Content uploaded by Werner Kuhn
Author content
All content in this area was uploaded by Werner Kuhn on Nov 17, 2014
Content may be subject to copyright.
Core Concepts of Spatial Information:
AFirstSelection
Werner K u h n
Institute for Geoinformatics (ifgi)
University of Muenster (Germany)
kuhn@uni-muenster.de
Abstract
The work reported here explores the idea of identifying a small set of
core concepts of spatial information. These concepts are chosen such
that they are communicable to, and applicable by, scientistswhoarenot
specialists of spatial information. They help pose and answer questions
about spatio-temporal patterns in domains that are not primarily spatial,
such as biology, economics, or linguistics. This paper proposes a first
selection of such concepts, with the purpose of initiating a discussion
of their choice and characterization, rather than presenting a definitive
catalog or novel insights on the concepts.
1Introduction
Research and development concerned with spatial information have been try-
ing for at least two decades to integrate spatial informationintomainstream
information technology as well as science and society at large. For example,
the vision of OGC (the Open Geospatial Consortium) went from,originally,
“The complete integration of geospatial data and geoprocessing resources into
mainstream computing” to today’s “Realization of the full societal, economic
and scientific benefits of integrating electronic location resources into commer-
cial and institutional processes worldwide.” The expectation behind the many
large-scale efforts to create Spatial Data Infrastructures,nationallyandinterna-
tionally, is that more of the purported 80% of decisions in society with a spatial
component will eventually become informed by spatial information, thereby im-
proving business, governance, and science.
Progress toward this goal has been modest and often happened outside or
even despite the efforts of spatial information experts, rather than through them.
Today ’ s m o s t p o p ular sources o f spa t i a l i n formati o n a r e i m a gery and map data
from Google, Microsoft and other companies, who acquired anddevelopedtheir
technology largely outside the communities of spatial information experts. The
problem has been recognized (and in part dealt with) in terms of a need for
simpler models and standards than those typically produced by experts. As
aresult,simplystructuredspatialdatacannowbeaccessed,andelementary
analyses performed on them, using non-specialist data formats and services built
around them.
Yet, f o r s patial in f o r m a t ion to b ecome a cross-cut t i n g e nabler of knowledge
and analysis, such bottom-up technological solutions alonearenotsufficient.
Their confusing letter soup of acronyms and their growing plethora of “standards
du jour” do not encourage a broader understanding and communication. An
effort at the conceptual level is needed, in order to present a coherent and
intelligible view of spatial information to those who may notwanttodiveinto
the intricacies of standards and data structures. Spatial information is too
valuable for so ciety to le ave it up to the sp eci alist s; bu t as specialists, we can
do a better job explaining it and demonstrating its benefits.
Spatial information answers questions about themes in space and time.All
its varieties result from treating these three components asfixed,controlled,
or measured (author?) [15]. For example, information about objects is pro-
duced by fixing time, controlling theme (i.e., choosing the objects of interest),
and measuring space. A unified treatment of time and space turns Sinton’s
structure into the geo-atom <x, Z, z(x)>, which links a point xinspace-time
to a property-value pair <Z, z(x)>, where z(x) is the value of property Z at x
[10]. The geo-atom answers the questions what is there? (looking for z, given
x) and where is this? (looking for x, given z), two questions whose dualism is
characteristic for spatial information.
Spatial data as such are not spatial information, but generate it once humans
interpret them. For example, 1-5-3 Yaesu is spatial data and can be interpreted
in the right context as the address of a post office in a ward of Tokyo; the same
goes for 6p21.3, the locus of a gene in a chromosome. Concepts are the mental
mechanisms needed to interpret data. For example, the concept of location is
needed to interpret addresses or gene loci, and the concept ofvalueisneeded
to interpret copyright regulations for spatial data.
The overall goal of this work is to explain spatial information and its poten-
tial for science and society through a set of core concepts. These are concepts
of spatial information, defined here as concepts used to answer questions about
themes in space and time as well as being represented by data or services. They
include spatial concepts,whichservetoreasonaboutspace1and information
concepts,inthesenseofconceptsabout spatial information, which may be spa-
tial or not. An example of the former is location, referring to space and being
represented digitally. An example of the latter is value,whichreferstospatial
information, but is not spatial. An example of both is resolution,whichisa
spatial measure, but also describes spatial information. The co-existence of such
content and meta concepts, and the consequent need to understand both, are
characteristic for information sciences in general.
1Acomprehensiveoverview withanemphasisonlearningishttp://www.teachspatial.org.
2Spatialinformationinscienceandsociety
Why bother with spatial information at all? First of all, because some of
the biggest societal and scientific challenges require a better understanding of,
and better decisions about, the location and interaction of things in space and
time. Consider climate change, biodiversity, financial systems, poverty, security,
health, energy or water supply – spatial information is essential in addressing
each of these global as well as many regional and local challenges. In fact, key
notions in today’s scientific and social debates on these challenges are essen-
tially spatio-temporal - consider risk, sustainability, vulnerability, or resilience.
Secondly, studying spaces at smaller scales (atoms and subatomic particles,
molecules, crystals, biological cells) as well as larger ones (planets and galaxies)
remains among the most fascinating - as well as most costly - scientific ad-
ventures. Thirdly, solutions to non-spatial problems often use spatial analyses
in real or metaphorical spaces, the latter ranging from the data cubes used in
data mining to the spatializations used for information retrieval or mnemonic
techniques.
Dealing with these and other challenges requires approachestranscending
those of single disciplines, even if these disciplines have themselves broad scopes,
as geography or computer science do. Transdisciplinary research addresses chal-
lenges that span multiple disciplines and have direct socialrelevance. Itsgoalis
to make progress in solving these problems, not just to gain knowledge. Its
approaches often benefit from exploiting spatial information, either because
the problems are inherently spatial or because space and timeactasunifiers
and organizers for all phenomena and our knowledge about them. For exam-
ple, combining satellite imagery of the Amazon rain forest with ground sensor
measurements and socio-economic models reveals deforestation patterns, whose
presentation to farmers, decision makers, and the general public helps reduce
the depletion of this planet’s lungs [1]. Recent developments in the social sci-
ences and humanities, referred to as a spatial turn [17], together with numerous
technological advances (navigation systems, mobile computing, high-resolution
satellite imagery, virtual globes, sensors, and crowd-sourced information) fur-
ther amplify and exemplify the opportunities for spatial information in science
and society.
To be able to b r i ng spati a l i nformat i o n to trans d i s ciplin a r ywork,scientists
of any disciplines need to be supported in understanding and exploiting spa-
tiality in their theories and models. This requires an explanation ofspatial
information in a theoretically sound and technically informed way, maximizing
the scope of applications and minimizing technological jargon. The spaces to
be considered are those where transdisciplinary challengesarise,i.e.,primarily
those of human experience in one to three dimensions plus time. They include
geographic spaces (such as a neighborhood in a city or a river catchment), indoor
spaces (such as a room or a hallway), body spaces (such as a human body or
organ), tabletop spaces (such as a desktop or workbench), andimages. Smaller
and larger spaces (such as cells, atoms or galaxies) as well ashigher-dimensional
ones (such as those used in statistics or data mining) are typically understood
through mappings to these experiential spaces. Among them, geographic spaces
are the ones with which we have the richest set of experiences,givinggeographic
information science a privileged status in dealing with spatial information.
3Conceptsofspatialinformation
Surprisingly, a comprehensive treatment of concepts of spatial information with
atransdisciplinaryscopedoesnotyetexist. Allexplanations of spatial infor-
mation have a technological bias toward Geographic Information Systems (GIS)
or a disciplinary one toward geography, surveying, or computer science. Where
they discuss core concepts at all, these are often limited to spatial concepts or
even geometric ones. While there is no shortage of calls to improve this situation
(for an early GIS-oriented example with a broad view, see [2]), comprehensive
results are missing or only starting to emerge1.
This lack of a conceptual consensus on spatial information across disciplines,
spaces, and technologies may be both a reason for, and a resultof,thefactthat
the “science behind the systems” [7] remains largely there - behind the systems.
Two d e cades aft e r i t s t arted to c a l l i tself a scie n c e , a nd despite significant ac-
complishments [9], efforts to present geographic information science to outsiders
as an intellectual (rather than just technological) venturelackacoherentcon-
ceptual basis2.
As a consequence, biologists, economists, or linguists interested in testing a
hypothesis using spatial information have to dig into GIS text books and stan-
dards documents, even just to understand what spatial information and reason-
ing could possibly do for them. Those interested in smaller orlargerspacesare
even worse off, since no textbooks or standards address these across disciplines.
Geographic information science, thus, has to ask itself where economics would be
today if its core concepts were limited to either micro- or macro-economics and
hidden in specialized textbooks and manuals for spreadsheets and accounting
software.
Taki n g u p this ch a l l enge, th e work pr e s e nted h e r e is neith e r about technolo-
gies, nor about particular disciplines or domains. It attempts to cut across their
boundaries, targeting a bigger role for spatial informationinscienceandsociety.
It strives for an explanation of what is special about spatial for those who are
not specialists of spatial information.
4Afirstselectionofconcepts
The discussion of the proposed nine core concepts of spatial information in this
section begins with location and the dual pair of field and object information.
It continues with the spatial concepts of network and process, and ends with
the information concepts of resolution, accuracy, semantics, and value. Each
2Some examples (with a variety of goals) are http://spatial.ucsb.edu/,
http://us pati al.umn. edu/, http: //spatial.uni- muenster.de, http://lodum.de/
concept gets only briefly characterized, in order to initiateandguidediscussions.
Detailed explanations exceed the available space and will beprovidedinarevised
and extended article [13].
4.1 Location
The starting point for a journey through concepts of spatial information has
to be location: spatial information is always linked to location in some way -
but what exactly is location and how does it play this central role? Location
information answers where questions: where are you? where is the appendix?
where are this morning’s traffic jams? Perhaps counter-intuitively, location is a
relation, not a property. This is so, because nothing has an intrinsic location,
even if it remains always in the same place. The house you live in can be
located, for example, by a place name, an address, directions, or various types
of coordinates. All of these location descriptions express relations between the
figure to be located (your house) and a chosen ground (a named region, a street
network, coordinate axes). How one locates things, i.e., what ground and what
relation one chooses, depends on the context in which the location information is
produced and used. Spatial reference systems, for example the World Geodetic
System 1984 (WGS84), standardize location relations and turn them into easier
to handle attributes within such a system. Yet, when data use multiple reference
systems (for example, latitude and longitude as well as projected coordinates),
locations need to be understood as relations and interpretedwithrespectto
their ground (for example, the Greenwich meridian or a projected meridian).
Relating different phenomena through location is fundamental to spatial
analysis. An early and famous application is Dr. Snow’s 1854 finding that
Cholera is spread by drinking water, after he had observed that many Cholera
deaths had taken place around a certain water pump. The great power of such
locational analyses results from the fact that nearby thingsaremorerelated
than distant things, which has been dubbed Tobler’s First Law, based on its
first explicit statement in [16].
4.2 Field
Fields describe phenomena that have a value everywhere in a space of interest,
such as temperature. Generalizing the field notion from physics, field-based
spatial information can also represent values that are statistically constructed,
such as probabilities or population densities. Field information answers the
question what is there?, where “there” can be anywhere in the space of interest.
Fields are one of two fundamental ways of structuring spatialinformation,the
other being objects 4.3. Both fix time, with fields resulting from controlled
space and measured theme, and objects resulting from controlled theme and
measured space. Time can also be controlled, rather than fixed. Controlling it
together with space leads to space-time fields; controlling it together with theme
produces ob ject animations. Fields have been shown to be morefundamental
than objects, capable of integrating field and ob ject views intheformofGeneral
Field models [14].
Since it is not possible to represent the infinitely many values of a field, they
need to be discretized for explicit digital storage. There are two ways to achieve
this, either through a finite set of samples with interpolation between them or
through a finite number of cells with homogeneous values, jointly covering the
space of interest. The cells can all have the same shape (forming a regular grid
of square, triangular, hexagonal, or cubic cells) as in the so-called raster model
for spatial data, which is best known from digital images. Or they can have
irregular shapes, adaptable to the variation of the field, as in the finite element
models used in engineering or the triangulated irregular networks (TIN) used
to represent terrains.
An important kind of fields captures values on two-dimensional surfaces such
as those of the earth or of the human body. These fields are typically organized
into thematic layers.Theideaofalayerisrootedintraditionalpaper-orfilm-
based representations of spatial information, such as maps,andtheproduction
of models from stacked transparent layers of data about a theme. The main
computational use of layers is to overlay them in order to relate information
about multiple themes or from multiple sources.
4.3 Object
After fields, ob jects provide the second fundamental way of structuring spa-
tial information. They describe individual things with spatial, temporal, and
thematic properties and relations. Object information answers questions about
properties and relations of objects, such as where is this?, how big is it?, what
are its parts?,which are its neighbors?, how many are there?.
For many a pplicati o n s , o n e i s inte r e s t e d i n t h ings that a r e features in the way
that a nose is a feature of a face, i.e. parts of a surface. Features are important
siblings of objects, but can be understood as a special case. The simplest way
to carve out features from a surface is to name regions on it. Geographic places
are the prototypical examples, carved out of the earth’s surface by naming
regions; but the same idea applies, for example, to models of airplane wings,
sails, or teeth. Object and feature models can co-exist, and the general tendency
today is to complement two-dimensional feature models with three-dimensional
object models and provide more or less seamless transitions between them. For
example, your house may be represented as a feature of the earth’s surface in
adigitalmap,asafeatureofstreetviewimages,andasathree-dimensional
object. The resulting blended feature-object notion pervades geography, but
also exists in biology and medicine (features of cells, organs, or bodies), and
generally in imaging (features extracted from images of anything).
Many questions about objects and features can be answered based on simple
representations as points with thematic attributes. For example, doing a blood
count or determining the density of hospitals in an area require only point rep-
resentations. On the other hand, some questions do require explicit boundaries
enclosing or separating objects. For example, determining the neighbors of a
land parcel, the extent of a geological formation, or the health of blood cells may
require boundary data. The frequent occurrence of boundaries in object infor-
mation, however, has mainly historical reasons, since analog representations like
images or maps were digitized by drawing or following lines onthem.
The so-called vector models for spatial data capture objects with boundaries
at various levels of sophistication. Like surface fields in raster data, collections
of features in vector data can be organized into thematic layers.Processing
vector data exploits the geometry of boundaries to compute sizes, shapes, buffers
around the objects, and overlays. Yet, many ob jects, particularly natural ones,
do not have crisp boundaries [3]. Examples are geographic regions or body
parts such as the head. It can be harmful to impose boundaries on such objects
only for the sake of storing them in vector models. Differencesbetweenspatial
information from multiple sources are indeed often caused bysuchmoreorless
arbitrary delimitations. For example, boundaries of climate zones are vague by
nature, and the variation in boundaries between different definitions matters
much less than the overall extent and location of the zones. Thus, whether
modeling ob jects with explicit boundaries is necessary or even desirable has to
be carefully assessed for each application. It is certainly not something that the
concept of an object implies.
4.4 Network
Connectivity is central to space and spatial information. The concept of a
network captures binary connections among arbitrary numbers of objects, which
are called nodes or vertices of the network. The nodes can be connected by any
relation of interest. Network information answers questions about connectivity,
such as are nodes m and n connected?,what is the shortest path from m to
n?,how central is m in the network?,where are the sources and sinks in the
network?,how fast will something spread through the network?,andmanyothers.
The two main kinds of networks encountered in spatial information are trans-
portation and social networks. Transportation networks (inthewidestsense)
model systems of paths along which matter or energy is transported, such as
roads, utilities, communication lines, synapses, blood vessels, or electric circuits.
Social networks capture relationships between social agents, such as friendships,
business relations, or treaties.
All networks can be spatially embedded, which means that their nodes are
located. This is often the case for transportation networks and increasingly for
social networks. If the embedding space is a surface, networks can be organized
into thematic layers, like the surface fields and feature collections encountered
above.
Network applications benefit from the well studied representations of net-
works as graphs and the correspondingly vast choice of algorithms. Partly due to
this sound theoretical and computational basis, networks are the spatial concept
that is most broadly recognized and applied across disciplines.
4.5 Process
Processes are of central interest to science and society - consider processes in
the environment, in a human body and its cells, and in machinesormolecules.
Processes that manifest themselves in field, object, or network information are
considered spatial. Information about spatial processes primarily answers ques-
tions about motion,change,andcausality.
Controlling time and measuring space generates informationaboutmotion;
controlling time and measuring theme informs about change.Timeistypically
controlled through time stamps in spatial information. Temporal reasoning on
time stamps (and on time intervals formed from them) is the basis for under-
standing motion and change. Migration or embolism are examples of motion.
Growth, such as that of vegetation or social networks, exemplifies the change
of objects or networks. Diffusion, for example in the form of climate change,
collapsing house prices, or spreading innovation, is an example of a change in
fields, objects or networks.
The most complex relation between spatial processes is that of causality.Dr.
Snow’s tracing of cholera to drinking water is a case of determining that one
process (drinking some water from a contaminated pump) is thecauseofanother
(contracting cholera), based on the patients being located near the pump.
Real-time spatio-temporal data from sensors and spatio-temporal simula-
tions are the two key sources of process information. In ordertomakessense
of these dynamic data and models, science needs better theories of change [4].
One of the main benefits to be expected from a list of core concepts of spatial
information is indeed to establish the conceptual foundations for such theories.
If the theories can be formulated in terms of the proposed coreconcepts,their
choice will be corroborated; if not, other concepts will havetojoinorreplace
them. For the current proposal, this means that all spatial change needs to
be explained in terms of operations on locations, fields, objects, networks, and
processes themselves.
4.6 Resolution
Resolution is the first and most spatial concept of information on this list. It
characterizes the size of the units ab out which information is reported and ap-
plies to all three components of space, time, and theme. For example, satellite
images have the spatial resolution of the ground area corresponding to a pixel,
the temporal resolution of the frequency at which they are taken, and the the-
matic resolution of the spectral bands pictured. Vote countshavethespatial
resolution of voting districts, the temporal resolution of voting cycles, and the
thematic resolution of parties or candidates. Resolution information answers
questions about how precise spatial information is, for example, when taking
decisions based on the information.
Resolution characterizes information about all concepts introduced so far:
location is recorded at certain granularities, fields are recorded at certain sam-
ple spacings or cell sizes, and the choice of the types of objects (say, buildings
vs. cities) and nodes (say, transistors vs. people) determines the spatial reso-
lution of object and network information. The choice of the spatial, temporal,
and thematic resolution at which spatial information gets recorded is primar-
ily determined by the processes studied, because these involve phenomena of
certain sizes, frequencies, and levels of detail. For example, migration, social
networking, and the diffusion of technological innovations all involve people over
months; embolism involves blood clots and vessels over hours; cancer involves
cells and organs over years; climate change involves large air and water masses
over decades; changing house prices involve land parcels andpeopleoverdays
or weeks.
Many processes need to be studied at multiple resolutions (for example, ero-
sion) or they connect to processes at other resolutions. For example, one can
think of all processes as involving some sort of motion at someresolution. All
five core spatial concepts on our list can be represented at multiple resolutions:
location descriptions are often hierarchical (for example,addresses);fieldsareof-
ten represented by nested rasters (called pyramids in the case of images); object
hierarchies are expressed as part-whole relations between objects (for example,
administrative subdivisions of countries); hierarchical network representations
allow for more efficient reasoning (for example, in navigation), process models
(for example, in medicine) are connected across levels of detail.
4.7 Accuracy
Accuracy, like precision, is a key property of information, capturing how in-
formation relates to the world. Information about accuracy answers questions
about the correctness of spatial information. The location of a building, given
in the form of an address, coordinates, or driving instructions, can in each case
be more or less accurate. The spatial, temporal, and thematiccomponentsof
spatial information are all subject to (in)accuracy.
Assessing the accuracy of information requires two assumptions: that there
is in principle a well-defined correct value and that repeatedmeasurementor
calculation distributes in regularly around it. The first assumption requires an
unambiguous specification of the reported phenomenon and of the procedure to
assign values. For example, if temperatures are reported fordifferentplaces,
one may need to specify the level above ground to which they refer. The second
assumption requires an understanding of measurement as a random process.
Choosing a particular form of distribution (called a probability density function)
allows for estimating the probability that a measured or computed value falls
within a given interval around the correct value. Mean errorsandanyother
accuracy data are based on these two assumptions.
Accuracy connects to resolution through the established practice of reporting
all data at a resolution corresponding to the level of expected accuracy. If
information is collected at multiple levels of resolution, one level can sometimes
be considered as accurate when assessing the others. For example, positions
determined from high-precision measurements serve as “fix points” for lower
precision measurements, and ob jects get extracted from remotely sensed images
by determining “ground truth” for parts of an image.
4.8 Semantics
Understanding the semantics of spatial information is crucial to its adequate
use. When it comes to analyzing spatial information, determining whether the
same things are called the same (and different things differently) is essential
to producing meaningful results and making sense of them. Thechallengeis
to capture what the producer means with some data or services and to guide
the user on how to interpret them. For example, when navigation systems use
road data, they make assumptions on what the data producer meant by “road
width” (paved or drivable?, number of lanes or meters or feet?). When using a
spatial information service, operational terms such as distance also have to be
interpreted adequately.
Semantic information answers the question how to interpret the terms used
in spatial information. It concerns the spatial, temporal, and thematic com-
ponents. While the semantics of spatial and temporal data have long been
standardized through spatial and temporal reference systems, the semantics of
thematic data and operations remain hard to capture and communicate. What
is meant with data about land use or body tissue, for example, depends on a
complex interaction between defining the intended use of someterms(say,forest
or muscle) and delineating the spatio-temporal extents of their application to
land or tissue.
Data and services do not have a meaning by themselves, but are used to
mean something by somebody in some context. Therefore, it is impossible to
fix the meaning of terms in information. However, one can make at least some of
the conditions for using and interpreting a term explicit. This is what ontologies
do: they state constraints on the use of terms. But language use is flexible and
does not always follow rules, even for technical terms. An empirical account of
how some terms are actually used can therefore provide additional insights on
intended meaning or actual interpretation. This is what folksonomies deliver:
they list and group terms with which information resources have been tagged.
Semantic information consists of necessarily incomplete collections of con-
straints from ontologies and folksonomies on the use and interpretation of terms.
The constraints can use binary logic (for example, stating that a term refers to
asubsetofthethingsthatanothertermrefersto)orfuzzylogic (where such a
statement is neither true nor false, but possibly true). The latter is an attempt
to account for the inherent vagueness of many terms.
Yet, a l l c o n straint s d e p e n d o n co ntext.Termsareusedbysomebodyto
mean something in a given context. Ontologies and folksonomies capture some
aspects of context, but spatial information is often used in other contexts than
the ones it was produced in. For example, road width data produced by traffic
engineers may be quite different from those needed for navigation. In order to
map between different contexts, ontologies need to be grounded.Thismeans
that their constraints need to refer to something outside their context, to which
the constraints of other contexts can then refer as well. Spatial information
has successfully relied on grounding for centuries, throughspatialreferencesys-
tems. These systems refer coordinates to something outside their conceptual
framework, such as physical monuments or stars. Generalizing this idea from
location information to any terms used in spatial information leads to the idea
of semantic reference systems [12]. These systems, once established in practice,
are expected to support translations of terms used in spatialdataandservices
from one context to another. Grounding is the basis for analytical translations
of terms from one context to another. Since all constraints onmeaningarenon-
deterministic, stochastic approaches to translation are a valid alternative to
explicit grounding. For example, translations of terms across natural languages
are now routinely and successfully achieved through stochastic methods.
4.9 Value
The final core concept proposed is that of value. Information about the value of
spatial information answers questions about the many roles spatial information
plays in society. The main aspect of value is economic,butthevaluationof
spatial information as a good in society goes far beyond monetary considera-
tions. It includes its relation to other important social goods, such as privacy,
infrastructure maintenance, or cultural heritage.
Setting policies on public access to spatial information, for example, is a
pressing societal need requiring a better understanding of the many valuations
involved. It is further complicated by the fact that information about indoor
and geographic spaces can now be and is being collected and shared by almost
everybody. This phenomenon of crowd-sourced or VolunteeredGeographicIn-
formation (VGI, [8]) is profoundly altering the value of spatial information, from
economic as well as institutional, ethical, and legal perspectives. For example,
akeynewchallengecreatedbyVGIistounderstandandmodeltrust in spatial
information.
Given these wide ranging aspects of spatial information value, no coherent
theoretical framework for it can be expected any time soon. Partial theories
of value, for instance about the economic value of spatial information, are still
sketchy and difficult to apply, because they involve parameters that are hard to
control or measure. The cost of spatial information is no goodguidetoitsvalue
either, because it often reflects the high expenses for collecting the information,
rather than the value of the result.
Value o f i n formatio n t e n d s t o a c c r u e holistic a l l y a n d u n predictably, by new
questions that can be asked and answered, new services that are provided.
Partly for this reason, spatial information holdings have b ecome significant as-
sets, not only for scientists and governments, but also for enterprises in all
sectors. Such assets need to be evaluated, for example in enterprise valuation,
reinforcing the need for theories of spatial information value.
Even at the level of personal information management, the value of access-
ing and analyzing information through its spatial and temporal properties has
barely been understood and tapped into yet [11]. For example,searchinginfor-
mation by where or when it was collected or stored is highly effective, but still
only weakly supported by the web, personal computers, and smart phones.
4.10 Also-ran
It may be useful to consider some arguments against core status for some other
concepts. Obviously, these may have to be reconsidered, so that this list of
also-rans is part of the material for discussion.
My earlier lists of concept candidates contained nearness, spatial relations,
feature, map, layer, motion, path, uncertainty, and scale. Typical reasons to
exclude them from the list were that they were too broad or too narrow. In
particular:
•nearness got generalized to spatial relations,buttheseservetospecify
location and are covered there;
•features are now treated together with objects;
•maps are visualizations of mostly geographic information that exists in
other forms;
•layers structure the representations of several concepts (fields, objects,
networks) and are dealt with there;
•motion is only one process in space, although the most important one;
•paths are covered as parts of networks;
•uncertainty covers several concepts, of which resolution, accuracy, and
semantics are covered;
•scale is also a catch-all for several concepts, of which resolutionison
the list, extent (of a study area) is rather trivial, and support is more
specialized (belonging to measurement ontology).
5Conclusions
Achieving a stronger role for spatial information in scienceandsocietyrequires
explaining its uses and benefits at a higher level than that of technologies and
acronyms. The small set of concepts of spatial information proposed in this
paper indicates a possible basis for such explanations. While it may miss or
misrepresent some concepts, it provides a starting point to reach a conceptual
view of our field that is accessible and intelligible to outsiders. The main goal
at the moment is, therefore, to receive critical feedback andsuggestionsofwhat
to add, drop, or change.
The concepts chosen and revised based on the expected feedback will then
be described in more detail over the coming year3.Thesedescriptionswillask
and answer four questions about each concept:
3To par tic i p at e in t h e d i scu s si o n , pl e ase vi s i t ht tp: / /if g i .uni-
muenster.de/services/oj s/index.php/ccsi/i ndex
1. what is the concept, i.e., what phenomena does it capture?
2. where does information about the concept come from, i.e., what are typical
sources of information about it?
3. how is the concept represented,i.e.,whatdatastructuresandalgorithms
implement it?
4. how is information about the concept used,i.e.,whatreasoningandanal-
yses does the concept support?
The expected result is a catalogue of core concepts that are meaningful and
useful across disciplines - a vocabulary to talk about spatial information to non-
specialists. Such vocabularies, when formalized, are referred to as ontologies.
While formalization is not a primary goal here, treating the concepts as nodes
in an ontology and relating them to an upper level ontology or embedding them
in ontology patterns will certainly help to clarify them further. Starting this
work as an ontology design exercise, however, would most likely not lead to a
useful set of concepts, because their relation to actual dataandcomputations
would be to o weak. A subsequent ontological analysis will produce an ontology
of spatial information that allows for interfacing with other domains, while
relating explicitly to information technology. As such, it will complement and
benefit from existing ontologies of spatial information [6, 5].
Acknowledgments
Countless discussions over the years with many colleagues and friends have
encouraged and influenced these thoughts. The members of http://musil.uni-
muenster.de and the students of my Introduction to Geographic Information
Science have been very helpful critics and supporters of this work. Some anony-
mous reviewers of GeoInfo2011 provided very useful comments.
References
[1] Ana Paula Dutra Aguiar, Gilberto Câmara, and Maria IsabelSobralEs-
cada. Spatial statistical analysis of land-use determinants in the Brazilian
Amazonia: Exploring intra-regional heterogeneity. Ecological Modelling,
209(2-4):169–188, December 2007.
[2] Peter A Burrough and Andrew U Frank. Concepts and paradigms in spatial
information: Are current geographic information systems truly generic?
International Journal of Geographical Information Science, 9(2):101–116,
1994.
[3] Peter A Burrough and Andrew U Frank, editors. Geographic Objects with
Indeterminate Boundaries.Taylor&Francis,December1996.
[4] Gilberto Câmara, Lúbia Vinhas, Clodoveu Davis, Fred Fonseca, and Tiago
Carneiro. Geographical Information Engineering in the 21stCentury. In
Gerhard Navratil, editor, Research trends in geographic information sci-
ence, pages 203–218. Springer Verlag, 2009.
[5] Helen Couclelis. Ontologies of geographic information.International
Journal of Geographical Information Science,24(12):1785–1809,November
2010.
[6] Andrew U Frank. Ontology for spatio-temporal databases.Spatiotemporal
Databases. In Timos Sellis, editor, The Chorochronos Approach,pages
9–77. 2003.
[7] Michael F Goodchild. Geographical information science.International
Journal of Geographical Information Science, 6(1):31–45, 1992.
[8] Michael F Goodchild. Citizens as sensors: the world of volunteered geog-
raphy. GeoJournal,69(4):211–221,November2007.
[9] Michael F Goodchild. Twenty years of progress: GISciencein2010.Journal
of Spatial Information Science,1(1):3–20,July2010.
[10] Michael F Goodchild, May Yuan, and Thomas Cova. Towards ageneral
theory of geographic representation in GIS. International Journal of Geo-
graphical Information Science, 21(3):239–260, 2007.
[11] Krzysztof Janowicz. The Role of Space and Time For Knowledge Organi-
zation on the Semantic Web. Semantic Web - Interoperability, Usability,
Applicability, 1(1-2):25–32, 2010.
[12] Werner Kuhn. Semantic Reference Systems. International Journal of Ge-
ographic Information Science (Guest Editorial), 17(5):405–409, June 2003.
[13] Werner Kuhn. The sciences before the systems: towards a transdisciplinary
role for spatial information. International Journal of Geographical Infor-
mation Science,(underreview):1–15,2012.
[14] Y. Liu, Michael F Goodchild, Q. Guo, Y. Tian, and L. Wu. Towards a Gen-
eral Field model and its order in GIS. International Journal of Geographical
Information Science, 22(6):623–643, 2008.
[15] David Sinton. The inherent structure of information as aconstrainttoanal-
ysis: Mapped thematic data as a case study. Harvard papers on geographic
information systems, 6:1–17, 1978.
[16] Waldo Tobler. A computer movie simulating urban growth in the Detroit
region. Economic Geography, 46:234–240, 1970.
[17] Barney Warf and Santa Arias, editors. The spatial turn: Interdisciplinary
perspectives,volume26.Taylor&Francis,November2009.
