dot density.
• Phenomenon– Discrete with smooth variation
• Data– Magnitude / count rather than derived
value • Population total rather than ratio
• Each dot is a spatial proxy for the value it represents
Dot density pro/con
• Advantages– Intuitive (more = more)– Reveals overall patterns of distribution
• Disadvantages– Difficult to estimate density– Density of dots can make interpretation
difficult– Literal interpretation (people are RIGHT
HERE)
Cartograms
• Cartographic maps are true to location and area
• Cartograms are true to attributes
• Physical geography/space distorted to emphasize “attribute space”
Cartogram Cartogram pro/con
• Advantages– Visually interesting
• Disadvantages– Can be difficult to interpret– Hard to deal with extremely large/small
values– More challenging to produce
Cartograms vs. Choropleths
10
Choropleth mapWhen to use choropleth
• Data– No real world phenomena that really fits– Generally represents aggregations of data
represented by spatial unit (e.g., average income per county)
– Don’t use count data• Example: Bigger counties tend to have more
people– You don’t mind giving the impression of
equal distribution of attributes (population evenly spread throughout county)
Is Choropleth right for you?
1. Is your data attributed to enumeration units? (counties, states, tracts, etc.)
2. What type of data?1. Total values / counts?
1. Population2. Total income
2. Derived values?1. Ratios involving area (density)2. Ratios independent of area (average income)
Making choropleth look goodhttp://http://www.colorbrewer.orgwww.colorbrewer.org
•Color contrast•Color blind okay•Several types of data
•Sequential•Qualitative•Diverging
Interesting, but useful color choices Classification methodsClassification methods
•• Jenks (Natural breaks)Jenks (Natural breaks)•• Standard deviationStandard deviation•• QuantileQuantile•• Equal intervalEqual interval•• CustomCustom
11
Jenks (natural breaks)Jenks (natural breaks)•• Based on natural Based on natural
groupings of datagroupings of data
•• Look for Look for ““gapsgaps”” between between clusters of valuesclusters of values
QuantileQuantile•• Each class has the Each class has the
same number of same number of features (e.g., 10 features (e.g., 10 polygons in each)polygons in each)
•• Total # of different data Total # of different data values / # classesvalues / # classes
•• Good for ordinal level Good for ordinal level data (e.g., high, data (e.g., high, medium, low)medium, low)
Equal intervalEqual interval•• Each class has the same Each class has the same
range of values (1range of values (1--10, 1110, 11--20, 2120, 21--30,30,……))
•• Best applied to data with Best applied to data with familiar ranges (e.g., familiar ranges (e.g., percentages)percentages)
•• Good with evenly Good with evenly distributed valuesdistributed values
Standard deviationsStandard deviations•• Shows the amount a Shows the amount a
feature varies from the feature varies from the meanmean
•• Good for looking at the Good for looking at the ““extremesextremes””
lower48POP90_SQMI
5 - 161
162 - 491
492 - 1030
1031 - 9187
lower48POP90_SQMI
5 - 32
33 - 78
79 - 189
190 - 9187
Comparing classificationsComparing classificationsJenks Standard deviations
Quantile
lower48POP90_SQMI
< 1 Std. Dev.
1 - 2 Std. Dev.
2 - 3 Std. Dev.
> 3 Std. Dev.
lower48POP90_SQMI
5 - 2301
2302 - 4596
4597 - 6892
6893 - 9187
Equal Interval
PopulationDensity
PopulationDensity
PopulationDensity
PopulationDensity
Population density of Indiana (1980)
Outliers are a problem for choropleth- Many counties have low density, few have high
1) How many classes?2) What variation is most important? (map purpose/intention)
* *
12
Equal interval – contrast too high Quantile – contrast too low Natural Breaks – too much emphasison extremes
Optimal (Jenks) – good balance
Time series dataTime series data
•• What if your data isnWhat if your data isn’’t just a onet just a one--time time snapshot?snapshot?
•• What if you want the reader to What if you want the reader to interact interact with your map?with your map?
Projections
• Projection choice affects the appearance of your map
• Projection also affects measurements
Attractive visualization of dataAttractive visualization of data
Gall - Peters Projection…somewhat reminiscent of wet, ragged long winter underwear hung out to dry on the Arctic Circle –Arthur Robinson
Perhaps a bit less like wet, ragged long winter underwear…
Goode’s Homolosine Projection
Why does projection matter? Why does projection matter?Appropriate visualization of dataAppropriate visualization of data
Mercator projection Mollweide (equal-area)
What if these maps were showing population density???
13
Where does projection matter? Where does projection matter?
Tools
• ArcMap– Data generation, light map-making
• Other graphic software – Illustrator, Freehand, Photoshop (in Star
Lab)• Macromedia Flash
– Time series data, interactive maps