|
Introduction – Where
Is GIS? |
Spatial Reasoning
book |
Is the GIS Cart in Front of the
Horse? — discusses driving forces, trends and forecasts
in contemporary GIS from the perspective of modeling interrelationships among
mapped variables
Explore a New Spatial Paradigm
— discusses
the movement from mapping and spatial inventories by technologists to spatial
reasoning and dialog involving enlightened users in development of solutions to
complex spatial problems
<Click
here>
for a printer-friendly version of this
topic (.pdf).
(Back to the Table of Contents)
______________________________
Is the GIS
Cart in Front of the Horse?
(GeoWorld, July 1996)
What
began in the 1960s as a cartographer's tool has evolved into a revolution in
many disciplines. General users have
become more directly engaged with GIS technology, radically changing the nature
of GIS applications. Early uses
emphasized mapping and spatial database management. Increasingly, applications have moved to
modeling the interrelationships among mapped variables. Most of these applications have involved
cartographic modeling, which employs GIS operations to mimic manual map
processing techniques, such as map reclassification, overlay, and simple
buffering around features. The new wave
of applications concentrates on spatial modeling, employing spatial statistics
and advanced analytical operations. The
spatial modeling approaches can be grouped into three broad categories: data
mining, predictive modeling, and dynamic simulation.
Data
mining uses the GIS to discover relationships among mapped variables. For example, a map of dead and dying
spruce/fir timber stands can be statistically compared to maps of driving
variables, such as elevation, slope, aspect, soil, and depth to bedrock. If a strong spatial coincidence (correlation)
is identified for a certain combination of driving variables, that information
can be used to direct management action.
In the sickly tree example, if the dead trees tend to be on high, steep,
northern slopes with thin, acid soils, then forest managers can ask the GIS to
identify areas of trees living in these conditions and take appropriate
preemptive action. It's like the remark
made by famous robber Willy Sutton when asked why he robbed banks: "That's
where the money is." Often the
simple relationships hidden in complex data are revealed by a slightly different
perspective.
Another
form of data mining is the derivation of empirical models. For example, a geographic distribution (3-D
surface) of PCB concentrations in an aquifer can be interpolated from water
samples taken at several wells. Areas of
unusually high concentrations (more than one standard deviation above the
average) are isolated. When a time
series of the high-concentration maps is animated, the contamination appears to
move through the aquifer-hence, an empirical ground water model. A "blob" moving across the map
indicates an event, whereas a steady "stream" snaking its way along
indicates a continuous discharge of a Pollutant.
Most
predictive modeling is nonspatial. Data
are collected by sampling large areas, then reducing the set of measurements to
a single typical value (arithmetic average).
The average values for several variables are used to solve a
mathematical model, such as a regression equation. For example, a prediction equation for the
amount of breakage during timber harvesting is defined in terms of percent
slope, tree diameter, tree height, tree volume, and percent defect, with big,
old, rotten trees on steep slopes having the most breakage. The nonspatial approach ignores the inherent
spatial information collected and substitutes the average of each variable into
the equation to solve for a single estimate of breakage for an entire
area. A GIS solution, however, spatially
interpolates the field data into mapped variables, then
solves the equation for all locations in space.
That approach generates a map of predicted breakage with
"pockets" of unusual breakage levels clearly identified. Analogous procedures can detect pockets of
unusually high sales of a product, levels of crop productivity, or incidence of
disease.
Dynamic simulation allows
the user to interact with a spatial model.
Model behavior can be investigated by systematically changing the
model's parameters and tracking the results.
This "sensitivity analysis" identifies the relative importance
of each mapped variable within the context of its unique geographic
setting. In the timber breakage example,
the equation itself may be extremely sensitive to steep slopes. In a project area with only gentle slopes of
less than l0 percent, however, tree height might be identified as the most
important factor.
A less
disciplined use of dynamic simulation enables a GIS to act like a spatial
spreadsheet and address "what if" questions. For example, the avoidance of steep slopes
and visual connectivity to houses might be considered in a highway siting
model. What if steep slopes are
considered more important? Where does
the proposed route change, and where does it not change? What if visual connectivity is considered
more important? This informal use of
dynamic simulation actively involves decision makers and interested parties in
the map analysis process. The induced
dialogue develops a common understanding that greatly exceeds the information
packed in a static data sandwich of maps.
What is
the reality of these futuristic tools?
In many respects, the new applications might have "the cart in
front of the horse." GIS can store
tremendous volumes of descriptive data and overlay myriad maps for their
coincidence. It has powerful tools for
expressing the spatial interactions among mapped variables. There is, however, a chasm between GIS and
applied science. The reality is that the
bulk of our scientific knowledge lacks the spatial specificity in the
relationships among variables demanded by these advanced applications. We have a tool that characterizes spatial
relationships (cart); we lack the research and understanding of its expression
in complex systems (horse).
For
example, a GIS can characterize the change in a relative amount of edge in a
landscape by computing a set of fractal dimension maps as a forest is
modified. That and more than twenty
other landscape analysis indices allow us to track landscape dynamics, but what
changes in the indices mean for nesting birds is beyond our current scientific
knowledge. similarly,
a GIS can characterize the effective sediment-loading distance from streams as
a function of slope, vegetative cover, and soil type. It is common sense that areas with stable
soil on gentle, densely vegetated intervening slopes are farther away from a
stream (in terms of sediment-loading potential) than areas with unstable soils
and steep, sparsely vegetated intervening slopes.
But how
are effective sediment-loading distances translated into fish survival? Exactly where can a developer dig up the dirt
and not have the dirt balls rain down on the fish? Similarly, neighborhood variability
statistics allow us to track the diversity, interspersion and juxtaposition
of
vegetative cover types. How then are
these statistics translated into management decisions about wildlife
populations? Exactly where can a logger
cut trees without destroying the last spotted owl? These (and many others) are serious questions
that can't be solved by technology or science alone.
The ability
of GIS to integrate multiple phenomena is well-established. The functionality needed to relate the
spatial relationships among mapped variables is in place. What is lacking is the scientific knowledge
to exploit these capabilities. Until
recently, GIS was thought of solely as a manager's technology focused on
inventory and record-keeping. Even the
early scientific applications merely used it as an electronic planimeter to
aggregate data over large areas for input into traditional, nonspatial models. I hope we are embarking on an era of
scientific research in which spatial analysis plays an integral part and
expresses its results in GIS modeling terms.
The opportunity to have the scientific and managerial communities use
the same technology is unprecedented.
Until then, however, foresighted yet frustrated managers will be forced
to use the analytical power of GIS to construct their own models— based on
their common (and occasionally uncommon) sense.
Explore a New
Spatial Paradigm
(GeoWorld, April 1995)
Wha'cha mean a pair-o-dimes …heck, I don't even
have two nickels to rub together.
The
mechanics of GIS were made a lot easier in the last decade, but the
relationships and assumptions built into GIS models remain mere sketches of
uncharted intellectual terrain. Such is
the challenge GIS presents to basic and applied science.
But
what about the rest of us?
Aren't the scientists going to do it all, leaving us to merely click on
the right icon? What does the evolving
technology have in store for the general user?
In short, it offers a mind-expanding (or quite possibly, mind-exploding)
paradigm shift in how we perceive, handle and employ maps. We are shifting from a product-focus to a
utility-focus in our map dealings. No
longer is it what a map contains, but how “that map combined with this map and
eye of newt can produce what we really need.”
That takes us beyond mapping to spatial
reasoning, meaning that the process and procedures of manipulating maps
transcend the mechanics of GIS interaction.
The ability to think spatially becomes as important as “how do I do
that?"
Most
GIS users have cognitive skills reflecting their experiences (both good and
bad) with manual map processing and procedures.
Their data analysis experience has been with nonspatial data, or
measurements in which the spatial component was removed surgically, leaving
only an average value. But GIS offers a
host of new tools for analyzing mapped data.
It follows that these new tools will spawn a new way of doing business
with maps, beyond the vocational mastery of a system's user interface.
Cornerstone
to this new perspective is an appreciation that maps are data-numbers first,
pictures later. That is a radical
departure from our 8,000-year history of mapping. In the past, maps primarily were descriptive. They showed the precise placement of physical
features, usually for navigation purposes.
Increasingly, maps have become prescriptive,
serving as data in determining appropriate management actions. They tell us where it is (inventory), and
they provide insight into how it could be (analysis).
Map analysis is an
emerging discipline, recognizing fundamental map analysis operations
independent of specific applications.
These analytical tools extend mapping and management of spatial data to
GIS modeling, expressing relationships within and among mapped data. Familiarity with this map analysis toolbox is
the initial step toward spatial reasoning.
Bizarre concepts, such as a Standard Normal Variable surface or
Coefficient of Variation map, must become second nature before you can maneuver
your souped-up GIS like a race car driver.
Spatial
reasoning is the effective application of these tools to solve problems. That involves developing an understanding of
their appropriate use for particular applications. Conceptualization of an elk habitat model,
for example, requires an understanding of the important factors (mapped
variables) and how to express their interaction (tools) to identify areas of
suitable habitat (GIS solution). The GIS specialist, wildlife biologist and
natural resource manager must work closely at the onset of model
development. The problem can't be
dissected into separate pieces and solved independently. Their collective strength lies in the ability
to communicate various perspectives of the problem and its comprehensive
solution. Because elk habitat is
inherently a spatial problem, spatial reasoning becomes the medium of
intellectual exchange.
As the
civil rights adage says, "…the people with the problems are the people
with the solutions." Thus, the GIS
user must be involved in developing GIS solutions. You can't abdicate the responsibility of GIS
model development to someone who just happens to know how to boot-up the GIS
black box. Nor can you abandon the years
of indigenous knowledge about the unique character of your area to a scientist
in another region. Your obligations
extend even beyond spatial reasoning skills to spatial dialog deftness.
During
that final step, GIS is used as a decision support system. The focus is on consensus building and
conflict resolution among interested parties.
The GIS is used as a means to respond to a series of “what if” scenarios
in which any single map solution isn’t important. It is how maps change as different
perspectives are tried that develops enough information to support a
decision. That process includes an
understanding of the sensitivities involved in the decision. It also involves decision makers in the
analysis process instead of just choosing among a set of tacit decisions
(static alternatives) produced by detached analysis. Using GIS in this manner is a radical
departure from current spatial reasoning and dialog methodologies. When you reach this plateau, you and your
high-powered GIS are ready for the races.
One
more point needs to be made. GIS is
moving rapidly from the domain of the GIS specialist to the general user, and
is about to face the utilitarian user who lacks the sentimental attachment of
the earlier GIS zealots. In the past,
end users were content with automating existing manual processing and data
retrieval systems. As they become more
knowledgeable and adept in map analysis and modeling, we can expect increasing
demands on GIS that will push at the envelope of traditional concepts of map
content, structure and use. Maps will be
viewed less as descriptive images and more as mapped data expressing user
understanding of spatial interactions.
Effective communication of the spatial reasoning that supports modeled
maps will become as important as statements of map scale and projection.
What a
change of events. The technologists’
have been pushing GIS for years-both its virtues and products. We are at a point in time when things are
about to reverse, and that may be the technologists worst nightmare. Instead of a small, friendly user community
gratefully accepting the bug fixes and new features in each software update,
there is a growing community tugging at the existing set of GIS capabilities
and applications. The growing cadre of
enlightened users is pushing GIS into new areas the technologists never dreamed
of. But it's the only way to get your
nickel’s worth out of GIS ...maybe even a new pair of dimes.
_____________________
(Back
to the Table of Contents)