Environmental Modelling & Software 17 (2002) 287–294

www.elsevier.com/locate/envsoft

Examining GIS decision utility for natural hazard risk modelling

*
A. Zerger
Centre for GIS and Modelling, Department of Geomatics, The University of Melbourne, Parkville, Melbourne, Vic. 3010, Australia

Received 24 April 2001; received in revised form 15 August 2001; accepted 3 September 2001

Abstract

The research examines the incompatibility between resolution and scale of spatial data in a spatial modelling environment and
the resolution and scale of human spatial decision-making. This incompatibility is rarely considered when a geographic information
systems (GIS) is applied to natural hazard risk assessment. However, the ultimate success of a risk management project should be
assessed in the context of improved decision-making. Decision utility is an emerging theme in GIS literature that focuses on
cognitive issues of GIS and human interactions. The research presents a technique for flood risk modelling using GIS and digital
elevation models to map relative risk in urban communities. The study is based on Cairns located in far northern Australia. The
risk model accounts for uncertainties inherent in the elevation data by adapting an existing error simulation technique. Techniques
for making spatial model assumptions and model error explicit to flood risk managers are introduced and evaluated for risk manage-
ment decision-making.  2002 Elsevier Science Ltd. All rights reserved.

Keywords: Natural hazards; Geographic information systems; Decision utility

1. Introduction ever, GIS spatial analysis techniques may introduce

Our knowledge of hazard phenomena and the pro-
cesses that drive them are imperfect. It is therefore
necessary to develop appropriate models (process, spa-
tial and temporal) to fill the gap. The synthesis of data
and the mapping of the relationships between the hazard
phenomena and the elements at risk require the use of
tools such as geographic information systems (GIS).
There are advantages in developing a fusion between the
philosophy of risk management and the strength of GIS
as a decision support tool. However, the ultimate success
of such an approach should be assessed in the context
of improved decision-making.
Coppock (1995) notes that there are grounds for
believing that GIS has an important function to play
because natural hazards are a multi-dimensional
phenomena which have a spatial component. Examples
include cartographic approaches for mapping the physi-
cal hazard, integrative hazard modelling and spatial
decision support systems (de Silva et al., 1993) and dis-
aster response planning (Zografos et al., 1994). How-
* Fax: +61-3-934-2916.
E-mail address: a.zerger@unimelb.edu.au (A. Zerger).
problems unique to the technology during the data inte-
gration and analysis process (Rejeski, 1993). Problems
include the degree of uncertainty that can be associated
with model results owing to the choice of the model
used, and the role of error in the input data and how it
effects the outcomes of the risk assessment. These fac-
tors contribute to the overall uncertainty in the results
of risk models. Because the objective of natural hazard
risk assessment is to reduce uncertainty, this should be
extended to the GIS and spatial analysis process as well
(Rejeski, 1993; Emmi and Horton, 1995; Murillo and
Hunter, 1996; Zerger, 1998).
This paper first examines existing approaches to
assessing model decision utility with a particular empha-
sis on GIS and spatial modelling. A GIS-based sensi-
tivity analysis technique for modelling the uncertainty in
storm surge inundation risk for Cairns, Australia is then
described. The technique accounts for storm surge inun-
dation model and database uncertainty. These spatial
measures of uncertainty are included in the final hazard
risk assessment. The objective is to develop a technique
which communicates uncertainty and has practical utility
for emergency management decision-makers in Cairns.
The final stage of the research examines the decision
utility of these techniques and map outputs for practical

1364-8152/02/$ - see front matter  2002 Elsevier Science Ltd. All rights reserved.
PII: S 1 3 6 4 - 8 1 5 2 ( 0 1 ) 0 0 0 7 1 - 8
288 A. Zerger / Environmental Modelling & Software 17 (2002) 287–294
risk management decision-making in Cairns. Issues,
including user perceptions, cognitive responses and
decision utility, are expected to receive greater attention
as GIS shifts from its technological focus, to a broader
geographic information science and decision support
emphasis. It is also expected that information systems
managers will question the overall benefits of GIS
implementation for risk management. The recurring
question in this research and in a broader sense is likely
to be: Does the model improve decision-making and to
what extent?
2. Assessing cognitive responses in spatial sciences
The success of a modelling initiative should be
assessed in the context of improved decision-making.
However, few GIS studies have examined the conse-
quences of spatial modelling on decision-making.
Nyerges (1995) suggests the reason is that few GIS prac-
titioners know how to study people. This is a concern for
natural hazard risk management because the objective of
GIS modelling is to improve the accuracy and effective-
ness of decision-making for improved public safety.
Hearnshaw (1995, p.196) summarises the importance of
user interaction studies for GIS below.
Human–computer interaction comprises phenomena
and a discipline which takes those phenomena as its
scope. The phenomena involve systems consisting of
people — both as individuals and as social organis-
ations; computers — both stand-alone and as net-
works; and their interaction. Since the systems are
physical and informational, so too are their interac-
tions. The discipline is concerned to support the
optimisation of the interactions between humans and
computers to perform work more effectively. The
concern, then, is not with the interactions in isolation.
Humans use computers to do work and also have per-
formance requirements for the work which is carried
out. Interactions and their optimisation, then, need to
be developed in the context of work and performance.
The transformation of GIS to a decision support tool
creates new challenges that inevitably include human
interaction issues. Traditional empirical user perception
studies have focused on the cartographic representation
of analogue maps. The studies have addressed carto-
graphic issues such as colour, shade, hue, map generalis-
ation, map symbols and other cognitive variables. Mon-
monier (in Rejeski, 1993) notes that because mapped
messages can be manipulated, a considerable amount of
care must be given to the design of cartographic presen-
tations. Thorndyke and Stasz (1980) assessed knowledge
acquisition from maps by people who had both training
and no training in map use, using quantitative statistical
techniques. The study focused on assessing an individ-
ual’s spatial knowledge acquisition, rather than assessing
cartographic representations. Similarly, Wood (1995)
used quantitative techniques such as eye movement
tracking to assess whether map readers notice map gen-
eralisation. The objective of the research was to examine
whether reducing the amount of detail in a map will
reduce the amount of processing required to perform a
map analysis task.
The interaction of a decision-maker with decision aids
using process tracing methods to understand the interac-
tion between tasks, decision-makers and information
technology in the spatial domain is discussed by Todd
(1995). Such techniques are the domain of decision, cog-
nitive sciences and psychology. Tracing methods include
studying computer logs, information display boards,
tracing eye movements and verbal protocols. Two
approaches nested within the verbal protocol tracing
method are relevant to this research. The first is struc-
tured elicitation which is analogous to the structured
interview. In this model the subject is asked specific
questions about the decision-making process. The
second is neutral-probing, which is analogous to
unstructured interviewing. In this model, the subject is
encouraged to ‘think aloud’ and observations are
recorded. As with structured interview techniques, the
structured tracing method is more suited to quantitative
analysis and comparison across subjects (Todd, 1995).
Regardless of the approach chosen, the objective of the
sampling is to obtain information-rich responses.
Schweizer and Goodchild (1992) assessed the ability
of shades of grey to represent spatial uncertainty in chor-
opleth maps. The study is unique because it examines
the utility of presenting uncertainty information to map
readers. Results indicate that representing spatial uncer-
tainty presents significant challenges to cartographers
because map readers need to be familiar with the
uncertainty/colour association through repeated use.
Similarly Hernandez (1995) notes that novice users will
have different spatial concepts than expert users (also
see Nyerges (1995) for a detailed treatment of user needs
and GIS). Schweizer and Goodchild (1992, p. 696)
note that:
Maps do not customarily display quality information
and the introduction of such a variable can be confus-
ing to map readers. Verbal communication with parti-
cipants indicated the mapping of data quality is an
unfamiliar concept and selections were made arbi-
trarily or according to individual biases. This suggests
education of users may improve the effectiveness of
displaying data quality.
Few studies have examined the consequences of car-
tography for improved decision-making. Crossland et al.
(1993) note that ‘no studies have been identified which
 A. Zerger / Environmental Modelling & Software 17 (2002) 287–294
address cognitive style as it relates to decision-making
related to spatial problems, particularly those
addressable by a geographic information system’. This
theme is examined by Crossland et al. (1993) who sug-
gest that we are seeing a fundamental shift in focus from
visual thinking in the private realm, to visual communi-
cation in the public realm. Cognitive cartographic-stud-
ies have focused on private realm visual thinking. The
proposed extension of cognitive studies into decision-
making assessment represents visual communication in
the public realm.
A feature of formal cartographic training is a treat-
ment of cognitive issues associated with map production.
The emphasis on cognitive user-interaction issues is
commonly lacking in formal GIS training. A GIS expert
is typically knowledgeable with software, spatial data
and spatial modelling issues rather than human percep-
tion issues. This is reflected in a general lack of consider-
ation of human interaction issues in GIS development.
The only example of cognitive research applied to natu-
ral hazard risk management that could be identified was
the study by Dymon and Winter (1993) that examined
map design criteria in risk maps. The study firstly exam-
ined visual representation and found that a majority of
maps lacked scale bars, north points and other key carto-
graphic features. The study then established a decision
model based on a series of questions a family, or an
individual, will use in making emergency evacuation
decisions. The study found that the most important carto-
graphic element needed is a clear map hierarchy. The
hierarchy refers to the immediate operative needs of the
map user. A clear map hierarchy is one that allows users
to obtain all relevant information with a quick glance.
In this research, map users are emergency managers in
Cairns and the analogy of a clear map hierarchy is used
to test the efficacy of uncertainty-based risk maps for
Cairns.
Crossland (1992) presents a quantitative technique for
assessing whether people make better decisions using
GIS. Crossland compared the use of paper maps and a
GIS display to assess the time taken to make a decision
and the error rate of the decisions. The scenario was
hypothetical (a utility company site location problem)
and tested the responses of 110 subjects. The study con-
cluded that GIS did improve the response time and pro-
vided more accurate solutions. This research differs from
Crossland’s study in two ways. First, the study popu-
lation in Crossland’s study had no prior experience with
the objective of the decision task. Second, the objective
in this research is to assess the utility of presenting
uncertainty results, rather than to compare computer-
based displays with map-based displays. Ives (1982, p.
8) similarly addresses some of the deficiencies of such
quantitative cognitive studies and proposes a different
research agenda focusing on decision utility. There are
also assumptions in Crossland’s research that a faster
289
decision is a better decision and that there is indeed a
correct decision (see also Crossland et al., 1993). For
storm surge hazard risk management none of these
assumptions are necessarily applicable.
3. Study site and surge risk modelling
A storm surge is the term used to describe an anomal-
ous elevation of ocean water typically some 50 km
across generated by the action of a cyclone and the coas-
tal bathymetry. Loss of life in Australia from storm surge
inundation has been minimal in comparison to events
elsewhere. Modelling surge inundation in urban areas
vulnerable to storm surge is an important component of
risk management for the following reasons:
앫 emergency management evacuation planning,
앫 developing urban zoning that accounts for storm surge
inundation and hence risk,
앫 as an educational tool to inform citizens of the risks
present in their community,
앫 identifying risk for possibly establishing insurance
premiums,
앫 developing building codes to minimise the impacts
of surge,
앫 cost-benefit analysis for developing mitigation stra-
tegies,
앫 managing post-disaster recovery.
Cairns is located in far northern Australia and has a
population of approximately 100,000. This coastal area
has been labelled ‘The New California’ with growth
rates that place it in the top 10 fastest growing urban
areas in the developed world (Skinner et al., 1993). The
linear nature of the coastal range in Cairns combined
with the desire for beach frontages restricts urbanisation
to a North–South corridor (Fig. 1). This leads to a greater
Fig. 1. Cairns study site in far northern Australia.
290 A. Zerger / Environmental Modelling & Software 17 (2002) 287–294
probability of massive losses from storm surge. In
addition, Cairns is amongst the most remote regional
coastal centres in Australia which magnifies the conse-
quences of any cyclone disaster. Although the incidence
of major cyclones in Australia is relatively uncommon,
events such as cyclone Tracy and Justin remind residents
and policy makers alike that a risk exists.
The key to developing emergency response strategies
for storm surge in Cairns is to identify spatially where
the risk will be greatest. A key indicator of surge risk is
the buildings and roads that may be inundated. Detailed
building databases were developed for Cairns from field
work and council databases. Fig. 1 shows the distribution
of the road network, and hence urbanisation in Cairns.
Approximately 25,000 buildings were integrated in a
vector Arc/Info (ESRI, 1997) spatial database. This
included attribute information such as building floor
heights, building addresses and the number of stories of
each building.
Until recently an overland storm surge model for
Cairns did not exist. To establish a model, a flatwater
sensitivity analysis approach was used. Inundations are
modelled by systematically increasing the inundation
level based on a digital elevation model (DEM) and
assessing the resulting building and road flooding from
50 to 1500 cm above the Australian Height Datum
(AHD) in increments of 50 cm. Sensitivity analysis is
particularly valuable for hazard risk assessment because
앫 The relative importance of input parameters can be
assessed (floor heights, DEM accuracy and
resolution).
앫 Risk managers can identify emergency management
‘hot-spots’. These are regions that have a relatively
high risk in the study domain.
앫 It can identify probable maximum risks where the tra-
ditional approach considers probable maximum haz-
ard (in this case storm surge hazard). This paper con-
tends that probable risk is a more appropriate
construct for risk management.
앫 Sensitivity analysis can identify ‘cold-spots’. These
are regions that can be used as evacuation safe areas.
앫 It can identify and define emergency management
‘catchments’ or regions within the study domain that
have a similar risk in the context of evacuation plan-
ning and decision-making.
앫 It can identify the robustness of risk analysis to errors,
scales and uncertainties in GIS data and models.
The model is implemented using the ARC/INFO GIS
and the ARC/INFO Macro Language (AML). Inundation
is a function of the ground elevation at each building
derived from a 20 m cell resolution DEM constructed
for Cairns using ANUDEM (Hutchinson, 1998) surface
interpolation algorithms, building floor heights derived
from field work and the flatwater surge inundation
model.
4. Modelling uncertainty
Uncertainty, in contrast to error, assumes that no prior
knowledge of the accuracy of the data exists. Few
examples of GIS applications to natural hazards risk
assessment consider data or model accuracy, or com-
municate the inherent uncertainties in data and models
in the final risk assessment. Commonly, model and data
uncertainty estimates are esoteric concepts which have
little practical use to risk managers. These may include
measures such as root mean square errors, probability
surfaces and classification error matrices.
Three possible sources of error are present in the
Cairns risk model including the vertical error in the
DEM, error in the floor heights and the error in the storm
surge model. The accuracy of the elevation model is a
function of the scale and accuracy of input topographic
data and the ANUDEM interpolation process. DEM
error is the focus of the uncertainty modelling method-
ology in this research. As a first step a measure of DEM
error has been derived from high accuracy permanent
survey markers in Cairns. A detailed DEM error assess-
ment found that 90% of the DEM was correct to within
±200 cm (root mean square error).
The concern with these errors is the implications for
risk predictions for low magnitude and high frequency
inundation events in Cairns. Significant increases in the
number of buildings inundated over floor level are
observed at increments less than the vertical accuracy of
the DEM. A critical question is: how much faith can
be placed in the inundation results when the model is
dependent on DEM accuracy? One outcome is that our
confidence in surge estimates must be questioned under
particular low magnitude scenarios. These events are
important in Cairns due to their relatively higher fre-
quency of occurrence.
Openshaw (1989) noted that the problem is not as
much the existence of uncertainty, but rather that the
traditional response has been to ignore it on the grounds
that methods to handle it do not exist. Goodchild et al.
(1992) note that with the existence of uncertainty in
models and spatial data there are three options to deal
with it: (a) omit all reference to it, (b) attach some form
of description to the output, and (c) show samples from
the range of maps or outputs possible. The first option
is unacceptable for hazard risk management and evacu-
ation planning and the second may not adequately com-
municate this complex concept to risk managers. Exam-
ining the third option is preferable because “it would
appear to have the greatest potential benefit in both com-
municating uncertainty and at the same time educating
the user community to the significance of the issue”
 A. Zerger / Environmental Modelling & Software 17 (2002) 287–294
(Hunter et al., 1994). The third option is adapted for
modelling storm surge inundation uncertainty and risk
in Cairns.
4.1. Grid cell model of uncertainty
A variety of methods exist for handling uncertainty in
spatial data including methods such as analytical error
propagation models and empirical stochastic approaches.
The research adapts the grid cell model of uncertainty
(GCUM) (Hunter et al., 1994) which has been used pre-
viously to determine the uncertainty of slope and aspect
estimates derived from spatial databases, for estimating
the influence of uncertainty on wildfire mapping using
remotely sensed imagery and for landslide suscepti-
bility mapping.
The GCUM applies noise to a source DEM to simu-
late the error present in the data by randomly perturbing
the elevation values to create new realisations of the
DEM. This realisation is then used in the model and new
risk results are derived. The realisations must be both
stochastic, yet sufficiently spatially correlated to truly
represent the spatial dependencies inherent in elevation.
This is achieved by constraining the random pertur-
bations by a spatial autocorrelation index (Cliff and
Ord, 1981)
5. Inundation risk modelling results
The output provided by the model is a probability of
over floor flooding for each building in the range 0–
100% for each inundation scenario. Continuous risk
probability maps can be re-classified into binary maps
of risk by selecting a threshold value below which
flooding is not likely. This should be avoided because
valuable information would be lost and such represen-
tations reinforce the perception of risk as an absolute.
Results from the model of building and road network
inundation for Cairns are firstly stored as attributes in
the databases and as digital animations at
http://www.geom.unimelb.edu.au/zerger/maps.html. Sample
inundation results are shown in Fig. 2 for four inun-
dation scenarios.
Results confirm that the greatest changes in risk occur
at relatively low inundation (⬍400 cm above AHD).
Although probable maximum storm surges
(approximately 600 cm above AHD) may have dramatic
consequences, results show that these would not be sig-
nificantly greater than consequences for relatively low
inundations. The animations also highlight the impor-
tance of critical threshold inundations in the study
domain.
The inclusion of uncertainty estimates into inundation
risk models has perceived advantages over binary maps
of risk. First, it includes measures of uncertainty into the
291
risk model and data and model assumptions are made
explicit to end-users. In other words, spatial uncertainty
can be visualised in the final risk maps. And second, the
techniques provide an indication of the relative risk of
inundation. Prioritising evacuation zones based on these
results may lead to more informed emergency manage-
ment decision-making. The following discussion exam-
ines the practical decision utility of these results for risk
management in Cairns.
6. Assessing GIS decision utility
Computer and paper maps are powerful tools for com-
municating risk analysis and assessment results and to
aid decision-making. A responsibility exists for GIS risk
practitioners to ensure that model assumptions have been
presented and that the cartographic representation is con-
ducive to improved decision-making. Decision utility is
an emerging theme in GIS literature that focuses on cog-
nitive issues of GIS and human interactions (Rejeski,
1993). It recognises that a fundamental objective of GIS
modelling is to reduce the uncertainty inherent in spatial
decision-making. It is therefore critical that GIS design
should be assessed in the context of improved
decision-making.
Because the essence of evacuation planning is the
rapid spatial movement of people, goods and services,
maps are one of the best media to depict this (Dymon
and Winter, 1993). This paper contends that in GIS-
based natural hazard risk assessment (and other spatial
modelling for that matter), an incompatibility exists
between the resolution and scale of spatial data in GIS
and the resolution and scale of human spatial decision-
making. Rajeski (1993, p. 323) highlights this incom-
patibility:
There can be no doubt that increasing the believability
and honesty of GIS products will increase their utility
for decision-makers. However, developing a close fit
between map analysis and policy goals is of such
critical importance that it requires a considerable level
of effort and foresight on the part of the analysts and
map makers. When maps fail to have impact, it often
has little to do with model nuances or data gaps, but
more to do with a lack of fit between cartography and
decision reality.
In this research, decision-making is concerned with
evacuation planning by emergency managers. Few stud-
ies have attempted to integrate decision-scale and resol-
ution into GIS-based risk modelling and fewer have
assessed the utility of uncertainty-based cartographic
representations. Semi-structured interviews were
presented to key emergency management stakeholders
to determine the utility of uncertainty-based risk maps
292 A. Zerger / Environmental Modelling & Software 17 (2002) 287–294
Fig. 2. Risk model results for 2, 3, 4 and 5 m inundation in Cairns showing the probability of flooding for each building derived from uncertainty
model (relative risk).
for evacuation planning and decision support (15 stake-
holders in total). Stakeholders included Cairns City
Council risk managers, Queensland Department of
Emergency Services policy staff and Queensland Police
Service officers.
6.1. Results of decision utility assessment
The aim of the interview process was to determine
the utility of uncertainty-based risk maps for evacuation
planning and decision support. The overwhelming con-
clusion was that for practical storm surge evacuation
planning, the maps provide too much spatial detail to be
practical. However, this conclusion should be considered
in the context of the functional purpose of the maps i.e.
risk mapping for storm surge emergency management
evacuation planning. For other functional purposes
including urban planning, riverine flood evacuation and
mitigation cost-benefit analysis the conclusions will dif-
fer. In general, risk managers concluded that storm surge
evacuation decision-making operates at a coarser spatial
resolution than the uncertainty-based GIS risk mapping.
In addition:
앫 Risk managers already account for a range of other
uncertainties when delineating evacuation zones such
as the uncertainty inherent in Bureau of Meteorology
forecasts. These were noted to overwhelm the uncer-
tainties inherent in inundation models and hence the
modelling of spatial uncertainty is somewhat redun-
dant.
앫 Because the risk management objective in Cairns is
to encourage self-evacuation, prioritisation of evacu-
ation is not as critical as first contended. Also, risk
managers will prioritise on the basis of vulnerability.
They will target hospitals and other groups that can-
not self-evacuate, rather than on the basis of the raw
hazard zones.
앫 Risk managers commonly assume a worst-case scen-
ario. At high inundations (⬎6 m AHD) the maps
effectively become binary maps of risk because the
coastal regions are topographically homogenous and
any delineation of relative risk is ignored.
앫 Risk managers currently evacuate entire regions
rather than individual buildings. The resolution of the
database is too fine for evacuation planning.
앫 Although risk management hot-spots have been ident-
ified in the maps, risk managers prefer to identify ‘risk
catchments’. Catchments are areas that have similar
risk estimates and require similar risk treatments
(evacuation instructions).
앫 There is a resistance to provide uncertainty-based
maps to the public because they imply policy uncer-
tainty and may result in individual decision ambi-
guity. This is a major problem when evacuation
orders are issued.
Responses also highlight the common incompatibility
between scientific modelling objectives and practical
decision-making. Coarser resolution and less accurate
spatial data may be adequate for evacuation decision-
making. If user considerations are assessed early in the
risk modelling process, GIS practitioners can minimise
data capture, avoid unnecessary levels of complexity in
the spatial modelling and generally improve the utility
of the risk modelling for decision-making. A decision
 A. Zerger / Environmental Modelling & Software 17 (2002) 287–294
utility assessment should be a key step in future appli-
cations of GIS to hazard risk management.
Although it was concluded that uncertainty-based risk
maps are too detailed for evacuation planning, these rep-
resentations of relative risk are nevertheless useful for
other hazard risk management activities. Most respon-
dents identified the suitability of this approach for slow-
onset riverine flooding. Risk managers working on pol-
icy formulation suggested the methodology would be
important for assessing mitigation strategies where a
sensitivity analysis is required. The uncertainty method-
ology is more appropriate for planning, preparedness and
assessing mitigation options, than for hazard event
response.
A unique behavioural interpretation of the maps was
observed with all risk managers. Most had little or no
experience with root mean square error estimates (as cal-
culated for DEMs) and few had practical GIS experi-
ence. Describing the concept of spatial uncertainty
proved challenging and describing the uncertainty mod-
elling technique was difficult. However, when risk man-
agers were challenged to make evacuation decisions on
the basis of the maps, they intrinsically identified the
dark grey points as high risk areas (Fig. 2) and the light
grey as having a lower risk. The underlying method-
ology behind this relative risk modelling was rarely
questioned and there was an intrinsic acceptance of
uncertainty-based results, regardless of their understand-
ing of this complex concept. Therefore cartographic rep-
resentations are important for presenting uncertainty
information without explicitly identifying it as such. This
is both a strength and a weakness of GIS-based risk
modelling.
7. Conclusions
The application of a GIS and spatial modelling for
natural hazard risk management is an emerging science
and the ability to provide a spatial context for risk is
critically important for risk reduction. A precautionary
outlook is required because GIS-based risk modelling
introduces concepts such as uncertainty that are rela-
tively new to risk managers. The recent interest by com-
mercial GIS vendors in emergency services and risk
management places undue emphasis on software and
hardware issues.
Research has shown that non-technical issues such as
assessing and planning for decision utility and the fitness
of use of spatial data may be greater barriers to success-
ful GIS model implementation. Regardless of the natural
hazard, GIS software, spatial data or the risk model, the
use of a GIS for risk management should be assessed
in the context of improved decision-making. This has
important implications for other spatially explicit model-
ling that aims to address decision support needs. If user
293
considerations are assessed early in the risk modelling
process, GIS practitioners can minimise data capture,
avoid unnecessary levels of complexity in the spatial
modelling and generally improve the utility of risk mod-
elling for decision-making.
Acknowledgements
This research was funded by the Australian National
Committee for the IDNDR and was made possible with
the assistance of Cairns City Council and the Tropical
Cyclone Coastal Impacts Program. Thanks are extended
to Dr Gary Hunter, The University of Melbourne for
making the Grid Cell Uncertainty Model available and
to Dingle Smith (CRES-ANU) for his ongoing and sig-
nificant contribution to the research. The author
acknowledges the contribution of an anonymous
CSIRO reviewer.
References
Cliff, A.D., Ord, J.K., 1981. Spatial Processes: Models and Appli-
cations. Pion Limited, London.
Coppock, J.T., 1995. GIS and natural hazards: an overview from a
GIS perspective. In: Carrara, A., Guzzetti, F. (Eds.), Geographical
Information Systems in Assessing Natural Hazards, vol. 6. Kluwer
Academic, Netherlands, pp. 21–34.
Crossland, M.D., 1992. Spatial decision support systems: an overview
of technology and a test of efficacy. In: AM/FM International. San
Antonio, Texas.
Crossland, M., Scudder, J.N., Herschel, R.T., Wynne, B.E., 1993. Mea-
suring the relationships of task and cognitive style factors and their
effects on individual decision-making effectiveness using a geo-
graphic information system. In: Nunamaker, J.F., Sprague, R.H.
(Eds.), Proceedings of the 26th Annual Hawaii International Con-
ference on Systems Sciences, vol. 4. IEEE Computer Society Press,
Hawaii, pp. 575–584.
de Silva, F., Pidd, M., Eglese, R., 1993. Spatial decision support sys-
tems for emergency planning: an operational research/geographic
information systems approach to evacuation planning. In: Sullivan,
J.D. (Ed.), International Emergency Management and Engineering
Conference. Society for Computer Simulation, pp. 130–133.
Dymon, U.J., Winter, N.L., 1993. Evacuation mapping: the utility of
guidelines. Disasters 17, 12–24.
Emmi, P.C., Horton, C.A., 1995. A Monte Carlo simulation of error
propagation in a GIS-based assessment of seismic risk. Inter-
national Journal of Geographical Information Systems 9, 447–461.
ESRI, Arc/Info, 1997. Environmental Systems Research Institute.
Redlands, CA.
Goodchild, M.F., Guoqing, S., Shiren, Y., 1992. Development and test
of an error model for categorical data. International Journal of Geo-
graphical Information Systems 6, 87–104.
Hearnshaw, H., 1995. Psychology and displays in GIS. In: Hearnshaw,
H.M., Unwin, D.J. (Eds.), Visualisation in Geographic Information
Systems. Wiley, Chichester, pp. 193–199.
Hernandez, D., 1995. HCI aspects of a framework for the qualitative
representation of space. In: Nyerges, T., Mark, D.M., Laurini, R.,
Egenhofer, M.J. (Eds.), Cognitive Aspects of Human–Computer
Interaction for Geographic Information Systems, vol. 83. Kluwer,
Dordrecht, pp. 45–60.
294 A. Zerger / Environmental Modelling & Software 17 (2002) 287–294
Hunter, G.J., Goodchild, M.F., Robey, M., 1994. A toolbox for
assessing uncertainty in spatial databases. In: Masters, E.G. (Ed.),
Proceedings of the 22nd Annual Conference of the Australasian
Urban and Regional Information Systems Association Inc., Sydney.
Hutchinson, M.F., 1998. Calculation of hydrologically sound digital
elevation models. In: Proceedings: Third International Symposium
on Spatial Data Handling International Geographical Union, Syd-
ney, Australia, pp. 117–133.
Ives, B., 1982. Graphical user interfaces for business information sys-
tems. MIS Quarterly, 14-42.
Murillo, M.L., Hunter, G.J., 1996. Evaluating uncertainty in a landslide
susceptibility model. In: Second International Symposium on spa-
tial data accuracy, Fort Collins, USA.
Nyerges, T.L., 1995. Cognitive issues in the evolution of GIS user
knowledge. In: Nyerges, T., Mark, D.M., Laurini, R., Egenhofer,
M.J. (Eds.), Cognitive Aspects of Human–Computer Interaction for
Geographic Information Systems, vol. 83. Kluwer, Dordrecht, pp.
61–74.
Openshaw, S., 1989. Learning to live with errors in spatial databases.
In: Goodchild, M., Gopal, S. (Eds.), Accuracy of Spatial Databases.
Taylor and Francis, London, p. 290.
Rejeski, D., 1993. GIS and risk: a three culture problem. In: Goodchild,
M.F., Parks, B.O., Steyaert, L.T. (Eds.), Environmental Modeling
with GIS. Oxford University Press, New York, pp. 318–331.
Skinner, J.L., Gillam, M.E., O’Dempsey, T.M., 1993. The new Califor-
nia? Demographic and economic growth in Queensland. In: Briton,
N.R., Oliver, J. (Eds.), Catastrophe Insurance for Tomorrow: Plan-
ning for Future Adversities. Griffith University, Australia, pp.
249–274.
Schweizer, D.M., Goodchild, M.F., 1992. Data quality and choropleth
maps: an experiment with the use of color. In: GIS/LIS, vol. 2.
URISA, San Jose, California, pp. 686–699.
Thorndyke, P.W., Stasz, C., 1980. Individual differences in procedures
for knowledge acquisition from maps. Cognitive Psychology 12,
137–175.
Todd, P., 1995. Process tracing methods in the decision sciences. In:
Nyerges, T., Mark, D.M., Laurini, R., Egenhofer, M.J. (Eds.), Cog-
nitive Aspects of Human–Computer Interaction for Geographic
Information Systems, vol. 83. Kluwer, Dordrecht, pp. 77–96.
Wood, C.H., 1995. Do map readers really notice and use generaliz-
ation? The perceptual consequences of line simplification in a task
oriented thematic map analysis experiment. In: 17th International
Cartographic Conference, Barcelona, Spain, vol. 1., pp. 118–127.
Zerger, A., 1998. Including model uncertainty estimates into tropical
cyclone risk modelling in Northern Australia. In: GIS 98/Resource
Technology 98. GISWORLD. Toronto, Canada, pp. 8.
Zografos, K.G., Douligeris, C., Tsoumpas, P., 1994. Using a GIS plat-
form for design and analysis of emergency response operations. In:
Sullivan, J.D. (Ed.), The International Emergency Management and
Engineering Conference, Hollywood, FL, pp. 14–19.