=Paper=
{{Paper
|id=Vol-1557/paper6
|storemode=property
|title=Thinking about Spatial Computing
|pdfUrl=https://ceur-ws.org/Vol-1557/paper6.pdf
|volume=Vol-1557
|authors=Andrea Ballatore,Werner Kuhn
|dblpUrl=https://dblp.org/rec/conf/cosit/BallatoreK15
}}
==Thinking about Spatial Computing==
https://ceur-ws.org/Vol-1557/paper6.pdf
Thinking about Spatial Computing
Andrea Ballatore1 and Werner Kuhn1
Center for Spatial Studies
University of California, Santa Barbara
kuhn|aballatore@spatial.ucsb.edu
Workshop on Teaching Spatial Thinking from Interdisciplinary Perspectives
(SPATIALTHINKING2015), COSIT 2015
Abstract. We propose to include the perspective of spatial computing
in interdisciplinary courses on spatial thinking. Specifically, we recom-
mend developing and applying a set of spatial lenses through which learn-
ers of Geographic Information Systems (GIS) get to see geographic space
and choose spatial computations. These lenses are based on the core
concepts of spatial information proposed by the authors. While there is
intentionally nothing new about the concepts per se, their explicit use as
lenses through which to see geographic information and select GIS oper-
ations is innovative. Thus, we propose a lightning talk on core concepts
of spatial information as a form of spatial thinking to support learning
GIS.
Keywords: spatial computing, spatial lenses, core concepts of spatial
information, GIS learning
1 Our Perspective: Spatial Computing
Spatial thinking is normally discussed in contexts of wayfinding or moving and
arranging objects. When developing educational materials for spatial thinking,
it is worth considering the spatial thinking that goes into spatial computing.
In particular, conceptualizing geographic phenomena, for example as fields or
networks, is a spatial thinking skill that gets taught less explicitly. There is
broad consensus that choosing how to conceptualize geographic phenomena for
analysis is a key spatial thinking skill [1]. Yet, GIS learners are often taught
about data formats, file types, databases, and standards before learning to think
about GIS contents. In other words, using a GIS is taught more through software
thinking than through spatial thinking about contents.
As a consequence, GIS users are taught how to answer questions about geo-
graphic space by memorizing computational techniques to answer them, rather
than by being presented with a way to think about questions that make sense
on certain contents. The questions typically remain implicit and get lost in the
necessary translation to the answering methods captured by GIS commands. For
example, the question which farms in some agricultural region are at risk from a
25
2 Ballatore and Kuhn
bird flu outbreak needs to be translated into commands for computing bird den-
sity maps, which are then visually inspected for risk areas. The conceptualization
of the question in terms of bird density surfaces is implicit and alternative choices
(such as farm objects connected through social and transportation networks) are
hard to think of.
2 Our Recommendation: Teaching through Spatial
Lenses
One can view the conceptualization phase of a GIS project as a process of choos-
ing lenses through which to look at geographic space. For example, a GIS user
may need to decide whether to consider some terrain as a surface (with eleva-
tion values for any position, allowing for computations like slope and aspect)
or as a network of peaks, pits, ridges and valleys (allowing for flow analyses).
Each choice of a lens (field or network, in this case) comes with a set of suitable
analysis operations (map algebra and network analysis). GIS user interfaces are
currently designed without such explicit lenses, o↵ering the users a bewildering
collection of analysis operations cutting across di↵erent conceptualizations.
Based on this idea of spatial lenses, we suggest including a computational
perspective in an interdisciplinary spatial thinking course. What spatial thinking
informs the choice of spatial computations and how? How is that kind of thinking
about geographic space best taught? As an answer to the first question, we
suggest the previously published idea of core concepts of spatial information.
Today’s answer to the second questions is, in practice, “through GIS commands”.
Instead, the answer could be “through the lenses of core concepts of spatial
information”. The core concepts have been defined in [2] and further specified
in [3]. Table 1 lists the terms adopted for them. A more extensive table with the
computations proposed for each core concept is omitted here for space reasons
and can be found in [3].
Core%Content%Concepts% Core%Quality%Concepts%
Granularity(
Location( Field( Object( Network( Event(
Accuracy(
Table 1: Overview of the core concepts of spatial information
! Fig. 1. Overview of the core concepts of spatial information
The first results from testing the set of core concepts as a vehicle for compu-
tational spatial thinking in an undergraduate introductory GIS course at UCSB
are encouraging. The idea for the course was that students should learn about the
kinds of questions they can ask a GIS before they study how to produce answers
through often obscure system commands. The course design went hand-in-hand
26
Thinking about Spatial Computing 3
with further research on these core concepts. The learning of the core concepts
was grounded by performing computations around each concept on local data
from UCSB Campus and surroundings.
3 Our Resources: An interactive tool, a book, open
source code, and labs
To support our recommendation, we plan to provide the following resources for
teaching an interdisciplinary spatial thinking course:
1. An interactive cube with lenses on its sides to see geographic information
from six di↵erent perspectives;
2. A set of short introductory texts, as chapter drafts of a book on the core
concepts underlying the lenses;
3. A GitHub repository1 containing formal specifications of the core concepts
and Python scripts to use the lenses for actual GIS queries;
4. A set of GIS labs organized around the core concepts.
These materials can be made available in any form deemed useful by the design-
ers of the envisioned interdisciplinary spatial thinking course.
4 Conclusion
Our workshop contribution will be to make a case for spatial thinking in the
teaching of spatial computing. We will present the idea of spatial lenses, defined
by a set of core concepts of spatial information. The main idea of this approach
is that computations need to be organized conceptually around the concepts
they compute with. This is a step forward from past attempts of GIS designers
and GIScience researchers to organize existing GIS commands in a bottom-up
fashion. Ongoing work concentrates on refining the specifications of core concept
computations and continuing the implementation of Python scripts to link them
to existing GIS commands and spatial computing libraries.
Acknowledgments
This ongoing research is made possible through funding from the University of
California Santa Barbara for its Center for Spatial Studies.
References
[1] Golledge, R. G. (1995). Primitives of spatial knowledge. In: Nyerges, T. L.,
Mark, D. M., Laurini, R., Egenhofer, M. J. (Eds) Cognitive Aspects of
Human-Computer Interaction for Geographic Information Systems, Springer,
pp. 29–44.
1
https://github.com/spatial-ucsb/ConceptsOfSpatialInformation
27
4 Ballatore and Kuhn
[2] Kuhn, W. (2012). Core concepts of spatial information for transdisciplinary
research. International Journal of Geographical Information Science, 26 (12:
Special Issue in honor of Michael Goodchild), 2267–2276.
[3] Kuhn, W., Ballatore, A. (2015). Designing a Language for Spatial Comput-
ing. In: Baçao, F., Santos, M. Y., Painho, M. (Eds) Geographic Information
Science as an Enabler of Smarter Cities and Communities, AGILE Confer-
ence, Lisbon (Portugal) (pp. 309–326, Best Paper Award), Springer.
28
Spatial Thinking and GIS
Diana S. Sinton1,2
1
University Consortium for Geographic Information Science
2
Cornell University
Ithaca, New York 14850
dianasinton@cornell.edu
Abstract. The practice of spatial thinking involves applying multiple spatial
concepts during processes of reasoning, and it is a constant component of our
daily professional and personal lives. Certain concepts such as location, scale,
representation, and distance play particularly important roles in the teaching and
learning of geographic information systems (GIS). As GIS is increasingly used
by people without a geographic foundation of thinking spatially, identifying why
and how to make spatial thinking more explicit can have positive learning
outcomes.
1 Introduction
Thinking spatially integrates spatial concepts with processes of reasoning, often relying
on internal or external representations to enable or facilitate and support the experience
[10]. A single act of spatial thinking also involves visualizing and interpreting location,
position, distance, direction, patterns, relationships, movement, and change through
space and time [13], [15]. Because these processes are such fundamental parts of our
everyday cognition, they are rarely thought of explicitly or deliberately. Yet with
evidence growing for the links between spatial thinking and professional success in the
STEM disciplines, not to mention the daily practices of navigation and information
visualization, the need for increased and focused attention on spatial thinking is
paramount.
1.1 Visual and Spatial Thinking
Significant and substantial overlap exists between visual and spatial thinking, yet
they also differ on cognitive, and experiential levels. Frequently, people describe a
preference for visual thinking when they are describing their preference to see
information versus hear it. Just as frequently, though, the preference is instead for
information to be conveyed via images, charts, graphs, or figures, versus through the
use of words. These are all visually dependent forms of communication, since it is
through one’s sense of vision that the information is perceived. But in this case, to opt
for an image or a chart over words or text derives in part from the benefit that spatial
thinking provides. That is, the meaning from the image or chart is being extracted via
the spatial arrangement or other spatial characteristics of items or data in the image or
29
chart. As a simple example, familial relationships between generations of grandparents,
parents, siblings, cousins, and grandchildren are possible to be described through a
lengthy paragraph of text. Those same relationships can be intuited much more
efficiently and succinctly through the conventional arrangement of a family tree, with
subsequent generations arranged vertically and same generations aligned horizontally.
There is also a directional component to interpreting relative ages.
1.2 Spatial Thinking, Geography, and Maps
The skills used to extract meaning based on the spatial arrangement of individuals in a
family tree are analogous to those applied to the arrangement of objects or phenomena
represented on a map. Geographers have long practiced spatial thinking in terms of a
disciplinary focus on location, distance, arrangements, patterns, and other space-based
phenomena. In practice, geographers are likely to first make note of arrangement or
patterns and then deduce the natural or social processes that would have resulted in
such a pattern being observable. Thus, “Why is it like this, here?” is a quintessential
question for geographers. Historically, this was consistent with direct observations, and
this was one reason why field work has always been an essential component of
geography.
To study and review areas beyond what could be seen via direct observation, maps have
always been the obvious geographical tool. Historically maps would have been
derived from direct experience, but since the use of aerial photography and satellite
imagery became widely available, geographers are more likely to use these as sources
of data. What has changed more recently in the last few decades is not so much the use
of remotely-sensed data but the practice of using geographic information systems (GIS)
to manipulate the data and generate maps. The ability to interact with and visualize
geographic space via digital technologies has dramatically expanded our ability to
conceptualize space in novel ways. This has opened up new avenues for both
knowledge and research, invoking multiple and distinctive notions of scale [7].
2 Location, Scale, and Representation
Through the use of digital technologies such as GIS, spaces become manipulable
(able to be grasped, rotated, scaled through zooming, moved), and one can insert
oneself into the space and recreate the perspective of direct observation (locomotion)
[2]. Depending on the purpose of the interaction and the nature and overall size of the
space being explored, the experience that occurs during manipulation or locomotion
would vary greatly. Appreciating these distinctions has significant implications for the
design and use of GIS, such as how to implement data editing functionality or add
functionality to modify perspectives [8]. How research questions are framed and how
research is designed are affected by these experiential differences. For example, the
study of navigation in real-world spaces may or may not mimic the augmented reality
30
experience of locomotion in virtual spaces [9]. The differences further extend to other
domains of science learning [4], [17], [19].
Concepts of location, representation, and scale are central to the connections
between spatial thinking, GIS, and GIScience. Most people are unlikely to know the
capabilities and technological constraints of how a computer is able to represent the
features and the characteristics of the natural and social world, much less the differences
among spatial data models themselves. Smith and Mark [16] found that people were
more comfortable identifying “geographic features” such as mountains and rivers as
things that “can be portrayed on a map,” but were much less likely to be able to envision
ta cartographic representation of a geographic “object” or “concept.” This lack of
familiarity with the digital representation of geographic phenomena that exist beyond
one’s direct level of observation, regardless of scale, affects how people understand the
nature of GIS data sets [14]. For example, it can be particularly challenging to
appreciate how we represent 3-dimensional phenomena, such as geologic bedrock,
ground water, or air temperature, with 0-, 1-, or 2-dimensional digital data structures.
Making decisions about data models and aggregation affect not only representation
but analysis and interpretation as well. For example, the classic example of John
Snow’s 1854 map of cholera deaths around a certain water pump on Broad Street in
London generated a pattern that suggested the correlation between deaths and that water
source. This is an example of spatial thinking that considers how two different 0-
dimensional “point” patterns may be related to one another. However, if the pattern
of points (the individual deaths) had instead only been available in an aggregated form,
such as at a neighborhood or a census tract as a two-dimensional polygon, the formation
of those polygons could either support or refute a correlation with the Broad Street
pump. For an illustration if this particular example, see Mark Monmonier’s mapped
version [6].
The London cholera example illustrates the risks inherent with undertaking spatial
analysis without a clear understanding of the spatial scale at which the different
variables naturally operate and interact. Related to this is the modifiable areal unit
problem (MAUP), often invoked in GIS-based spatial analysis when one is analyzing
the relationships between data sets aggregated at different scales and across different
administrative units, such ZIP code areas and Census tract areas.
Another area in which scale and representation conflict with regards to spatial
thinking involves spatial reference and coordinate systems, map projections, and the art
and science of cartography. Technically, a GIS can accommodate virtually an unlimited
number of coordinate systems associated with an equally large number of map
projections, assuming the software provides the functionality to change key parameters.
However, by design the system does not expect and cannot easily permit these to be
manipulated or distorted across map scales or map extents. That is, a single map that is
“not drawn to scale,” that simultaneously depicts multiple, geographically-coincident
data sets, or a single but multi-part data set, each at a different scale, cannot be
generated. So, a data set of the fifty United States cannot, in one single frame, have
Alaska appear to be the size of Texas, unless it has it as an inset map with its own
31
different scale than the other contiguous states. Again, this is by design because a GIS
is expected to maintain the georeferencing of its digital data sets with absolute
consistency and reliability.
In contrast, our minds do not maintain our knowledge of geographic space either “to
scale” or in correct georeferenced alignment. This has significant implications for the
ways in which we learn to navigate in new locations [3] or interpret You-Are-Here
maps [5]. However, it can be to our educational benefit to deliberately distort and
manipulate fixed and deterministic spatial reference systems to design alternative
representations of geographic features. The 1856 map titled “Mountains and Rivers”
(Figure 1) designed by G. W. Colton depicts both the world’s longest rivers and its
tallest mountains, arranged side-by-side respectfully to allow comparison between
these geographic features. Such a map could be produced via a GIS only with extensive
manipulation of the software and data structures, to trick the system into ignoring scale
and location.
Fig. 1. Mountains & Rivers, a map by G. W. Colton. Published by J.H. Colton &
Co.(1865). Map courtesy of the David Rumsey Map Collection.
3 Distance and Directionality
The spatial concept of distance may be the most important one connecting spatial
thinking and geographic information science. The role of distance in interpreting and
predicting patterns of natural and social phenomena is so essential that it forms the basis
of the so-called First Law of Geography: that everything is related to everything else,
but near things are more likely to be related than far things. This observation, first made
by geographer and computer scientist Waldo Tobler in 1970 [18], holds for many
phenomena and at many scales, yet because it is not a universal truth and has
32
exceptions, it therefore becomes a factor to be considered and addressed during analysis
before it can be dismissed.
Evaluating it, however, is in itself problematic. How “near” and “far” are interpreted
makes all the difference. GIS operations rely on the distance measurements calculated
within and between data sets. Distances form the basis of how buffers and all other
tools of proximity and adjacency are implemented. Appreciating distance as a spatial
concept seems deceptively obvious, but to correctly apply distance-based tools during
a spatial analysis requires an understanding of the scale at which the data’s patterns and
processes exist and operate. Moreover, “near” and “far” will always vary by context,
application, scale, and setting. Programming a computer to be sensitive to those
variables is not yet easily done.
Another challenge being introduced are the inconsistent and idiosyncratic methods
that are applied to analyze patterns based on distance. Conceptually the role of distance
may be a simple concept to understand, but there are numerous computational
approaches to its analysis and little communication between researchers from different
fields. For instance, a Generalized Spatial Association Rule (GSAR) is being used in
business and logistics [1], but it is identical to existing methods for detecting spatial
autocorrelation that are already well-known and implemented in GIS. This exemplifies
the significant gaps that exist between tool developers and researchers across different
fields and industries, all of whom are grappling with how best to analyze distance.
An understanding of directionality is also relevant, as spatial processes are non-
uniform [11]. Water flows downhill, wind blows things down-wind rather than up-
wind, and people tend to move back-and-forth along roads, trails, and other corridors,
rather than randomly across the landscape. Thus there is prior knowledge about both
space and geography that ought to be considered when one is making choices about
analysis via a GIS. Unfortunately, the science behind geographic information does not
systematically or consistently result in ideal tool or system development.
4 Curricular Instruction around Spatial Thinking and GIS
Given the important connections between spatial thinking, spatial concepts, and GIS,
it makes sense to leverage the knowledge to enhance and improve learning. Knowledge
about location, scale, representation and distance serves as “pre-GIS” background to
make their GIS learning as efficient and effective as possible. This is particularly
important for students who lack formal education in geography, and for whom the
practices of spatial thinking are not familiar.
Curricula like these have been developed in several different educational settings
[13]. Typical introductory GIS classes cover numerous topics that can be explained
with their spatial context explicitly provided. For example, when students are learning
about digital elevation models (DEMs) used to represent topography, have students first
venture outdoors and sketch any nearby vista containing topographical relief, from both
frontal and planar perspectives (as contour lines). As a second indoors step, have
33
students create a physical model of the topography with clay, and once the model is
complete and placed within a clear plastic or glass container, they can view it from its
side and sketch what would be its contour lines (with a marker on the outside of the
container itself). Lastly, have the students place different small pieces of mesh netting
(like an assortment of 5” by 5” pieces of window screens, with different mesh sizes,
available for purchase at hardware stores) over the top of the containers and view their
topographical model through the mesh. This mimics the experience of having elevation
represented in a GIS as pixels of different sizes or resolutions. Together, these steps
contribute important prior knowledge about scale and digital constraints for how natural
features are represented in a GIS, and they can be completed during a class session prior
to beginning work with DEMs themselves. In these ways, a DEM will not be such an
unfamiliar abstraction of reality and students gain confidence in the analyses they
conduct with them.
A GIS class is one particular type of course which benefits immensely from explicit
instruction on spatial concepts and spatial thinking, but there are many others as well.
Spatial thinking can also be the topic of its own class. Because it is such a multi-faceted
and varied topic, it fits well into a learning situation designed for flexibility and
creativity, such as a First-Year Seminar for undergraduate students or a professional
development workshop for teachers or faculty, for example. Ideas and suggestions for
specific curricular ideas can be found at teachspatial.org.
5 Conclusion
The role of spatial thinking in, with, and about GIS and GIScience is necessarily
important but inevitably complex. There are substantial needs and opportunities for
educational research in these areas but significant progress has been hindered by the
lack of robust and reliable assessment instruments and the absence of funding available
for lines of inquiry that must necessary be multi-disciplinary. Even twenty years ago,
when the digital technologies were first being used more widely by researchers and
academics, it was clear that the relationships needed to be better understood.
What are the fundamental spatial concepts humans use and understand'? We
believe this question has wide significance for many aspects of GIS design,
education, social effects, etc. An important question yet to be answered,
however, is whether we should concentrate more on modifying GIS for
humans, or on training users to understand and use GIS effectively. Should
GIS be developed to mirror human capabilities or to compensate for human
limitations? The answer to these questions undoubtedly lies somewhere
between these choices. [8:175]
34
References
1. Dong, W., Li, L., Zhou, C., Yu Wang; Li, M.; Tian, C., Sun. W.: Discovery of
generalized spatial association rules. Service Operations and Logistics, and Informatics
(SOLI), 2012 IEEE International Conference Proceedings. (2012) 60-65
2. Freundschuh, S., Egenhofer, M.: Human Conceptions of Spaces: Implications for
Geographic Information Systems. Transactions in GIS 2(4) (1997): 361-375
3. Golledge, R.: The Nature of Geographic Knowledge. Annals of the Association of
American Geographers 92(1) (2002) 1–14
4. Jones, M. G., Gardner, G., Taylor, A., Wiebe E., Forrester, J.: Conceptualizing
Magnification and Scale: The Roles of Spatial Visualization and Logical Thinking.
Research in Science Education 41(3) (2011) 357-368
5. Lobben, A.: Tasks and Cognitive Processes Associated with Navigational Map
Reading. The Professional Geographer 56(2) (2004) 270–281
6. Monmonier, M.: How to Lie with Maps. 2nd edition. University of Chicago Press
(1996)
7. Montello, D.R.: Scale in Geography. In: Smelser, N. J., Baltes, P. B. (eds.),
International Encyclopedia of the Social and Behavioral Sciences Oxford: Pergamon
Press (2001) 13501-13504
8. Montello, D. R., Freundschuh, S. M.: Sources of spatial knowledge and their
implications for GIS: An introduction. Geographical Systems (2) (1995) 169-176
9. Montello, D. R., Waller, D., Hegarty, M., Richardson, A. E.: Spatial memory of real
environments, virtual environments, and maps. In: Allen, G. L. (ed.), Human spatial
memory: Remembering where. Lawrence Erlbaum Associates, Mahway, New Jersey
(2004) 251-285
10. National Research Council: Learning to Think Spatially. National Academies Press,
Washington, DC (2006)
11. O’Sullivan, D., Unwin, D.: Geographic Information Analysis, 2nd Edition. Wiley
Hoboken New Jersey (2010)
12. Sinton, D. S.: Spatial Thinking. In: Stoltman, J. (ed.), 21 st Century Geography: A
Reference Handbook. SAGE Publications, Thousand Oaks, California (2011) 734-744.
13. Sinton, D. S.: Spatial Learning in Higher Education. In: Montello, D. R., Grossner, K.,
Janelle, D. G. (eds.), Space in mind: Concepts for spatial learning and education. MIT
Press, Cambridge, Massachusetts (2014) 219-238
14. Sinton, D. S., Bednarz, S. W.: About that G in GIS. In: Sinton, D., Lund, J. (eds.):
Understanding Place: GIS and Mapping across the Curriculum. Redlands, CA: ESRI
Press, Redlands, California (2007) 19-34
15. Sinton, D. S., Kolvoord, B., Gersmehl, P., Bednarz, S., Uttal, D.: The People’s Guide
to Spatial Thinking. National Council for Geographic Education, Washington, DC
(2013)
16. Smith, B., Mark, D.: Geographical categories: an ontological investigation. Int. J.
Geographical Information Science 15(7) (2001) 591-612
17. Taylor, A., Jones, M. G.: Proportional reasoning ability and concepts of scale: surface
area to volume relationships in science. International Journal of Science Education
31(9) (2009) 1231–1247
18. Tobler, W.: A computer movie simulating urban growth in the Detroit region.
Economic Geography 46(2) (1970) 234-240
19. Tretter, T., Jones, M. G., Andre, T., Negishi, A., Minogue, J.: Conceptual boundaries
and distances: students’ and experts’ concepts of the scale of scientific phenomena.
Journal of Research in Science Teaching 43 (2006) 282–319
35