MSc in Geographical Information Science 2008

Using KML for Thematic Mapping

Awarded with Distinction

Bjørn Sandvik

Institute of Geography
School of GeoSciences
University of Edinburgh
Primary supervisor:
Bruce Gittings,
Institute for Geography, School of GeoSciences, University of Edinburgh

External supervisor:
James Reid,
EDINA, University of Edinburgh

Copyright of this dissertation is retained by the author and The University of

Edinburgh. Ideas contained in this dissertation remain the intellectual property of the
author and their supervisors, except where explicitly otherwise referenced.

Part 1: Research Paper

All rights reserved. The use of any part of this paper reproduced, transmitted in any
form or by any means, electronic, mechanical, photocopying, recording, or otherwise
or stored in a retrieval system without the prior written consent of the author and The
University of Edinburgh (Institute of Geography) is not permitted.

Part 2: Supporting Document

The Supporting Document is available under a Creative Commons Attribution-Share
Alike License Version 3.0. The Thematic Mapping Engine (TME) and the
thematicmapping.org website solely remain the intellectual property of the author.
The TME source code is available under a GNU General Public License Version 3.

Statement of originality and length

“I declare that this dissertation represents my own work, and that where the work of
others has been used it has been duly accredited. I further declare that the length of
the components of this dissertation is 5050 words for the Research Paper and 8807
words for the Supporting Document.”

13 August 2008
 Part 1: Research Paper

Using KML for Thematic Mapping

Bjørn Sandvik

Institute of Geography
School of GeoSciences
University of Edinburgh

Abstract

The use of geobrowsers has increased considerably over the last few years. Thematic
mapping has a long history in cartography, but the new geobrowsers (like Google
Maps and Earth) tend not to focus on this aspect of geographical information
representation. This paper examines how Keyhole Markup Language (KML) can be
used for thematic mapping. KML is not targeted towards thematic mapping, but it is
possible to use KML elements in ways that were probably not intended. Current
possibilities for making proportional symbol maps, chart maps, choropleth maps and
animated maps with KML will be presented. These experiments show that KML and
geobrowsers offer great potential for thematic mapping, but that there are significant
issues that need to be resolved.

Keywords: Cartography, Visualisation, KML, Keyhole Markup Language, Thematic

mapping, Google Earth, GIS
Research Paper Bjørn Sandvik

1. Introduction

The purpose of this paper is to examine how KML, Keyhole Markup Language
(Wilson 2008), can be used for thematic mapping. By using KML, thematic maps can
be visualised in accessible and powerful geobrowsers like Google Earth. Thematic
mapping has a long history in cartography, but the new geobrowsers tend to have a
stronger focus on detailed satellite imagery and general-reference maps than on more
abstract data sources.

“From the perspective of the social sciences, the focus on content that is visible from
above is problematic, given the abundance of more abstract data sources. A new
generation of techniques is needed that can mash such data with the Google Earth base,
creating more powerful ways of communicating what social scientists know about the
surface of the Earth.”

Michael F. Goodchild 2008:23

This paper shows how the current KML

standard (Version 2.2; Wilson, 2008) can
Thematic
be used for common thematic mapping mapping
KML
techniques. KML recently became an standard
techniques
official standard of the Open Geospatial
Consortium1 (Bacharach, 2008).
Becoming an international standard will
Visualisations
bring KML to centre-stage and greatly
broaden its use and general acceptance.

A fully functional Thematic Mapping Public Geo-

datasets browsers
Engine (TME) has been developed as a
proof-of-concept. The application allows
the user to make thematic maps through
an easy-to-use web interface2. The KML
examples shown in this paper were created using the TME and displayed in Google
Earth. TME is further described in the Supporting Document (Sandvik, 2008).

This paper begins by discussing why geobrowsers are appropriate tools for thematic
mapping, followed by an introduction to the Keyhole Markup Language (KML) and
the most common thematic mapping techniques. The paper then proceeds to describe
and discuss how proportional symbol maps, chart maps, choropleth maps and
animated maps can be created with KML.

1
The Open Geospatial Consortium (OGC) is an international industry consortium of 368 companies,
government agencies and universities participating in a consensus process to develop publicly available
standards for geospatial and location based services (www.opengeospatial.org).
2
http://thematicmapping.org/engine

2
Research Paper Bjørn Sandvik

2. Background

It is common among cartographers to distinguish between general-reference maps3

and thematic maps (Slocum et al., 2005). A general-reference map emphasises the
location of a variety of different features, such as coastlines, lakes and roads
(Robinson et al. 1995). A thematic map displays the spatial pattern of a social or
physical phenomenon, such as population density, life expectancy or climate change.
Thematic mapping has a long history in geography (MacEachren, 1979), and thematic
maps can readily be created with a wide range of Geographical Information Systems
(GIS) and specialist software. Yet, the new geobrowsers have tended not to focus on
this aspect of geographical information representation.

2.1 Why use geobrowsers for thematic mapping?

Powerful and accessible tools for geographical visualisation have appeared in the last
few years. Google Earth, Google Maps, Microsoft Virtual Earth, NASA World Wind
and ArcGIS Explorer (see table 1) are all tools that are freely available to the public.
The term geobrowser is used throughout this paper to describe these tools. All
geobrowsers are capable of accessing georeferenced4 data over the internet, and the
view can be two and/or three dimensional. A geobrowser can be a desktop program or
an application embedded in the web browser.

Proprietary / Performance
Geobrowser 2-D / 3-D KML support
Open Source (vector handling)
5
Google Earth Proprietary 3-D Full Good
6
Google Earth Plug-in Proprietary 3-D Intermediate Good
7
Google Maps Proprietary 2-D Basic Intermediate
Microsoft Live Search Proprietary 2-D Basic Intermediate
8
Maps
9
Microsoft Virtual Earth Proprietary 3-D Basic Good
10
ArcGIS Explorer Proprietary 2-D / 3-D Basic Good
11
OpenLayers Open Source 2-D Basic Intermediate
12
NASA World Wind Open Source 3-D Basic Good
13
ossimPlanet Open Source 3-D Basic Good

Table 1: Common proprietary and open source geobrowsers with KML support. It is unsurprising that
Google Earth has the best KML support, as it was where KML originated. The vector graphic
rendering in current web browsers is too slow to handle large KML documents with many features,
although newer web mapping clients based on various add-ons, like the new Google Earth plug-in,
achieve better performance.

3
Large-scale general reference maps of land areas are usually called topographic maps (Robinson,
1995)
4
To georeference means to relate information to a geographic location (Hill, 2006).
5
http://earth.google.com
6
http://code.google.com/apis/earth
7
http://maps.google.com
8
http://maps.live.com/
9
http://www.microsoft.com/virtualearth
10
http://www.esri.com/software/arcgis/explorer
11
http://openlayers.org
12
http://worldwind.arc.nasa.gov
13
http://www.ossim.org/OSSIM/ossimPlanet.html

3
Research Paper Bjørn Sandvik

Geobrowsers have attracted millions of users amongst the general public, who are
seduced by the ease with which they can zoom from space right down to street level.
Geobrowsers are also used by a growing number of scientists as they enable data with
a spatial component to be overlain on top of the background imagery (Butler, 2006).
Thus, geobrowsers have been more oriented towards supporting satellite imagery and
general-reference maps (i.e. roads, buildings and political borders) than thematic
maps. Due to their huge public interest and relative accessibility, geobrowsers are
potentially capable of bringing thematic maps or visualisations to a wider audience.

A great advantage of the new geobrowsers is the possibility of adding and visualising
user generated content. This can be done through an application programming
interface (API) or XML14 scripting. An API consists of different pre-built
functionalities which can be used by programmers to create interactive map
applications (Gibin et al., 2008). XML scripting is described in subsequent paragraphs
relating to the KML standard.

By allowing the user to visualise statistical data in an external geobrowser instead of a

built-in mapping application, data from various sources can be more readily
combined. The value of the statistical data can be increased if the user is allowed to
tie it to other data sources. This can be both quantitative and qualitative data, like text,
pictures and videos. Furthermore, geobrowsers can also be embedded in other web
applications to enhance their utility and general ability to contextualise information.

Using geobrowsers for thematic mapping can, however, be problematic. For 2-D
geobrowsers, the big players in the web mapping world (Google Maps and Microsoft
Live Search Maps) use the Mercator projection15. Mercator is a good choice for
zoomable satellite imagery, but less suitable for thematic world maps due to the great
area distortions. Current 2-D viewers also have problems rendering large vector-based
maps, due to web browser limitations (Hudson-Smith et al. 2007; LeMay, 2005).

3-D virtual globes, like Google Earth and Microsoft Virtual Earth, partly avoid the
distortions of map projections. Since images still have to be projected onto a flat
screen, virtual globes use the Perspective Orthographic projection, which is also the
projection inherent in the human visual system (Goodchild, 2008). Unfortunately, the
ability to see the whole Earth at once is lost when thematic maps are rendered on a
globe. It is also difficult to estimate the area or the volume of proportional symbols
when seen in perspective (Shepherd, 2008).

2.2 Keyhole Markup Language (KML)

KML is an XML grammar used to encode and transport representations of
geographical data for display in a geobrowser (Wilson, 2008). KML was originally
created as a file format for Keyhole's Earth Viewer, which later emerged as the
Google Earth application (Ratliff, 2007), allowing users to overlay their own content
on top of the base maps and satellite imagery. In 2007, Google submitted KML to the
Open Geospatial Consortium (OGC). KML was adopted as an OpenGIS standard in
14
Extensible Markup Language (XML) is a general purpose specification for creating custom markup
languages (www.w3.org/xml). XML facilitates the sharing of structured data across different
information systems, particularly over the Internet. KML is an XML-based language.
15
OpenLayers is an open source geobrowser that supports different projections. http://openlayers.org

4
Research Paper Bjørn Sandvik

2008 (Wilson, 2008), and the OGC has now the responsibility for maintaining and
extending the standard.

KML is focused on visualisation of geographic features on map or a globe. The XML

language also includes controls of the user’s navigation in the sense of where to go
and where to look (Wilson 2008). However, thematic mapping or thematic
cartography are not mentioned in the 233 page OGC KML standard document
(Wilson, 2008), suggesting KML is not targeted towards this kind of usage.

In common with other XML

grammars, KML uses a tag-based
structure with nested elements and
attributes. A KML document is
processed in a similar way that
HTML documents are processed by
web browsers. The basic building
blocks of the language are called
elements, and a tag is the way an
element is represented as KML
code (Crowder, 2007).

KML documents and their related

images and 3-D objects (if any)
may be compressed using ZIP
encoding into KMZ files (Google,
2008a). This greatly reduces the file
size and makes data transfer more
efficient, overcoming one of the Figure 1: Diagram showing the hierarchy of KML elements
major criticisms of XML-based (Google, 2008a).
structures.

2.3 Alternatives to KML

KML was chosen due to its popularity and simplicity, and because the standard is
supported by many of the popular geobrowsers. This section compares KML with
other standards of interest.

OpenGIS Geography Markup Language (GML) is an XML grammar which helps in

the storage, exchange and modelling of geographical information containing both
spatial and non-spatial attributes (Portele, 2007; Lu et al., 2007). The encoding is
comprehensive in the way in which it can represent features with complex 3-D
geometry, features with 2-D topology, dynamic features and coverages (Lake and
Farley, 2007; Peng and Zang, 2004). The uptake of GML in the mainstream web
community has been slow, primarily because GML is viewed as a complex format
(Reed, 2006; Torkington, 2006; Lu et al., 2007). GML is not a visualisation language;
it does not provide any information regarding how the data is to be displayed.

OpenGIS Styled Layer Descriptor (SLD) and Symbol Encoding (SE) are two related
XML languages for styling information (Lupp, 2007; Müller, 2006). SLD/SE is
capable of describing the rendering of vector and raster data. Only lines, polygons,

5
Research Paper Bjørn Sandvik

points, text and raster images are integrated in the description language, which makes
it impossible to visualise multiple data values, e.g. pie charts or bar charts (Sykora et
al., 2007). Various proposals exist for an OGC Styled Layer Descriptor/Symbology
Encoding extension for thematic cartography (Dietze and Zipf, 2007; Sae-Tang and
Ertz, 2007; Iosifescu-Enescu, 2007).

Scalable Vector Graphics (SVG) is an XML based web standard for 2-D vector
graphics from the World Wide Web Consortium (Jackson, 2003). SVG is also a
suitable format for GIS and mapping applications (Dunfey et al., 2006). While GML
provides a means of storing and transporting geographical features, SVG makes it
possible to display these features as vector maps (Peng and Zhang, 2004). The
graphics description capabilities of SVG are much stronger than those of KML, but
SVG is only 2-D and does not incorporate concepts for navigation (Lake, 2008).

Compared to these standards, KML does a bit of everything. It defines geographic

objects, their styling and their graphical representation. There are some overlaps
between GML and KML in the way in which the basic geometrical objects are
represented (Lake, 2005). KML contains styles, but is not a styling language in the
manner of SLD/SE. In SVG, the canvas is the 2-D surface of a computer screen,
whereas KML provides the mechanisms for visualising geographical features on a
map or a globe.

GML and KML are also logical partners, like GML and SVG. SLD/SE provides
styling rules to transform data encoded in GML into a target visualisation language
(e.g. KML or SVG). Future versions of KML may be harmonised with other relevant
OGC standards to encourage a broader implementation and greater interoperability
(Wilson, 2008; Reed, 2007; Lake, 2008).

2.4 Data as a public good

One of the biggest bottlenecks in using KML and geobrowsers for thematic mapping
is the lack of data access and restrictive licensing rules. Many data providers are
concerned about reverse engineering, and do not readily permit open and human-
readable formats to be used. The licensing terms of many data providers are also in
conflict with the terms and conditions of geobrowser vendors. Google requires
applications that use their APIs to be freely accessible to end users, without any
subscription or restricted access (Google, 2008b). Even though there are initiatives
towards open data and open access (Onsrud et al. 2004; OKF, 2008; Quint, 2004),
these are still highly controversial issues and the policies vary greatly across the world
(Weiss, 2002).

A new data portal from the United Nations turned out to be the solution for this
project. The United Nations Statistics Division (UNSD) has experienced a dilemma
between meeting an official and public need for free access to information, and
maintaining the potential for earning revenues through sales (Statistical Commission
2007). UN member states and data customers have insisted that statistical products
should be available free of charge, and this concern resulted in the launch of a
“Statistics as a Public Good” project in 2005. The objective of this project was to
develop a new internet based data service whereby UNSD would increase the

6
Research Paper Bjørn Sandvik

dissemination, use and understanding of multidisciplinary statistical data. The new

UNdata portal was launched in February 2008 under a liberal licensing policy:

“All data and metadata provided on UNdata’s website are available free of charge and
may be copied freely, duplicated and further distributed provided that they are not put up
for sale or otherwise commercially exploited and that UNdata is cited as the reference.”

(UNdata, 2008)

The UN has played a major role in measuring world phenomena and quantifying the
importance of different human activities, and the creation of a universally
acknowledged statistical system is “one of the great and mostly unsung successes of
the UN organisation” (Ward 2004:2). By adopting a “data as a public good” policy,
the UN allows international statistics to be utilised in creative new ways by third
parties.

2.5 Thematic mapping techniques

MacEachren (1979) has suggested that thematic cartography has exerted a substantial
influence on the development and dissemination of knowledge. With the advent of
national censuses, and the growth of social sciences, thematic maps have become a
primary mechanism for summarising and communicating the increased volumes of
geographically related information.

Jaques Bertin (1967) established a graphic system of visual variables, which

represents an indispensable and universally recognized theory of the cartographic
transcription of geographical information (Koch, 2001). Visual variables describe the
perceived differences in map symbols that are used to represent geographical
phenomena (Slocum et al., 2005). Bertin’s system has been subsequently modified by
various cartographers (Koch, 2001), and the visual variables used in this paper are
based on Slocum et al. (2007), which add 3-D symbolisation.

Cartographers commonly distinguish between point, line, area and volume

symbolisation (Robinson et al., 1995; Slocum et al., 2005). These distinctions may be
summarised as follows.

A point symbol refer to a particular location in space, and is used when the
geographical phenomena being mapped is located at a place or is aggregated to a
given location (MacEachren, 1979). Differentiation among point symbols is achieved
by using visual variables, like size, colour and shape. Common thematic mapping
techniques using point symbols are dot maps and proportional symbol maps. On a dot
map one dot represents a unit of some phenomena, and dots are placed at locations
where the phenomenon is likely to occur (Slocum et al., 2005). A proportional symbol
map is constructed by scaling symbols in proportion to the magnitude of data
occurring at point locations. These locations can be true points or conceptual points,
such as the centre of a country for which the data have been collected. Various ways
of making 2-D and 3-D proportional symbols in KML are discussed later in this
paper.

Line symbols are used to indicate connectivity or flow, equal values along a line and
boundaries between unlike areas (MacEachren, 1979). Line symbols are differentiated

7
Research Paper Bjørn Sandvik

on the basis of their form (e.g. solid line versus dotted line), colour and width.
Common thematic mapping techniques using line symbols are flow maps and
isarithmic maps. Flow maps utilise lines of differing width to depict the movement of
phenomena between geographical locations (Slocum et al., 2005). Isarithmic maps
depict smooth continuous phenomena, like rainfall or barometric pressure (Slocum et
al., 2005). It is beyond the scope of this project to explore ways of creating flow maps
and isarithmic maps in KML, as it requires other types of data.

Area symbols are used to assign a characteristic or value to a whole area on a map.
Visual variables used for area symbols are colour, texture and perspective height
(Slocum et al., 2005). The choropleth map is probably the most commonly employed
method of thematic mapping, and is used to portray data collected for enumeration
units, such as countries or statistical reporting units. While choropleth maps reflect the
structure of data collection units, dasymetric maps assume areas of relative
homogeneity, separated by zones of abrupt change. The country statistics used in this
paper can be considered as areal phenomena, because the statistical values are
associated with political units specified as enclosed regions.

Volume symbols can be considered as 2½-D or true 3-D (Slocum et al. 2005). The first
can be thought of as a surface, in which a geographical location is defined by x and y
coordinate pairs and the value of the phenomenon is the height above a zero point. An
example is prism maps which uses perspective height as the visual variable. 3-D
symbols can be used to represent true 3-D phenomena, like the concentration of
carbon dioxide (CO2) in the atmosphere or geological material underneath the earth’s
surface (Slocum et al., 2005). In this paper, 3-D proportional objects are used to
visualise statistical data.

8
Research Paper Bjørn Sandvik

3. Thematic mapping with KML

It is important to understand the style-handling mechanism when creating thematic

maps with KML. Styles are used to define the visual variables in a thematic map.
KML only supports a limited set of styles (no texture, dotted lines, etc.). A style can
be defined once and assigned an identifier (see example below). It can then be
referenced multiple times within the KML file which defined it or, indeed, within
other KML files. This is called a shared style (Wilson, 2008).

<kml>
<Document>
<Style id='sharedStyle'>
<IconStyle>
<Icon>
<href>files/symbol.png</href>
</Icon>
</IconStyle>
<Style>
<Folder>
<Placemark>
<name>China</name>
<Snippet>1,312,978,855 (2005)</Snippet>
<styleUrl>#sharedStyle</styleUrl>
<Style>
<IconStyle>
<color>e50066ff</color>
<scale>7</scale>
</IconStyle>
</Style>
<Point>
<coordinates>106.514,33.421,0</coordinates>
</Point>
</Placemark>
</Folder>
</Document>
</kml>

This sample KML document shows how to create a proportional symbol in KML (see section 3.1.1).
The shared styles are defined within the Document element. This style includes a reference to the
symbol image. Each feature (e.g. country) is defined with a Placemark element. This element contains
the name of the feature, a reference to the shared style and styles that are specific for this particular
feature (e.g. colour and scale). The position (i.e. longitude/latitude) of the symbol is defined by the
Point element.

3.1 Proportional symbol maps

The use of proportional symbols is a common thematic mapping method used to
represent numerical data associated with point locations. The examples shown are
conceptual point data (Slocum et al., 2005), which are collected over areas, but
conceived of as being located at points for the purpose of symbolisation. KML lacks
native support for proportional symbols, but it is possible to create such symbols
using three different techniques:

9
Research Paper Bjørn Sandvik

3.1.1 Scaled image icons

The easiest and most efficient way
to make proportional symbols in
KML is by scaling image icons.
KML icons are used to visualise
various point data, and custom
icons can be added by referencing
an image stored on the local file
system or a remote web server.
Vector images, like Scalable Vector
Graphics (SVG), are not supported.

The best way to create a

proportional circle is to reference
Figure 2: AIDS estimated deaths (aged 0-49) in southern
an image file containing a white Africa in 2005, represented as proportional image icons on top
circle on a transparent background. of Google Earth satellite imagery. Statistics from UNdata.
As the image is not vectorised, it
gets a jagged (antialiased) look when given a large scale value. Images used as
proportional symbols should therefore have a larger size than normal map icons.

If a KML Point element is contained by a Placemark element, the point determines

the position of the Placemark’s name and icon. A reference to the icon image can be
specified in the associated style. These images can be scaled and colourised by
changing the style attributes.

These symbols are attached to the ground at just one point, and have a “floating in the
air” appearance. The size of the image icons are not affected by perspective. This is
beneficial as it makes country comparison easier. Another advantage is that icons are
clickable so the user can easily assess the data value by clicking on the symbol.
As this is probably not the originally intended way of deploying image icons, there
are some problems that need to be solved by the geobrowser vendors. In Google
Earth, the image icons are scaled properly when the user zooms in or out, but not
when the size of the Google Earth window is changed. Other geobrowsers are
currently not able to scale or colourise image icons.

3.1.2 Drawing regular polygons

Another way of making proportional symbols in KML is to use regular polygons. A
regular polygon is a polygon whose all sides have the same length and all vertex
angles are equal. KML has no support for regular polygons, so drawing regular
polygons requires the longitude/latitude of each vertex or corner point to be
calculated. A complex function is needed which takes the curvature of the Earth into
account.

When the vertices of the regular polygons are calculated, it can be drawn by using the
KML Polygon element. The polygon size is determined by the position of the
vertices, while the colour can be changed in an associated style. It is very difficult, if
not impossible, to create perfect shapes that look the same in 2-D and 3-D viewers.

10
Research Paper Bjørn Sandvik

The benefit of this technique is that the Polygon element is supported by many
geobrowsers.

Figure 3: Regular polygons drawn on top of the terrain Figure 4: The same KML document visualised in
layer in Google Maps. The squares represent the Google Earth. The squares are skewed due to the
number of internet users in each country (UNdata). curvature of the Earth.

3.1.3 Proportional 3-D Collada objects

The last method of making proportional symbols in KML is to use 3-D Collada
objects. Collada is an open XML format than enables 3-D authoring applications to
exchange digital assets (Collada, 2008). Google Earth has adopted Collada as its
native format for 3-D models. A KML document cannot contain Collada objects, but
it can reference objects stored on the local file system or a remote web server.

The Collada interchange file

format is normally used to make
3-D buildings in Google Earth,
but, since there are no scale
restrictions, they can also be used
for thematic mapping. By using 3-
D objects, it is possible to create
quite visually appealing maps.

Collada objects can be loaded and

scaled by using the KML Model
element. 3-D objects can be scaled
in 3 directions (x, y, z). This
means that they can also be used Figure 5: Mobile phone subscribers in South East Asia in 2004,
visualised using a 3-D Collada object. The volume of the object
as bars. One variable could is proportional to the statistical value (UNdata).
represent the x-y dimension, while
another could represent the z (height) dimension. Collada objects can also be placed
on top of each other to create stacked bars.

Styles are not supported for Collada objects, which makes it difficult to colourise the
objects. Colours have to be changed in the Collada file itself. The objects are currently
not clickable in Google Earth, which makes it hard to provide a route to the
underlying statistical values.

11
Research Paper Bjørn Sandvik

3.2 Chart maps

KML has no built-in support for
charts (e.g. pie charts or bar charts),
but it is possible to utilise an
external charting service. For
example, the Google Chart API16
allows chart generation by
specifying data and formatting
parameters in an URL. The API
returns a PNG image of the chart.
Various chart types are supported,
like pie, bar and line charts. The
most common usage of this API is
to embed charts on a web page, but
Figure 6: Example of how the Google Chart API can be
the charts can also be overlaid on a utilised by the KML Icon element. The charts are scaled
map by including the chart URLs in according to total population and the pie shows the age
a KML document. distribution for each country. Statistics from UNdata.

The charts are loaded by employing the same technique as that used for image icons.
The chart appearance is specified in the URL and generated on-the-fly when the KML
document is rendered by the geobrowser. This technique is dependant on a fast
internet connection, as all the charts have to be loaded from an external web service.
It would be better to include chart functionality as an integral part of the KML
standard.

3.3 Bar maps

KML polygons can be extruded
from the ground. This utility,
combined with the regular
polygon technique described
above, makes it possible to create
3-D bars. This can be achieved by
adding an altitude value to each
vertex of the polygon. This
altitude value, measured in
metres, should be directly
proportional to the statistical
value. The statistical values
represented by bars are easier to
assess than for 2-D and 3-D Figure 7: GDP per capital in European countries in 2006,
symbols. represented as 3-D bars. Statistics from UNdata.

16
http://code.google.com/apis/chart

12
Research Paper Bjørn Sandvik

3.4 Choropleth maps

The thematic mapping
techniques discussed thus far
all relate to various point
symbols. Choropleth maps use
area symbols, where the
shading of each area varies in
sympathy with the data value
associated with it (Krygier and
Wood, 2005).

KML contains two useful

elements for representing
areas: Polygon and
MultiGeometry. The Polygon Figure 8: Unclassed choropleth map showing the infant mortality
element, which can be used to rate in Africa. Statistics from UNdata.
create regular polygons, can
also be used for areas that can be represented by a single feature. Several polygons
can also be wrapped in a MultiGeometry element with one associated style. This is
useful for countries which consist of separated land areas or islands. Polygons can be
filled by specifying colour and transparency within the Style element.

3.5 Prism maps

Another visual variable for
choropleth maps is perspective
height, which produces what is
commonly termed a prism map
(Slocum et al., 2005). An
unclassed prism map will
portray ratios correctly, since a
value twice as large as another
will be represented by a prism
twice as high. A prism map can
be created by adding an altitude
value to each vertex in the
polygon and specifying that the
polygon is to be extruded. Figure 9: Infant mortality rate visualised as a 3-D prism map.
Statistics from UNdata.
There is a problem with how
prism maps are rendered in 3-D geobrowsers, which lead to the so-called polygon
hole problem. Areas having a low value on a statistical indicator will also be
represented with a prism with a low altitude value. Only the vertices of the polygon
are extruded to the given altitude, and not the centre of the geometry. Since the “roof”
of the prism is not following the great circle17, holes appear in the centre due to the
curvature of the earth.

17
The great circle is a circle on the surface of a sphere that has the same circumference as the sphere.

13
Research Paper Bjørn Sandvik

3.6 Animated maps

One commonly used and
effective technique for
visualising geographical
processes is map animation.
By using the time primitives
in KML, animated thematic
maps showing how a variable
changes over time can be
produced. All
features/symbols for a time
unit can be wrapped in a
KML Folder with a TimeSpan
element.

Whenever time is specified in

a KML file, a time slider Figure 10: This visualisation shows the world population for four
appears in the upper right different years (1950, 1975, 2000 and 2050). Statistics from UNdata.
The image also illustrates the polygon hole problem described above.
corner of Google Earth. By
dragging this slider, the user can see how the changes from one year to another.

All features can be animated in KML, but the problem is that features have to be
duplicated for each time step, even though it is only the statistical value that is
changing. The result is large KML files and, consequently, reduced performance.

3.7 Raster images and map tiles

The thematic mapping techniques presented so far exploit the KML’s vector
capabilities. The benefit of using point and area symbology is increased interactivity,
as these features can respond to various user interactions (Gibin et al., 2008). The
symbology itself can also be changed in a capable geobrowser.

This functionality is more

difficult to achieve when
thematic maps are distributed
as raster images. Nonetheless,
a raster approach can be more
suitable when working with
more complex or detailed
data, as multi-vector payloads
often reduce the geobrowser’s
performance. Serving
thematic maps as raster
images might also avoid the
infringement of intellectual
property rights, since it is
impossible to recover the Figure 11: This image of Earth’s city lights (NASA 2008) was
original vector or statistical loaded as raster map tiles into Google Earth. The bars show CO2
emissions from each country (UNdata).
data from image tiles (Gibin
et al., 2008).

14
Research Paper Bjørn Sandvik

The KML GroundOverlay element can be used to draw an image overlay in KML.
This element includes a reference to an image stored on a local file system or on a
web server. The image is aligned to the ground by specifying the longitude and
latitude positions of the bounding box. Thematic maps covering a large area should be
partitioned into an image pyramid, a hierarchical set of images, each of which is an
increasingly lower resolution of the original image. Every image in the pyramid is
subdivided into tiles that are stored separately. Image tiles that fit together like this are
said to tessellate, and the tiling patterns they create are called tessellations (Johnson
and Neill, 2003). Image tessellation increases the geobrowser performance because it
only loads the portions of the map image that are in view (Wilson, 2008).

By using this technique, it is possible to combine datasets from different sources.

Figure 11 shows how data of CO2 emissions is visualised on top of a raster image of
Earth’s city lights (NASA, 2008). Various base maps can also be supported, so the
user can more easily distinguish between the map symbols and the background
information.

3.8 Map legends

Map legends are indispensable in thematic mapping, but KML has no built-in legend
support. The legend explains the map symbols and the variable range, and allows the
user to interpret the map. The best way of adding a legend is to use the ScreenOverlay
element. With this element you can add an overlay that is displayed within the
geobrowser. Unfortunately, this element only supports image overlays and not HTML
layers. This makes it harder to create a dynamic legend that reflects changes in the
statistical data. It is also difficult to make a symbol legend in KML, as symbol size
varies with scale (zoom). One possible solution is to duplicate symbology and add a
colour legend for all thematic mapping techniques. The colour legend informs the
user about the value range (min and max), and where the colourised symbols are
positioned on this range.

4 Conclusion

It is possible, with some inventiveness, to use KML for thematic mapping. There are,
at least, three different ways of making proportional symbol maps: by scaling image
icons, by calculating the longitude/latitude vertices of regular polygons, or by scaling
3-D Collada objects. Furthermore, an external web service, like Google’s Chart API,
can be utilised to create chart maps. Choropleth maps can be generated by using
shaded polygons, and turned into prism maps by adding an altitude value to each
coordinate tuple. The KML time primitives can be used to create temporal or
animated maps. There is also an option to distribute thematic maps as (raster) map
tiles, instead of via a vector format.

However, the current KML standard is not optimised for thematic mapping, as it was
not explicitly designed for this purpose. Many of the techniques presented in this
paper can be characterised as “hacks”, because KML elements are used in ways for
which they were probably never originally intended. There are consequently issues
that need to be resolved in future KML versions, like the “polygons-hole problem”
that occurs when prism maps are rendered on a 3-D globe. There is also a lot of

15
Research Paper Bjørn Sandvik

redundancy in the data, as height (altitude) values have to be repeated for each vertex
in a polygon, and entire feature structures need to be repeated for every time step in
an animation. Proportional symbols would be easier to create if KML supported
regular polygons natively. Collada objects would be more suitable for thematic
mapping if they were made clickable.

KML needs to be harmonised with other OGC standards to achieve greater

interoperability and more extensive use. Particularly important for thematic mapping
are styling rules for transforming feature and attribute data into various 2-D and 3-D
representations. Such styling rules are already present in the OGC Symbology
Encoding specification. KML also needs enhanced graphics functionality and
symbolisation constructs. Instead of reinventing the wheel, this functionality could be
modelled after the Scalable Vector Graphics (SVG) specification.

The KML standard, combined with powerful and accessible geobrowsers, like Google
Earth, has a great potential for thematic mapping. Hopefully, this potential will also
be realised by the OGC and geobrowser vendors, and lead to future improvements and
native support for thematic mapping functionality in the open standards promulgated
by OGC and the vendors.

16
Research Paper Bjørn Sandvik

References

Bacharach, S., 2008, “OGC Approves KML as Open Standard”, Geospatial Press Releases. Avalable
online: http://geospatialpr.com/2008/04/14/ogc-approves-kml-as-open-standard [Last accessed:7
August 2008]

Bertin, J., 1967, “Semiologie Graphique”, Paris

Butler, D., 2006, “Virtual Globes: The web-wide world”, Nature volume 439, 16 February 2006,
Nature Publishing Group.

Collada, 2008, “COLLADA”, The Khronos Group Inc. Available online:

http://www.collada.org/mediawiki/index.php/COLLADA [Last accessed: 24 July 2008]

Crowder, D. A., 2007, “Google Earth for Dummies”, Wiley Publishing.

Dietze, L., Zipf, A. 2007, “Extending OGC Styled Layer Descriptor (SLD) for Thematic Cartography -
Towards the ubiquitous use of advanced mapping functions through standardized visualization
rules”, 4th Int. Symp. on LBS and Telecartography. Available online:
http://www2.geoinform.fh-mainz.de/~zipf/Thematic-SLD.LBS-Telecarto2007.pdf [Last
accessed: 8 August 2008]

Dunfey, R.I., Gittings B.M., Batcheller, J.K., 2006, “Towards an open architecture for vector GIS”,
Computers & Geosciences, pp. 1720-1732

Galdos, 2008, “KML released as an OGC Specification”, Galdos Systems Ic. Available online:
http://www.galdosinc.com/archives/498 [Last accessed: 13 June 2008]

Gibin, M., Singleton, A., Milton, R., Mateos, P., Longley, P., 2008, “An Exploratory Cartographic
Visualisation of London thorough the Google Maps API”, Applied Spatial Analysis and Policy
(2008) 1:85-97, Springer Netherlands

Goodchild M. F., 2008, “What does Google Earth Mean for the Social Sciences”. Book chapter in
Dodge, M., McDerby, M. and Turner, M., 2008, “Geographic Visualization: Concepts, Tools
and Applications”, Wiley

Google, 2008a, “KML Reference”. Available online:

http://code.google.com/apis/kml/documentation/kmlreference.html [Last accessed 13 June
2008]

Google, 2008b, “Google Maps API Terms of Service”, Google Code. Available online:
http://code.google.com/apis/maps/terms.html [Last accessed: 8 August 2008]

Hill, L. L., 2006, “Georeferencing. The geographic information associations of information”, MIT
Press

Hudson-Smith, A., Milton, R., Batty, M., Gibin, M., Longley P., Singleton, A., 2007, “Public Domain
GIS, Mapping & Imaging Using Web-based Services”, UCL Working Paper Series, Centre for
Advanced Spatial Analysis (CASA). Available online:
http://www.casa.ucl.ac.uk/working_papers/paper120.pdf [Last accessed: 7 August 2007]

Iosifescu-Enescu, I.,2007, “SE Implementation Specification Change Request – extensions for thematic
mapping”, EU FP7 ORCHESTRA Consortium

Kersitng, O., Döllner, J., 2002, “Interactive 3D visualization of vector data in GIS”, Proceedings of the
10th ACM international symposium on Advances in geographic information systems, McLean,
Virginia, USA. Available online: http://portal.acm.org/citation.cfm?id=585170 [Last accessed: 8
August 2008]

17
Research Paper Bjørn Sandvik

Koch, W. G., 2001, “Jaques Bertin’s theory of graphics and its development and influence on
multimedia cartography”, Information Design Journal 10(1), pp 37-43, John Benjamin
Publishing Company

Krygier, J., Wood, D., 2005, “Making Maps: A Visual Guide to Map Design for GIS”, The Guilford
Press

Lake, R., 2005, “GML and KML Syntax”, Galdos Systems Inc. Available online:
http://www.galdosinc.com/archives/183 [Last accessed: 8 August 2008]

Lake, R., 2007, “KML and GML Working Together”, Galdos systems. Available online:
http://www.galdosinc.com/archives/317 [Last accessed: 4 August 2008]

Lake, R., Farley, J., 2007, ‘Infrastructure for the Geospatial Web’. In The Geospatial Web – How
Geobrowsers, Social Software and Web 2.0 are Shaping the Network Society, A. Scharl and
K. Tochtermann (Eds), pp. 15-26, Springer, London

Lake, R., 2008, “KML released as an OGC Specification”, Galdos Systtems Inc. Available online:
http://www.galdosinc.com/archives/498 [Last accessed: 13 June 2008]

LeMay, R., 2005, “Take browsers to the limit: Google”, ZDNet Austrailia. Available online:
http://www.zdnet.com.au/news/software/soa/Take-browsers-to-the-limit-
Google/0,130061733,139204515,00.htm [Last accessed: 7 August 2008]

Lu, C.T., Dos Santos, R., Sripada, L.N., Kou, Y., 2007, “Advances in GML for Geospatial
Applications”, Geoinformatica, 11, pp. 131-157

Lupp, M., 2007, “Styled Layer Descriptor profile on the Web Map Service Implementation
Specification”, Version 1.1.0 (revision 4), OGC 05-078r4, Open Geospatial Consortium.
Available online: http://portal.opengeospatial.org/files/?artifact_id=22364 [Last accessed: 8
August 2008]

MacEachren, A. M., 1979, “The Evolution of Thematic Cartography / A Research Methodology and
Historical Review”, The Canadion Cartographer Vol 16, No 1 June 1979, pp 17-33

Müller, M., 2006, “Symbology Encoding Implementation Specification”, Version 1.1.0 (revision 4),
OGC 05-077r4, Open Geospatial Consortium. Available online:
http://portal.opengeospatial.org/files/?artifact_id=16700 [Last accessed: 8 August 2008]

NASA, 2008, “Earth’s City Lights”, The National Aeronautics and Space Administration (NASA).
Available online: http://visibleearth.nasa.gov/view_rec.php?id=1438 [Last accessed: 28 July
2008]

OKF, 2008, “The Open Knowledge Foundation. Protecting and Promoting Open Knowledge in a
Digital Age”, Website: http://www.okfn.org [Last accessed: 8. August 2008]

Onsrud, H., Camara, G., Campbell, J., Chakravarthy, N. S., 2004, “Public Commons of Geographic
Data: Research and Development Challenges”, Lecture Notes in Computer Science, Volume
3234/2994. Available online:
http://www.dpi.inpe.br/gilberto/papers/commons_giscience2004.pdf [Last accessed: 17
November 2007]

Peng, Z., Zhang, C., 2004, “The roles of geography markup language (GML), scalable vector graphics
(SVG), and Web feature service (WFS) specifications in the development of Internet geographic
informations systems (GIS)”, Journal of Geographical Systems, 6, pp. 95-116

Portele, C., 2007, “OpenGIS Geography Markup Language (GML) Encoding Standard”, version 3.2.1,
OGC 07-036, Open Geospatial Consortium. Available online:
http://portal.opengeospatial.org/files/?artifact_id=20509 [Last accessed: 8 August 2008]

18
Research Paper Bjørn Sandvik

Quint, Barbara, 2004, “OECD Ministers Supports Open Access for Publicly Funded Research Data”,
Information Today. Available online:
http://newsbreaks.infotoday.com/nbreader.asp?ArticleID=16519 [Last accessed: 8 August 2008]

Ratliff, E., 2007, “Google Maps is Changing the Way We See the World”, Wired Magazine, Issue
15.07. Available online: http://www.wired.com/techbiz/it/magazine/15-
07/ff_maps?currentPage=all [Last accessed: 8 August 2008]

Reed, C., 2007, “KML 2.1 Reference – An OGC Best Practice”, Open Geospatial Consortium.
Available online: http://portal.opengeospatial.org/files/?artifact_id=21469 [Last accessed: 1.
oktober 2007]

Reed, C. (editor), 2006, “An Introduction to GeoRSS: A Standards Based Approach for Geo-enabling
RSS feeds”, OGC White Paper, Open Geospatial Consortium. Available online:
http://www.opengeospatial.org/pt/06-050r3 [Last accessed: 8 August 2008]

Robinson, A. H., Morrison, J.L., Muehrcke, P.C., Kimerling, A. J., Guptill, S. C., 1995, “Elements of
Cartography”, Sixth Edition, John Wiley & Sons

Sandvik, B., 2008, “Supporting Document: Thematic Mapping Engine”, MSc GIS Dissertation,
University of Edinburgh

Sae-Tang, A, Ertz, O, 2007, “Towards Web Services Dedicated to Thematic Mapping”, OSGeo
Journal, Vol. 3. Available online: http://www.osgeo.org/files/journal/v3/en-
us/final_pdfs/ertz.pdf [Last accessed: 8 August 2008]

Sheperd, I.D.H, 2008, “Travails in the Third Dimension: A critical Evaluation of Three-dimensional
Geographical Visualization”. Book chapter in Dodge, M., McDerby, M. and Turner, M., 2008,
“Geographic Visualization: Concepts, Tools and Applications”, Wiley

Slocum, T. A., McMaster, R. B., Kessler, F. C., Howard, H. H., 2005, “Thematic Cartography and
Geographic Visualization”, Second Edition, Person Education Inc.

Statistical Commission, 2007, “Collection and dissemination of statistics by the United Nations
Statistics Division”, Report of the Secretary-General to the Statistical Commission, thirty-ninth
session, 26-29 February 2008. Available online: http://unstats.un.org/unsd/statcom/doc08/2008-
12-CollectionDissemination-E.pdf [Last accessed: 25 June 2008]

Sykora, P., Schnabel, O., Iosifescu-Enescu, I., Hurni, L., 2007, “Extended Cartographic Interfaces for
Open Distributed Processing”, Cartographica, volume 42, issue 3, pp.209-218, University of
Toronto Press.

Torkington, N., 2006, Where 2.0 Standards. O’Reilly Radar. Available online:
http://radar.oreilly.com/archives/2006/05/where_20_standards_1.html [Last Accessed: October,
2007]

UNdata, 2008, “Terms and Conditions of Use”, United Nations. Available online:
http://data.un.org/wiki/disclaimer.ashx [Last accessed: 8 August 2008]

Jackson, D., 2003, “Scalable Vector Graphics (SVG) 1.1 Specification”, W3C Recommendation 14
January 2003, The World Wide Web Consortium (W3C). Available online:
http://www.w3.org/TR/SVG11 [Last accessed: 8 August 2008]

Ward, M., 2004, “Quantifying the World: UN Ideas and Statistics”, United Nations Intellectual History
Project, Indiana University Press

19
Research Paper Bjørn Sandvik

Weiss, P., 2002, “Borders in Cyberspace: Information Policy and the Global Information
Infrastructure”, US Department of Commerce. Available online:
http://www.weather.gov/sp/Borders_report.pdf [Last accessed: 8 August 2008]

Wilson, T., 2008, “OGC KML 2.2.0”, Document #07-147r2, Open Geospatial Consortium. Available
online: http://www.opengeospatial.org/standards/kml/ [Last accessed: 5 May 2008]

20