1.2.3. GeoVisualization Mashup

Information

Geovisualization frequently requires us to consider spatial data of different types from a variety of 'formal' and 'informal' sources using vector and raster data models at various scales. In this tutorial we have used KML to specify 'maps' that are cartographically and geographically informed and have seen how spatial data that are transformed and manipulated in a GIS can be coded in KML .

Geobrowsers such as Google Earth allow us to visually integrate these various inputs into our geovisualization. They support the exploratory process with ancillary data from a variety of sources and some useful interactions.

Using KML to specify the geometry, cartography and structure of our data allows us to visualize it in Google Earth for consideration and visual comparison with a range of ancillary data (such as recent air photography, place names, road networks and data from the 'Geographic Web'.).

Using Google Earth allows us to explore detail in a number of ways : spatial detail by zooming in and 'geographic filtering' as well as attribute detail by inspecting individual items and requesting 'details on demand'. Despite its rich source of data and quick interactive interface Google Earth does not have sophisticated geographic information processing functionality currently, but using a GIS such as LandSerf allows us to produce visual output in forms that are acceptable to the geobrowser .

Our argument is that the mashup creates a whole that is greater than sum of parts as surfaces can be compared with point, line and area-based data, ancillary photographic data and multimedia from informal and formal sources. The KML / Google Earth combination provides the 'glue' to integrate data sources and facilitate interaction. It allows us to filter by space, time and attribute to support this process and provides a useful and flexible environment for the kind of open exploration that is typical of geovisualization in which data and functionality needs change rapidly (Dykes 2005). These capabilities may be useful for developing geovisualization mashups in applied situations (Wood et al. 2007).

Example

To use the analysis performed by LandSerf in Google Earth we need to export the new surfaces from LandSerf into a format Google Earth can read. You may have noted that the last line of the LandScript file in the previous section created a file called flickrChi.kmz. A .kmz file is simply a compressed version of one or more KML files. In this case the file contains two compressed files - image.png and image.kml. The former is a high resolution version of the map explored in LandSerf, the latter provides the georeferencing instructions for Google Earth. It is shown below :

image.kml

<?xml version="1.0" encoding="UTF-8"?>

<kml xmlns="http://earth.google.com/kml/2.1">
<Document xmlns:xlink="http://www.w3/org/1999/xlink">

<Region>
<LatLonAltBox>
<north>49.860764</north>
<south>24.985762</south>
<east>-66.68767</east>
<west>-124.97933</west>
</LatLonAltBox>
<Lod>
<minLodPixels>256</minLodPixels>
<maxLodPixels>-1</maxLodPixels>
</Lod>
</Region>

<GroundOverlay>
<name>Chi expectation surface</name>
<drawOrder>1</drawOrder>
<Icon>
<href>image.png</href>
</Icon>
<LatLonBox>
<north>49.860764</north>
<south>24.985762</south>
<east>-66.68767</east>
<west>-124.97933</west>
</LatLonBox>
</GroundOverlay>

</Document>
</kml>

The GroundOverlay element is used to position the image. In this example LandSerf also creates a Region that is used to define the area in which the image is viewed. In this case, the region simply repeats the georeferencing information used by the GroundOverlay, but Regions can be used to define a set of multiple images to be displayed depending on where the user's viewpoint is relative to the ground surface. In particular the Lod ('Level of Detail') element is used to determine which images are displayed depending on how many pixels wide an image would be drawn in Google Earth. This is the technique that is used by Google Earth to replace coarse satellite images with more detailed air photos as you zoom in towards a target.

A Googe Earth GIS Mashup

This final example integrates the various data sources that we have considered in this tutorial, and the exercise gives you the opportunity to explore them in Google Earth .

The following data have been generated and assembled for geovisualization in Google Earth and constitute the mashup example that you will be using in the exercise :

Flickr point data showing locations and times. The related photos can be viewed by clicking on the point icons in Google Earth
Population counts, densities and change by state, county and tract
Population density surface generated by LandSerf
Flickr and iPhone density surfaces generated by LandSerf
Flickr and iPhone chi distribution surfaces generated by LandSerf through LandScript
In addition, Google Earth provides aerial photography / satellite images and a range of data sets available from the 'Layers' window, including access to the 'Geographic Web'.

The figure below shows an example of using the Chi distribution and Flickr point data in Google Earth. General spatial patterns of the observed and expected photo distributions can be compared with other geographic data. Additionally, individual photographs may be queried as part of the data exploration process.

Flickr and GIS surface data displayed in Google Earth

Exercise

Start Google Earth and open FlickrChi.kmz and iPhoneChi.kmz created by LandSerf in the previous exercise.
Download >> http://www.gicentre.org/tutorials/gisGE/exercise3.kml
This KML file contains the data described above in the example section.
The data are stored in different files. NetworkLink elements are used to integrate the data in a single downloadable KML file.
Now we have access to a variety of spatial data in a 'mashup' we can use it to participate in some geovisualization.
Use the data to ...
- suggest relationships between the data from our two main sources : aggregated population counts and point-based photo locations.
- explore any variations in such relationships across space and over time.
- explore any variations relating to the scale at which we consider these phenomena.
- assess the effects of errors in the data generation process (e.g. can you explain the very low chi value near Lebanon, Kansas?).

You may find it useful to do so by reconsidering the questions that we posed at the outset ...

Can you suggest geographic relationships / trends in the individual data sets - do these vary geographically?
Can you identify errors or anomalies in the individual data sets - do these vary geographically?
Can you suggest geographic relationships / trends between the data sets - do these vary geographically?
Does viewing the data in this way help you generate ideas?

Notes

We believe that there is plenty of scope for using KML to describe our data in ways that are informed by principles and techniques from GI Science . We encourage evaluation and reflection about the benefits and problems associated with undertaking and supporting geovisualization in this way.

The vector representations used here in particular have limited scalability - there is a need to 'slice' the data as we have done here according to time, geography and 'interestingness'. Whilst Google Earth allows us to change scale rapidly and load appropriate data, switching between overview, filter and detail is cognitively taxing as well as technically difficult. The continual use of the Internet in this synchronous tutorial has exacerbated the issue. If you would like to download a larger sample of the Flickr data set then we encourage you to do so after any scheduled teaching has finished.

Download >> http://www.gicentre.org/infovis/data/flickr/
Web page provides access to Flickr data.
The data are searchable by term and can be output using the three alternative TimeStamp encodings introduced in the first session.

We conclude that there is considerable scope for supporting geovisualization using technologies and data sources that are open and that contain opportunities for development. Mashups, that combine data and functionality from a range of sources have potential for helping us understand and generate ideas from the data sets to which we have access.

We hope that this short 'hands-on' and example-based introduction has helped introduce you to some of these possibilities and given you some ideas as to how you might develop mashups.

We encourage you to use the examples provided here and the various references to inform your own visualization mashups. We would be pleased to hear about your work.