CommonGIS: A Software System for Exploratory Data Analysis

1
CommonGISA Software System for Exploratory Data
Analysis

• Tutorial
• Presenters Natalia and Gennady Andrienko
• FHG AIS (Fraunhofer Institute for Autonomous
  Intelligent Systems)
• http//www.ais.fraunhofer.de/and
• ISESS conference, Sesimbra Portugal, 24.05.2005

2
What Is CommonGIS?

• Originates from an EU project CommonGIS
• Has some standard GIS features, e.g. can draw
  maps
• However, not a common GIS but an innovative
  software system specially designed for
  Exploratory Data Analysis
• First of all, for spatially referenced data
• And also for temporally referenced data

3
What Is EDA?

• Emerged in statistics in 1970ies originator
  John Tukey
• A philosophy and discipline of unbiased looking
  at data What can data tell me? rather than Do
  they agree with my expectations?
• Similar to the work of a detective (J.Tukey)
• Tools first of all, visualisation
• Main principles
• look from multiple perspectives (? multiple
  linked displays)
• play with data (? interactive, manipulable
  displays, data transformation facilities)

4
CommonGIS Main Features (1)

• Intended for spatially referenced data ? offers a
  variety of thematic mapping techniques
• Maps are highly interactive and allow the user to
  manipulate data representation
• Not just for fun but for effective analysis
• Not only maps

5
CommonGIS Main Features (2)

• A variety of well-integrated tools for EDA
• Maps statistical graphics several mechanisms
  of display coordination
• Designed to gain synergy of
• Visualisation
• Display manipulation
• Data manipulation
• Querying
• Computational techniques

6
CommonGIS Specials

• Original tools for analysis of spatial data in
  raster (grid) format
• Advanced tools for analysis of spatio-temporal
  data
• Spatially distributed events (e.g. earthquakes)
• Space- and time-referenced thematic data (e.g.
  ecological monitoring or modelling)

7
See a live demo
8
Tools For Raster (Grid) Data Analysis
9
Raster (Grid) Data
Typically, represent continuous phenomena. Values
are measured at sample locations, usually in
nodes of a regular rectangular grid.
10
Grid Data Properties

• One grid contains values of a single attribute.
• The values refer to regularly spaced locations or
  uniform rectangular cells rather than to any real
  spatial objects or to administrative units.
• Values between grid nodes are derived by means of
  interpolation.

11
Grid Data Visualisation (1)
color encoding
Each grid value is encoded by colour. Various
rules of colour encoding (colour scales) may be
used.
12
Grid Data Visualisation (2)
This is an example of applying different colour
scales to the same data.
13
Manipulating Grid Visualisation 1
Interactive focusing only a selected value
interval is shown.
14
Manipulating Grid Visualisation 2
Interactive variation of the midpoint of the
diverging colour scale.
15
Some Problems with Grids
How to visualise and analyse multiple grids?
We can see only the topmost grid layer
Transparency does not really help
16
Interactive Grid-Table Bridge (1)
Step 1. Introduce an artificial territory
division. It is convenient to have uniform
areas, e.g. equal-size rectangles.
17
Interactive Grid-Table Bridge (2)
Step 2. Derive attributes of the rectangular
areas from the grids by means of aggregation
mean, median, minimum, maximum, etc. from the
grid values fitting in each cell
18
Interactive Grid-Table Bridge (3)
Step 3. Visualise and analyse the resulting
attributes as usual
19
Interactive Grid-Table Bridge (4)
The sizes of the rectangles (grid-table bridge
resolution) can be interactively changed. The
attributes are automatically re-computed and the
displays updated.
20
Further Possibilities for Analysis of
Grid-Derived Attributes
21
See a live demo
22
Forest Scenarios Comparison
23
Dominant Species And Age Class (1)
natural
selective
Russian
illegal
24
Dominant Species And Age Class (2)
natural
selective
Russian
illegal
25
Species Structure (1)
26
Species Structure (2)
27
Age Structure (1)
28
Age Structure (2)
29
Look at the German weather (live demo)
30
Recap What Has Been Shown (1)

• Interactive map displays applying various
  thematic cartography techniques
• area painting
• graduated symbols
• charts bars, pies, mosaics,
• classifications single attribute,
  cross-classification, multiple attribute
  (dominance)
• Map manipulation techniques
• outlier removal
• visual comparison
• modification of classes

31
Recap What Has Been Shown (2)
Dot plot
Dispersion graph
Distribution of attribute values within a range
Frequency histogram
Scatter plot shows how two attributes are related
Parallel coordinates object characteristics
profiles relationships between attributes (look
at line slopes)

• A variety of display types

32
Recap What Has Been Shown (3)
Let us examine characteristics of districts in
this area
Two distinct clusters in the value space of these
two attributes
The characteristics in terms of the upper 4
attributes are rather coherent
The values of this attribute greatly vary
Enclose the area in a frame
The values of this attribute are split in two
groups with a gap between
The districts fit in the left half of the
histogram, mostly in bars 1 and 4

• Display linking and coordination

33
Recap What Has Been Shown (4)

• Data manipulation
• Comparison to overall mean, median, etc.
• Temporal comparison previous moment, selected
  moment
• Normalisation (z-score)
• Aggregation, e.g. summing all age classes
• Raster ? table

34
Recap What Has Been Shown (5)
Temporal and thematic filters
Now we see only the earthquakes with magnitudes 4
and more that occurred at depths not less than 20
meters during the year 1999

• Dynamic querying

35
Recap What Has Been Shown (6)

• Computational tools (e.g. cluster analysis)

36
Recap What Has Been Shown (7)

• Specialised tools for spatio-temporal data

37
Conclusion

• All this is NOT about CommonGIS but about
  Exploratory Data Analysis!
• Main principles
• look from multiple perspectives
• play with data
• Tools for EDA
• Visualisation
• Display manipulation
• Data manipulation
• Querying
• Computational techniques
• Tool combination is important!

38
To Learn More

• Tutorials and demos look at http//www.ais.fraunh
  ofer.de/and
• Publications (see the list on the same site)
• Book to appear (? end 2005)
• N. and G. Andrienko
• Exploratory analysis of spatial and temporal
  data
• (Springer-Verlag)