Directory UMM :Data Elmu:jurnal:A:Aquacultural Engineering:Vol23.Issue1-3.Sept2000:
Aquacultural Engineering 23 (2000) 233 – 278
www.elsevier.nl/locate/aqua-online
Applications of geographical information
systems (GIS) for spatial decision support in
aquaculture
Shree S. Nath a,*, John P. Bolte b, Lindsay G. Ross c,
Jose Aguilar-Manjarrez d
a
Skillings-Connolly, Inc., 5016 Lacey Boule6ard S.E., Lacey, WA 98503, USA
Department of Bioresource Engineering, Oregon State Uni6ersity, Cor6allis, OR 97331, USA
c
Institute of Aquaculture, Uni6ersity of Stirling, Stirling FK9 4LA, UK
d
Food and Agricultural Organisation of the U.N, Viale delle Termi di Caracalla, 00100 Rome, Italy
b
Received 1 September 1999; accepted 2 October 1999
Abstract
Geographical information systems (GIS) are becoming an increasingly integral component
of natural resource management activities worldwide. However, despite some indication that
these tools are receiving attention within the aquaculture community, their deployment for
spatial decision support in this domain continues to be very slow. This situation is
attributable to a number of constraints including a lack of appreciation of the technology,
limited understanding of GIS principles and associated methodology, and inadequate
organizational commitment to ensure continuity of these spatial decision support tools. This
paper analyzes these constraints in depth, and includes reviews of basic GIS terminology,
methodology, case studies in aquaculture and future trends. The section on GIS terminology
addresses the two fundamental types of GIS (raster and vector), and discusses aspects related
to the visualization of outcomes. With regard to GIS methodology, the argument is made for
close involvement of end users, subject matter specialists and analysts in all projects. A
user-driven framework, which involves seven phases, to support this process is presented
together with details of the degree of involvement of each category of personnel, associated
activities and analytical procedures. The section on case studies reviews in considerable detail
four aquaculture applications which are demonstrative of the extent to which GIS can be
deployed, indicate the range in complexity of analytical methods used, provide insight into
* Corresponding author. Tel.: + 1-360-4913399; fax: +1-360-4913857.
E-mail address: snath@skillings.com (S.S. Nath)
0144-8609/00/$ - see front matter © 2000 Elsevier Science B.V. All rights reserved.
PII: S 0 1 4 4 - 8 6 0 9 ( 0 0 ) 0 0 0 5 1 - 0
234
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
issues associated with data procurement and handling, and demonstrate the diversity of GIS
packages that are available. Finally, the section on the future of GIS examines the direction
in which the technology is moving, emerging trends with regard to analytical methods, and
challenges that need to be addressed if GIS is to realize its full potential as a spatial decision
support tool for aquaculture. © 223 Elsevier Science B.V. All rights reserved.
Keywords: GIS; Aquaculture; Spacial decision support
1. Introduction
The rapid growth of aquaculture worldwide has stimulated considerable interest
among international technical assistance organizations and national-level governmental agencies in countries where fish culture is still in its infancy, and has resulted
in increased concerns about its sustainability in countries where the industry is well
established. Planning activities to promote and monitor the growth of aquaculture
in individual countries (or larger regions) inherently have a spatial component
because of the differences among biophysical and socio-economic characteristics
from location to location. Biophysical characteristics may include criteria pertinent
to water quality (e.g. temperature, dissolved oxygen, alkalinity/salinity, turbidity,
and pollutant concentrations), water quantity (e.g. volume and seasonal profiles of
availability), soil type (e.g. slope, structural suitability, water retention capacity and
chemical nature) and climate (e.g. rainfall distribution, air temperature, wind speed
and relative humidity). Socio-economic characteristics that may be considered in
aquaculture development include administrative regulations, competing resource
uses, market conditions (e.g. demand for fishery products and accessibility to
markets), infrastucture support, and availability of technical expertise. The spatial
information needs for decision-makers who evaluate such biophysical and socioeconomic characteristics as part of aquaculture planning efforts can be well served
by geographical information systems (GIS; Kapetsky and Travaglia, 1995).
Moreover, it is often the case that governmental agencies involved with issuing
new aquaculture permits need to perform spatial analysis on a proposed site to
assess its potential environmental, economic and social impacts on other locations.
This situation is analogous to the need for monitoring existing operations in terms
of environmental and/or other impacts. As noted by Osleeb and Kahn (1998), these
decision support needs cannot be effectively addressed without the use of GIS.
Finally, Kapetsky and Travaglia (1995) also point out that the individual investor
interested in aquaculture development also requires spatial information particularly
at the time of site selection from among a range of alternative locations with
different biophysical and socio-economic characteristics. GIS is potentially a powerful tool for assisting this class of decision-makers, and is already being effectively
used for such purposes in some places (Carswell, 1998; LUCO, 1998; Arnold et al.,
2000) where advanced capabilities in terms of infrastructure and trained personnel
exists. In general, however, increased deployment of GIS for practical decisionmaking in aquaculture is hampered by several constraints including: (1) a lack of
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
235
appreciation of the benefits of such systems on the part of key decision-makers; (2)
limited understanding about GIS principles and associated methodology; (3) inadequate administrative support to ensure GIS continuity among organizations; and
(4) poor levels of interaction among GIS analysts, subject matter specialists and end
users of the technology (see also Kapetsky and Travaglia, 1995).
The primary goals of this overview are to examine the above constraints in depth,
and to supplement previous reviews of GIS applications in aquaculture (Meaden
and Kapetsky, 1991; Kapetsky and Travaglia, 1995; Aguilar-Manjarrez, 1996;
Ross, 1998). Our overview includes an introduction to GIS terminology and
methodology, topics that are relatively well documented in the literature (Burrough,
1986), but to which personnel in aquaculture have had limited exposure, especially
from the perspective of GIS as a tool for practical decision-making. A summary of
selected cases where such tools have been applied in aquaculture is also provided.
Finally, a section of the paper focuses on the future of GIS in aquaculture, and
recommendations for its continued use.
2. GIS terminology
GIS is an integrated assembly of computer hardware, software, geographic data
and personnel designed to efficiently acquire, store, manipulate, retrieve, analyze,
display and report all forms of geographically referenced information geared
towards a particular set of purposes (Burrough, 1986; Kapetsky and Travaglia,
1995). An excellent introduction to GIS terminology, maintained by the Association of Geographic Information, is available on the World Wide Web (WWW) at
http://www.geo.ed.ac.uk/root/agidict/html/welcome.html.
There are essentially two types of GIS — vector and raster systems. These
systems differ in the manner by which spatial data are represented and stored.
Burrough (1986) and Meaden and Kapetsky (1991) provide useful comparisons of
vector and raster systems in summary form. In both systems, a ‘geographic
coordinate system’ is used to represent space. Many coordinate systems have been
defined, ranging from simple Cartesian X-Y grids to spatial representations that
correspond to the real world such as latitude/longitude pairings, State Plane
coordinates or the Universal Transverse Mercator system coordinate system (Burrough, 1986; Thompson, 1998).
2.1. Vector GIS
In vector GIS, spatial data are defined and represented as ‘points’, ‘lines’ or
‘polygons’ (Fig. 1). A point is defined by a single set of coordinates. Examples of
point data for aquaculture applications may include features of a farm such as
pump locations, supply wells and small buildings like storage sheds. Features such
as roads, rivers and canals are conveniently represented as lines (Fig. 1). Lines have
starting and ending points referred to as ‘nodes’, which may include a large number
of ‘vertices’ (Fig. 1). The segment of a line between two vertices is referred to as an
236
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
‘arc’. Polygons are used in GIS to represent enclosed areas (Fig. 1). A polygon
consists of a number of lines, but is distinguished by the characteristic that its
starting and ending nodes are the same. For aquaculture applications, examples of
polygons include a reservoir, a certain soil class, and a distinct type of land
classification (e.g. a mangrove forest). Based on the coordinate system used, a
vector GIS knows where the spatial feature (point, line or polygon) exists (i.e. its
absolute location), and its relationship to other features (topology or relative
location). As an example, a GIS would thus be aware that a reservoir supplying
water to a fish farm is located north of the farm ponds.
After spatial features are represented in vector GIS, their associated properties
can be specified in a separate database. As an example, properties of a soil polygon
such as pH, bulk density, cation exchange capacity, and infiltration rate can be
archived in a database. These properties are often referred to as ‘themes’, which can
be presented in so called ‘thematic maps’. Vector GIS lend themselves well to the
use of relational databases because once the spatial features are specified (once),
any amount of related data can be associated with them. The term ‘coverage’ is
used in the GIS literature to refer to the combination of a geographically referenced
feature and its associated data. Thus, a ‘stream coverage’ would refer to the line in
a vector GIS that represents its course, and the associated data such as flow rates,
temperature, water quality characteristics, and withdrawal rates (and possibly
spatial locations) for different uses.
2.2. Raster GIS
In a raster GIS, space is represented by a uniform grid, each cell of which is
assigned a unique descriptor depending on the coordinate system used (Fig. 2).
Fig. 1. Representation of spatial features (points, lines and polygons) in a vector GIS (adapted from
Thompson, 1998).
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
237
Fig. 2. Representation of spatial features (points, lines and polygons) in a raster GIS (adapted from
Thompson, 1998).
Thus, a cell in a grid that uses the latitude/longitude system would have a pair of
coordinates. Numerical data pertaining to spatial features that are represented in
the grid are assigned to the appropriate cell. Raster GIS can be conveniently
thought of as being a spreadsheet, and one of the earliest GIS applications in
aquaculture was essentially a spreadsheet-based system to assess carp culture
possibilities in Pakistan (Ali et al., 1991).
In raster systems, georeferenced data stored in a grid constitute a ‘layer’, as
opposed to a coverage in vector systems. Each grid contains a unique set of
information. Thus, each of the soil properties described above for vector GIS would
be stored in a separate layer if a raster system is used. In other words, there would
be as many layers as there are properties of the spatial feature. Consequently, data
storage requirements for raster GIS are generally higher than comparable vector
systems. Further, if there is a need to integrate different layers that describe a
certain feature, the process may also be more costly in terms of access time. These
disadvantages are perhaps less important in present day computers because large
hard disks are not only relatively inexpensive, but also allow rapid data retrieval. A
more important disadvantage of raster GIS is that the spatial resolution of analyses
is limited to the cell size of the coarsest layer. When the cell size is large relative to
the scale of analysis, maps created from raster data tend to be somewhat blotchy.
Nevertheless, raster systems are particularly useful when it comes to numerical
manipulations because of the uniform grid structure. For instance, they lend
themselves well to spatial modeling applications. As an example, a water temperature simulation model can be executed with weather datasets (e.g. air temperature,
solar radiation, cloud cover, relative humidity, and wind speed) in the form of
raster inputs. Model output can be exported to a gridded file, which in turn can be
238
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
imported into raster systems for further classification and display (Kapetsky and
Nath, 1997). Similarly, raster GIS can be used to model effluents discharged by
aquaculture operations into natural water bodies by applying appropriate equations
to cells in the spatial grid (Perez-Martinez, 1997). Another advantage of raster GIS
is that remote sensed data (which are usually stored in raster format) can often be
directly imported into the software and immediately become available for use
(Burrough, 1986).
2.3. Analytical scope, reporting and 6isualization
One of the powerful features of both vector and raster GIS packages is that
statistical summaries of layers/coverages, model stages or outcomes can easily be
obtained. Statistical data can include area, perimeter and other quantitative estimates, including reports of variance and comparison among images. A further
powerful analytical tool that aids understanding of outcomes is visualization of
outcomes through graphical representation in the form of 2D and 3D maps. For
example, entire landscapes and watersheds can be viewed in three dimensions,
which is very valuable in terms of evaluating spatial impacts of alternative decisions. Techniques have also been developed to integrate GIS with additional tools
such as group support systems, that allow interactive scenario development and
evaluation, and support communication among stakeholders via a local area
network (LAN; Faber et al., 1997). There is also currently rapid development and
deployment of Internet-enabled GIS tools, which allow a wider community of
decision makers to have instant access to spatial data. All of these tools are
constantly being added to GIS packages and are of great value if appropriately
used. Presently available tools include Arc/Explorer and Internet Map Server (from
ESRI, Inc) and MapXtreme (from MapInfo Corporation). Because GIS technology
is evolving at a very rapid rate, we have chosen not to evaluate specific products.
Information about GIS and related tools are, however, available from various Web
sites including: http://www.utexas.edu/depts/grg/virtdept/resources/vendors/vendors.htm and http://www.gislinx.com/. Guidelines for selecting GIS tools are
available in: Burrough (1986), Meaden and Kapetsky (1991) and Burrough and
McDonnell (1998).
3. GIS methodology
Ideally, any GIS study will consist of seven phases: identifying project requirements, formulating specifications, developing the analytical framework, locating
data sources, organizing and manipulating data for input, analyzing data and
verifying outcomes, and evaluating outputs (Fig. 3). In practice, it may be necessary
to iterate within the overall process, particularly among the first four phases. The
figure also indicates the degree to which different types of project personnel namely
end users, subject matter experts, and GIS analysts can be expected to be involved.
The degree of personnel involvement within each phase will vary according to the
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
239
organization(s) and/or types of personnel, as well as according to the requirements
of specific projects.
With regard to the first phase of any GIS project (basically a conceptualization
and planning step that precedes actual implementation), our opinion is that it has
traditionally been somewhat neglected both in the GIS literature as well as in
practice. This is true despite the fact that it will likely determine the extent to which
information generated by the use of GIS is used in real world decision making. The
latter criterion, of course, is the ultimate yardstick by which success of spatial
methods and technology will be measured over the long-term.
The phases involved in any GIS study (Fig. 3; described in detail below) occur
iteratively in the sense that project personnel may often conduct a pilot-scale study
Fig. 3. Schematic representation of the phases in a GIS project. In practice, most of the iteration within
the overall process is to be found within the first four phases. Involvement of end users (), subject
matter specialists ( ), and GIS analysts (
) within each of the phases is also indicated, with symbol
sizes reflecting the importance of their respective roles.
240
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
with available information, and then successively enhance and/or refine the analysis
until a satisfactory end point is reached. Alternately, subsequent phases in the
project cycle may result in new information that needs to be incorporated in the
preceding phase(s). For example, it may become evident during the development of
the analytical framework that requirements documented during the first iteration of
the project cycle (Fig. 3) may need to be revised.
Recognizing the iterative nature of a GIS study allows the entire project team to
develop an improved understanding of the issue on hand, and how spatial methods
and technology can be best be used to address it. Once project personnel explicitly
recognize that GIS work is iterative in nature and begin to document what is learnt
during each phase, it becomes easier to track how user requirements are explicitly
addressed during implementation (often referred to as traceability in the software
literature; Ambler, 1998), and to deliver timely as well as meaningful project
updates (e.g. GIS outputs and reports).
3.1. Identifying project requirements
The process of identifying requirements for a GIS is essentially a multiple
stakeholder decision-making situation. This is because such work is invariably
executed by a group of subject matter experts and analysts, and because results of
the analyses are potentially useful to a range of decision-makers. Individuals in
these three broad groups of stakeholders are likely to have different perspectives
and expectations with regard to the capabilities and limitations of GIS. It is
important that these perspectives and expectations are allowed to surface during the
early stages of GIS work so that an enabling environment can be created within
which decision support needs of end users are discussed and project goals
formulated.
Once project personnel, particularly end users, have had an opportunity to
present their spatial decision support needs, discussions can begin to examine how
GIS tools can address these needs, and clarify the limitations of such tools (e.g.
spatial data availability and quality, software/hardware resources that may be
needed, cost and time constraints, etc.). At this stage, it is usually not necessary or
even productive to invest an inordinate amount of time in discussing these issues.
Instead, the intent is to develop common understanding about decision support
needs, specify project goals, clarify GIS functionality and develop a listing of
general requirements. More importantly, such discussions can facilitate the development of a creative and productive project team, wherein each participant has an
opportunity to identify their own roles and responsibilities, and those of their
partners.
The above observations are consistent with the work by Campbell (1994) (see
also Campbell and Masser, 1995), who suggests that the following factors are
important in successful GIS implementation and its subsequent use in decisionmaking:
an information management strategy in which the needs of users are clarified,
and resources and values of the organization(s) are accounted for;
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
241
identification of simple applications that generate information fundamental to
the work of potential users;
implementation that is user-directed, and which has the commitment and participation of staff throughout the organization(s) involved; and
the ability for the organization(s) to create sufficient stability or to be able to
cope with change, both in terms of evolving specifications and rapidly changing
technology.
In our own experience, there has been a tendency among some of the technical
assistance organizations with which we collaborate to invest a considerable amount
of time, effort and financial resources in developing GIS capabilities and acquiring
expensive spatial datasets. However, it is often the case that this investment is made
without a clear understanding about what they wish to accomplish with such
capabilities, and the stakeholder groups (both within and outside the organization)
whose decision support needs they are attempting to fulfil. More often than not,
GIS capabilities of such organizations are primarily used as a tool for generating
and displaying maps. The current state of spatial methods and technology, on the
other hand, clearly indicates that GIS capabilities go beyond data management and
visualization alone.
For instance, process-based models and GIS (especially raster systems) are very
complementary tools in that the former can be used to generate new output layers
from multiple base maps already in GIS format. The derived layers often contain
information that is more useful to decision-makers. Moreover, it is now possible to
directly integrate within GIS models that have distributed (i.e. spatially explicit)
parameters, and to seamlessly run alternate scenarios within a single software
package. Prior to the availability of such techniques, it was necessary to execute
multiple model runs outside of a GIS environment, and to import the output into
GIS for visualization. Such indirect coupling often requires additional time and
technical expertise to handle model management and data transfer requirements,
besides limiting the range of scenarios that end users may wish to analyze. Direct
integration of models within GIS, however, has the potential to enable such users
in the aquaculture domain to explore alternate scenarios in a truly interactive way,
simultaneously reducing the need for the continued presence of analysts or at least
using their expertise in a manner that is more cost effective. The trend towards
interactive GIS use is already evident in both in the business world and in some
natural resource management domains such as conservation efforts.
Another observation to be made in the current context is that if the target
decision support audience is not adequately consulted to assess their needs, the
possibility exists that considerable effort may be expended by organizations to
generate information that is rarely used by the intended audience. The fundamental
message here is that if end users are involved early on in the planning process for
GIS projects and their needs better understood, experts and analysts are more likely
to generate strategies that better address these needs. Moreover, they are also more
likely to make optimal use of spatial methods and technology within any organizational constraints that may exist. It should be noted that our observations about
adopting a stakeholder-driven approach to project design and management, and
242
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
specific techniques to manage the process are not unique to GIS applications, but
are increasingly recognized within a number of domains including organizational
development (Adams, 1986; Senge, 1990), concurrent engineering design (Voland,
1992), and information technology (Ambler, 1998).
3.2. Formulating specifications
The goals and requirements of the project articulated in the first phase above are
by necessity of a general nature. This is largely because it is rarely productive to
embark upon an exercise to specify project components without a broad understanding of decision support needs. Furthermore, if team members begin to focus
too much on project specifics early on in the overall process, they run the risk of
not being able to adequately accommodate additional needs of end users (given that
such needs usually tend to evolve over time), and of already confining themselves to
a certain mindset about implementation strategies.
However, once an overall understanding of project requirements has been
developed among team members, it is helpful to develop a listing of more
functional specifications corresponding to each of the requirements that have been
identified. For instance, if the project requires that the final GIS be interactive
(implying that the end user can explore alternate scenarios on their own), one or
more of the following are representative of possible functional specifications:
capability of generating and editing different thematic maps;
features for querying attributes in spatial databases;
support for adding new themes and/or updating current information in the GIS
over time;
enabling application of the analytical approach to different geographical regions;
and
capability to modify existing models and/or link new models to the GIS.
The process of identifying functional specifications involves an in-depth analysis
of each of the project requirements. In practice, it is usually beneficial for the
subject matter experts and GIS analysts to first formulate project specifications
jointly, and to then share these with the end users. This approach tends to result in
time (and therefore cost) savings. Moreover, the need for communicating (potentially) difficult technical concepts is minimized because end users are likely to be
more interested in the actual specifications, rather than in the process by which they
were determined.
The benefits of formulating specifications for a GIS project far outweigh the
initial investment in terms of resource and time commitments. Some of the benefits
include an increased probability of implementing a system that does indeed meet
end user needs, laying the foundation for a work breakdown structure with
associated timelines and responsibilities (for different project personnel), and an
improved likelihood of building a GIS with the appropriate types of hardware and
software tools. Experience from the software industry also suggests that when
project specifications are formally identified, there is an improved probability of
minimizing cost over-runs and time delays. On the other hand, as noted by the
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
243
same author, it is safe to assume that no project can ever be fully specified
(particularly during the initial phases of development) and a judicious end point for
this phase must be identified to ensure that timely progress towards the overall
goals is maintained. It is also often the case that when functional specifications are
identified, project personnel may identify new requirements, which need to be
appropriately documented with existing information from the first phase of the
overall project (i.e. identifying project requirements; Fig. 3).
3.3. De6eloping the analytical framework
The previous two phases deal primarily with aspects of what is to be accomplished (i.e. project needs). Development of the analytical framework for a GIS
project, on the other hand, addresses issues of how these needs will be addressed.
This phase largely involves subject matter specialists and GIS analysts (Fig. 3).
However, consultations with end users are recommended to ensure that the project
will indeed address their needs, to accommodate new needs that may arise, and to
foster an improved understanding of analytical methods that may be used. An
understanding of the methods used, as well as their limitations, is critical in terms
of appropriate application of GIS outcomes for decision support.
Several methods have been used, either singly or in combination, by GIS
practitioners to integrate spatial information into a useful format for analysis and
decision making. The following represent analytical methods that have already been
used in aquaculture GIS or have potential for use in the future. Their inclusion here
is intended to provide readers with an overview of the range of analytical methods
that may come up for discussion during this phase of a GIS project.
3.3.1. Arithmetic operators
A large number of arithmetic operators can be used in GIS (Burrough, 1986).
For instance, a useful precursor is the scalar operation in which a constant term is
to be applied to spatial data (e.g. to estimate the potential annual consumption of
fish in a region, the population size of towns can be multiplied by a constant
representing the annual fish consumption per capita). Another example would be
the subtraction of monthly evapotranspiration and seepage layers from a precipitation layer to estimate mean monthly water requirements for ponds (Aguilar-Manjarrez and Nath, 1998).
3.3.2. Classification
It is almost always the case that the source data, whether in real or integer
format, will need to be further classified before further use. Classification is an
essential part of any data reduction process, whereby complex sets of observations
are made understandable. Although any classification process involves some loss of
information, a good scheme not only aims to minimize this loss, but by identifying
natural groups that have common properties, provides a convenient means of
information handling and transfer (Burrough, 1986). Further, in any classification
process, care must be taken to preserve the appropriate level of detail needed for
244
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
sensible decision making at a later stage (Burrough, 1986; Aguilar-Manjarrez, 1996;
Ross, 1998).
Aguilar-Manjarrez (1996) provides an exhaustive review of five methods that
have been explored to classify data on land types for various uses:
1. The FAO land evaluation methodology which assesses land suitability in terms
of an attribute set corresponding to different activities.
2. The limitation method in which each land characteristic is evaluated on a
relative scale of limitations.
3. The parametric method in which limitation levels for each characteristic are
rated on a scale of 0 to 1, from which a land index (%) is calculated as the
product of the individual rating values of all characteristics.
4. The Boolean method which assumes that all questions related to land use
suitability can be answered in a binary fashion, and that all important changes
occur at a defined class boundary.
5. The fuzzy set method in which an explicit weight is used to assess the impact of
each land characteristic. Fuzzy techniques are then used to combine the evaluation of each land characteristic into a final suitability index. Apart from a
dominant suitability class, the fuzzy set method equally provides information on
the extent to which a certain land unit belongs to each of the suitability classes
discerned.
For GIS applications, any of the above methods can be used to classify source
data into a four- or five-point scale of suitability (with one being the least suitable).
However, the choice among classification methods is dependent on the type of data
and intended uses of the output information. Classification allows normalization of
all data layers, an essential pre-requisite for further modeling. An example is the
grouping of soil data into four classes based on a number of properties important
for pond construction (Kapetsky and Nath, 1997).
3.3.3. Simple o6erlay
The most common technique in geographic information processing is that of
topological overlaying in which multiple data layers are overlain in a vertical
manner. Topological overlays may be broadly classified into simple and weighted
(see below) methods. After reclassification of thematic maps in terms of suitability
for an activity has been accomplished, simple overlay can be accomplished by
applying mathematical or logical operators to the layers. This can often generate
very valuable outcomes. Simple overlays operations assume that all layers have
equal importance. Outcomes of the operation can be further refined if they are
overlain with constraint layers by subtraction. A more common practice, however,
is to represent constraint layers in Boolean format, with disallowed and allowed
areas defined as 0 and 1, respectively. In this way, when the constraint layer is
multiplied by one containing suitability data, the former is automatically excluded
from the outcome. A series of such simple reclassifications and overlays can be
assembled as a tree of sequential choices (i.e. a decision tree) to provide a rational
outcome to a spatial problem. Examples of constraint layers are large water bodies
and forested lands that are unavailable for aquaculture (Kapetsky and Nath, 1997).
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
245
3.3.4. Weighted o6erlay
In decision making, it is usually the case that different attributes under consideration do not have the same level of importance. This calls for a weighted overlay
approach, in which each layer is assigned a weight that is proportional to its
importance. To accomplish this, each source layer is first reclassified onto a
common scale and then multiplied by a weighting factor. The resulting values are
used in further overlay operations to obtain an outcome (applying Boolean
constraints in the final stage if needed).
The derivation of weightings for source attributes is often assumed to be an
objective scientific matter. However, it is clear from a range of studies that even
expert opinions on ranking of attributes may differ substantially. Aguilar-Manjarrez (1996) showed that, from a list of attributes, a group of experts with similar
backgrounds generally agree upon which are the most relevant to use for any given
decision process. However, the ranking assigned to the attributes by these experts
can differ. In general, Aguilar-Manjarrez (1996) showed that from a list of ten
selected attributes, different experts would be in agreement on the priority of the
first three to four and the last three. However, attributes with mid-level priority are
often dealt with quite differently. Moreover, subject matter experts with different
backgrounds (e.g. aquaculturist, coastal zone planner, conservationist, etc.) bring
differing agendas to the same problem, resulting in a range of outcomes. At worst,
it is clear that without guidance, a range of prioritizations can be obtained which
are cumulatively meaningless. GIS analysts therefore need to ensure that weightings
applied personally are as objective as possible and that where expert group input is
used, the basis for this input is fully explained and carefully analyzed before its use
for decision making. Summary tables that transparently document factors used in
the evaluation and the weights assigned to them by different experts are a useful aid
in this regard (Aguilar-Manjarrez, 1996; Aguilar-Manjarrez and Nath, 1998).
As the number of layers in a model increases, the logic of processing them can
become confusing. Consequently, special routines to assist with this process have
been provided in certain GIS packages. For example, the decision support module
in IDRISI (Eastman, 1993) supports multi-criteria evaluation (MCE), which is
based on the analytical hierarchy process (AHP) described by Saaty (1980). This
module enables a structured, logical approach to weighted overlay with built-in
crosschecks at some stages of the weighting process to ensure consistency of logic
by the operator. AHP falls into the broader category of pairwise comparison
techniques in which attributes are ranked against each other to assess their relative
importance. Although AHP or other pairwise comparison variants do provide
opportunity for increased objectivity in the ranking of factors, they do not
adequately address the issue of different weightings that may be assigned by diverse
subject matter experts. For instance, different weightings were assigned to factors
important for pond aquaculture by different experts in continental-scale GIS
studies (Kapetsky and Nath, 1997; Aguilar-Manjarrez and Nath, 1998). The AHP
process was useful in these studies in terms of providing a logical framework for
comparing the factors. However, in both instances, the authors ultimately relied on
their own ranking schemes (which had similar patterns to those of the experts) to
246
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
assign weights to the factors. In general, it is unlikely that technical methods such
as AHP or other multi-criteria decision making tools (see review by Merkhofer,
1998) can be a complete replacement to a well-facilitated exercise in which
stakeholders (i.e. subject matter specialists, GIS analysts and end users in the
current context) are encouraged to seek consensus on the relative importance of
attributes for use in GIS. Instead, such tools should be viewed as providing
additional support to the consensus building process.
3.3.5. Neighborhood analysis
A key function provided by GIS, and one which cannot be easily addressed by
the use of any other decision support tool, is its capability to allow evaluation of
the characteristics of an area that surrounds a specific location, which is referred to
as neighborhood analysis. GIS software provides a range of neighborhood functions including point interpolation techniques in which unknown values are estimated from known values of neighboring locations. These techniques are typically
used to convert point coverages into grids or to generate elevation data either in
triangular irregular network (TIN) format for vector GIS or as a digital elevation
model (DEM) for raster GIS. DEMs have not as yet found widespread use in
aquaculture, but are often at the heart of environmental GIS that are targeted
towards analyses of entire landscapes. For instance, the DEM for a watershed of
interest is a key layer in many spatially explicit hydrological models. In the future,
such tools are likely to find applications in aquaculture from the perspectives of
assessing water availability and modeling environmental impacts of existing and/or
proposed aquaculture operations. A common type of neighborhood operation is
buffering, which allows the creation of distance buffers (or buffer zones) around
selected features. An example where this operation may be useful in aquaculture is
when there is a need to determine how many streams are available (or to be
avoided) within a specified distance of a proposed facility.
3.3.6. Connecti6ity analysis
This type of analysis is characterized by an accumulation of values over the area
that is being traversed. Support for connectivity analysis in commercially available
GIS software is not as standardized in comparison to the operators described
above. Network analysis is one type of connectivity operation, which is characterized by the use of feature networks such as hydrographic hierarchies (i.e. stream
networks) and transportation networks. As an example of the use of network
analysis in aquaculture, Kapetsky and Nath (1997) estimated the shortest transportation path from a given site to the nearest market.
The most sophisticated type of connectivity operation is three dimensional
analysis which is helpful in terms of visualizing impacts of alternate decisions on
entire landscapes (e.g. watersheds). Such advanced features of GIS are likely to
have applications in aquaculture but have apparently yet to be used in the systems
that have been developed in this domain.
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
247
3.3.7. Hierarchical models
In practice, comprehensive representation of the real world in a GIS often
involves the use of a large number of source variables, which when processed into
GIS using one or more of the analytical methods described above, may result in
considerable complexity. Experience suggests that when the number of layers
exceeds about 10, MCE becomes difficult, even to the experienced modeler. Such
situations warrant the development of a hierarchical modeling scheme.
In this approach, naturally grouped variables are first considered together to
produce ‘sub-model’ outcomes such as water needs, soil suitability, input availability, farm gate sales, and markets (Kapetsky and Nath, 1997). It is often the case
that a source variable or processed layer will be used in more than one sub-model
and that the layer may need to be transformed depending on the intended purpose.
Each of these sub-models may, in turn, be derived from lower-level models which
pre-process variable data into useful factors. Another example from aquaculture is
the estimation of wave height from wind velocity and fetch (Ross et al., 1993).
It is, of course, not necessary to embed all of the models and/or sub-models
within a single system per se. This is because commercial GIS can export and
import information quite seamlessly with external modules. For instance, growth
estimates derived from an external bioenergetic model (linked to spatial weather
and water temperature datasets) have been used to predict fish yields in continentalscale GIS (Kapetsky and Nath, 1997; Aguilar-Manjarrez and Nath, 1998). In a
different context, external rule-based systems are being used in conjunction with
GIS to generate future agricultural land use patterns for assessing policy alternatives in the Pacific Northwest region of the United States (Bolte, unpublished
information).
3.3.8. Multi-objecti6e land allocation
Once an activity has been modeled and quantified, it will invariably have some
potential to conflict with other uses of the space or resources. For example,
practices such as agriculture, forestry, and aquaculture may compete for available
land and associated resources. This calls for trade-off decisions to be made so that
activities can coexist. These decisions typically require consideration of economic,
environmental and social ramifications of alternative land use practices. Decision
support tools are available in some GIS packages to facilitate this process. The
multi-objective land allocation (MOLA) and multi-dimensional decision space
(MDCHOICE) tools in IDRISI are good examples. MOLA was used by AguilarManjarrez and Ross (1995) to identify areas suitable for agriculture and aquaculture across the Sinaloa state of Mexico (see section on case studies below). Inputs
for the MOLA technique originate from the results of weighted overlays. The
approach allows GIS analysts to set limits on areas required for different land uses
and assign requirements for these uses. An iterative process is then used to
successively reassign ranked cells to alternate activities depending upon how closely
they match the associated requirements.
248
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
3.4. Identifying data sources
Once the analytical framework has been developed, it is necessary to identify
data sources to be used in the project. This phase is largely restricted to GIS
analysts (Fig. 3), although subject matter specialists often provide helpful advice
(e.g. identifying non-spatial datasets). Information for spatial decision-making and
analysis is varied, and will usually consist of data describing the biophysical,
economic, social and infrastructural environments. These data can come from a
variety of sources ranging from primary data gathered in the field or satellite scenes
to all forms of secondary data, including textual databases and reports.
It is generally both costly and time consuming to collect field (primary) data first
hand. Therefore, all GIS practitioners attempt to locate the data they need from
existing secondary sources, either in paper or digital form. The initial consideration
is identifying what data are needed for the overall analysis. This is followed by
attempts to source the data, and to assess their age, scale, quality and relative cost.
Different data types are often developed in different geographic coordinate systems,
and must be re-projected onto a common coordinate system. Other common issues
are ensuring that features common across multiple layers are spatially coincident,
and understanding the resolution, constraints and uncertainty associated with the
data. For these reasons, data collection and preprocessing are typically the most
expensive and time-consuming component of an analysis.
Digital database availability is increasing at a rapid rate throughout the world.
These databases contain information ranging from natural resources (e.g. maps of
soils, water resources and temperature distributions), to population census and
cadastral (i.e. property ownership and associated boundaries) data. In many cases,
such data may only be available in hard copy reports, although information is also
available on CD-ROMs, from which databases for GIS use can be extracted.
A further source of data is the WWW, from which a wide range of mapped and
other spatial data can be found by carefully executed searches. Spatial information
at low resolution can often be obtained in this way, good examples being the
Global 30 arc second elevation (http://edcwww.cr.usgs.gov/landdaac/gtopo30/
gtopo30.html), and the 1-km Global Land Cover datasets. Other online datasets
include information on drainage basins, and weather, which can be downloaded
from Web sites at some universities and international agencies. In this regard, a
useful site that provides metadata search capabilities and access to a range of data
resources is the one maintained by the Center for International Earth Science
Information Network (CIESIN) at http://www.ciesin.org. In some countries such as
the USA, GIS clearinghouses have been established on the Web and offer access to
consolidated datasets from multiple agencies (see http://www.usgs.gov). Many of
these free spatial databases are of immense value and should not be overlooked.
The ‘lingua franca’ of GIS is the thematic map in which particular attributes of
the geographical region are represented digitally or on a specially prepared hard
copy map. Clearly, where such maps are available in digital form, they are directly
usable in a GIS. Hard copy maps can also be digitized, and the information
imported into GIS. The digitizing process, however, is very time consuming and
requires many hours of data editing to ensure superior quality.
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
249
Thematic data can be in the form of a choropleth (areas of equal value separated
by boundaries), isopleth (lines which connect points of equal value) or point maps
(Burrough, 1986), all of which are often valuable, a good example in terms of
aquaculture being the hydrographic chart. The more common topographic map
frequently contains many thematic data (including elevations, water bodies, roads,
cities, and woodlands) which are of value to the GIS analyst. These themes can be
extracted at the stage of digitization, and established as separate layers in the
spatial database.
3.5. Organizing and manipulating data
After datasets have been collected, it is necessary to organize and manipulate
them for use in the target GIS. This phase is also largely restricted to GIS analysts
(Fig. 3), although depending on the type of application, occasional interaction with
subject matter specialists may be warranted. Some of the key activities that occur
in this phase include verification of data quality, data consolidation and reformatting, creation of proxy data and database construction. Proxy data refer to
information that is derived from another data source, for which established
relationships may exist. Examples include estimation of water temperature from air
temperature (Kapetsky, 1994), extraction of semi-quantitative texture from FAO
soil distribution maps (Kapetsky and Nath, 1997) or calculation of maximum wave
heights from wind speed and fetch (Ross et al., 1993).
In terms of verification of data quality, the reliability of thematic maps worldwide is variable, as is the currency of their content. These aspects must be
considered where critical decisions are based upon such material. Although digital
maps are often quite up to date, the spatial accuracy of some printed material can
be suspect. Critical assessment and verification of source data quality is very
important. The value of such verification for all input data cannot be over-stressed,
before digitizing as well as after the fact. It is usually the case that at least some of
the layers required in a GIS database will not be of a high enough standard, and
verification on paper may very well need to become verification by survey, where
warranted. A detailed overview of technical methods to address data quality issues
is available in Burrough (1986).
Certain data sources such as satellite data are already in digital form, but all
others may requir
www.elsevier.nl/locate/aqua-online
Applications of geographical information
systems (GIS) for spatial decision support in
aquaculture
Shree S. Nath a,*, John P. Bolte b, Lindsay G. Ross c,
Jose Aguilar-Manjarrez d
a
Skillings-Connolly, Inc., 5016 Lacey Boule6ard S.E., Lacey, WA 98503, USA
Department of Bioresource Engineering, Oregon State Uni6ersity, Cor6allis, OR 97331, USA
c
Institute of Aquaculture, Uni6ersity of Stirling, Stirling FK9 4LA, UK
d
Food and Agricultural Organisation of the U.N, Viale delle Termi di Caracalla, 00100 Rome, Italy
b
Received 1 September 1999; accepted 2 October 1999
Abstract
Geographical information systems (GIS) are becoming an increasingly integral component
of natural resource management activities worldwide. However, despite some indication that
these tools are receiving attention within the aquaculture community, their deployment for
spatial decision support in this domain continues to be very slow. This situation is
attributable to a number of constraints including a lack of appreciation of the technology,
limited understanding of GIS principles and associated methodology, and inadequate
organizational commitment to ensure continuity of these spatial decision support tools. This
paper analyzes these constraints in depth, and includes reviews of basic GIS terminology,
methodology, case studies in aquaculture and future trends. The section on GIS terminology
addresses the two fundamental types of GIS (raster and vector), and discusses aspects related
to the visualization of outcomes. With regard to GIS methodology, the argument is made for
close involvement of end users, subject matter specialists and analysts in all projects. A
user-driven framework, which involves seven phases, to support this process is presented
together with details of the degree of involvement of each category of personnel, associated
activities and analytical procedures. The section on case studies reviews in considerable detail
four aquaculture applications which are demonstrative of the extent to which GIS can be
deployed, indicate the range in complexity of analytical methods used, provide insight into
* Corresponding author. Tel.: + 1-360-4913399; fax: +1-360-4913857.
E-mail address: snath@skillings.com (S.S. Nath)
0144-8609/00/$ - see front matter © 2000 Elsevier Science B.V. All rights reserved.
PII: S 0 1 4 4 - 8 6 0 9 ( 0 0 ) 0 0 0 5 1 - 0
234
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
issues associated with data procurement and handling, and demonstrate the diversity of GIS
packages that are available. Finally, the section on the future of GIS examines the direction
in which the technology is moving, emerging trends with regard to analytical methods, and
challenges that need to be addressed if GIS is to realize its full potential as a spatial decision
support tool for aquaculture. © 223 Elsevier Science B.V. All rights reserved.
Keywords: GIS; Aquaculture; Spacial decision support
1. Introduction
The rapid growth of aquaculture worldwide has stimulated considerable interest
among international technical assistance organizations and national-level governmental agencies in countries where fish culture is still in its infancy, and has resulted
in increased concerns about its sustainability in countries where the industry is well
established. Planning activities to promote and monitor the growth of aquaculture
in individual countries (or larger regions) inherently have a spatial component
because of the differences among biophysical and socio-economic characteristics
from location to location. Biophysical characteristics may include criteria pertinent
to water quality (e.g. temperature, dissolved oxygen, alkalinity/salinity, turbidity,
and pollutant concentrations), water quantity (e.g. volume and seasonal profiles of
availability), soil type (e.g. slope, structural suitability, water retention capacity and
chemical nature) and climate (e.g. rainfall distribution, air temperature, wind speed
and relative humidity). Socio-economic characteristics that may be considered in
aquaculture development include administrative regulations, competing resource
uses, market conditions (e.g. demand for fishery products and accessibility to
markets), infrastucture support, and availability of technical expertise. The spatial
information needs for decision-makers who evaluate such biophysical and socioeconomic characteristics as part of aquaculture planning efforts can be well served
by geographical information systems (GIS; Kapetsky and Travaglia, 1995).
Moreover, it is often the case that governmental agencies involved with issuing
new aquaculture permits need to perform spatial analysis on a proposed site to
assess its potential environmental, economic and social impacts on other locations.
This situation is analogous to the need for monitoring existing operations in terms
of environmental and/or other impacts. As noted by Osleeb and Kahn (1998), these
decision support needs cannot be effectively addressed without the use of GIS.
Finally, Kapetsky and Travaglia (1995) also point out that the individual investor
interested in aquaculture development also requires spatial information particularly
at the time of site selection from among a range of alternative locations with
different biophysical and socio-economic characteristics. GIS is potentially a powerful tool for assisting this class of decision-makers, and is already being effectively
used for such purposes in some places (Carswell, 1998; LUCO, 1998; Arnold et al.,
2000) where advanced capabilities in terms of infrastructure and trained personnel
exists. In general, however, increased deployment of GIS for practical decisionmaking in aquaculture is hampered by several constraints including: (1) a lack of
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
235
appreciation of the benefits of such systems on the part of key decision-makers; (2)
limited understanding about GIS principles and associated methodology; (3) inadequate administrative support to ensure GIS continuity among organizations; and
(4) poor levels of interaction among GIS analysts, subject matter specialists and end
users of the technology (see also Kapetsky and Travaglia, 1995).
The primary goals of this overview are to examine the above constraints in depth,
and to supplement previous reviews of GIS applications in aquaculture (Meaden
and Kapetsky, 1991; Kapetsky and Travaglia, 1995; Aguilar-Manjarrez, 1996;
Ross, 1998). Our overview includes an introduction to GIS terminology and
methodology, topics that are relatively well documented in the literature (Burrough,
1986), but to which personnel in aquaculture have had limited exposure, especially
from the perspective of GIS as a tool for practical decision-making. A summary of
selected cases where such tools have been applied in aquaculture is also provided.
Finally, a section of the paper focuses on the future of GIS in aquaculture, and
recommendations for its continued use.
2. GIS terminology
GIS is an integrated assembly of computer hardware, software, geographic data
and personnel designed to efficiently acquire, store, manipulate, retrieve, analyze,
display and report all forms of geographically referenced information geared
towards a particular set of purposes (Burrough, 1986; Kapetsky and Travaglia,
1995). An excellent introduction to GIS terminology, maintained by the Association of Geographic Information, is available on the World Wide Web (WWW) at
http://www.geo.ed.ac.uk/root/agidict/html/welcome.html.
There are essentially two types of GIS — vector and raster systems. These
systems differ in the manner by which spatial data are represented and stored.
Burrough (1986) and Meaden and Kapetsky (1991) provide useful comparisons of
vector and raster systems in summary form. In both systems, a ‘geographic
coordinate system’ is used to represent space. Many coordinate systems have been
defined, ranging from simple Cartesian X-Y grids to spatial representations that
correspond to the real world such as latitude/longitude pairings, State Plane
coordinates or the Universal Transverse Mercator system coordinate system (Burrough, 1986; Thompson, 1998).
2.1. Vector GIS
In vector GIS, spatial data are defined and represented as ‘points’, ‘lines’ or
‘polygons’ (Fig. 1). A point is defined by a single set of coordinates. Examples of
point data for aquaculture applications may include features of a farm such as
pump locations, supply wells and small buildings like storage sheds. Features such
as roads, rivers and canals are conveniently represented as lines (Fig. 1). Lines have
starting and ending points referred to as ‘nodes’, which may include a large number
of ‘vertices’ (Fig. 1). The segment of a line between two vertices is referred to as an
236
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
‘arc’. Polygons are used in GIS to represent enclosed areas (Fig. 1). A polygon
consists of a number of lines, but is distinguished by the characteristic that its
starting and ending nodes are the same. For aquaculture applications, examples of
polygons include a reservoir, a certain soil class, and a distinct type of land
classification (e.g. a mangrove forest). Based on the coordinate system used, a
vector GIS knows where the spatial feature (point, line or polygon) exists (i.e. its
absolute location), and its relationship to other features (topology or relative
location). As an example, a GIS would thus be aware that a reservoir supplying
water to a fish farm is located north of the farm ponds.
After spatial features are represented in vector GIS, their associated properties
can be specified in a separate database. As an example, properties of a soil polygon
such as pH, bulk density, cation exchange capacity, and infiltration rate can be
archived in a database. These properties are often referred to as ‘themes’, which can
be presented in so called ‘thematic maps’. Vector GIS lend themselves well to the
use of relational databases because once the spatial features are specified (once),
any amount of related data can be associated with them. The term ‘coverage’ is
used in the GIS literature to refer to the combination of a geographically referenced
feature and its associated data. Thus, a ‘stream coverage’ would refer to the line in
a vector GIS that represents its course, and the associated data such as flow rates,
temperature, water quality characteristics, and withdrawal rates (and possibly
spatial locations) for different uses.
2.2. Raster GIS
In a raster GIS, space is represented by a uniform grid, each cell of which is
assigned a unique descriptor depending on the coordinate system used (Fig. 2).
Fig. 1. Representation of spatial features (points, lines and polygons) in a vector GIS (adapted from
Thompson, 1998).
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
237
Fig. 2. Representation of spatial features (points, lines and polygons) in a raster GIS (adapted from
Thompson, 1998).
Thus, a cell in a grid that uses the latitude/longitude system would have a pair of
coordinates. Numerical data pertaining to spatial features that are represented in
the grid are assigned to the appropriate cell. Raster GIS can be conveniently
thought of as being a spreadsheet, and one of the earliest GIS applications in
aquaculture was essentially a spreadsheet-based system to assess carp culture
possibilities in Pakistan (Ali et al., 1991).
In raster systems, georeferenced data stored in a grid constitute a ‘layer’, as
opposed to a coverage in vector systems. Each grid contains a unique set of
information. Thus, each of the soil properties described above for vector GIS would
be stored in a separate layer if a raster system is used. In other words, there would
be as many layers as there are properties of the spatial feature. Consequently, data
storage requirements for raster GIS are generally higher than comparable vector
systems. Further, if there is a need to integrate different layers that describe a
certain feature, the process may also be more costly in terms of access time. These
disadvantages are perhaps less important in present day computers because large
hard disks are not only relatively inexpensive, but also allow rapid data retrieval. A
more important disadvantage of raster GIS is that the spatial resolution of analyses
is limited to the cell size of the coarsest layer. When the cell size is large relative to
the scale of analysis, maps created from raster data tend to be somewhat blotchy.
Nevertheless, raster systems are particularly useful when it comes to numerical
manipulations because of the uniform grid structure. For instance, they lend
themselves well to spatial modeling applications. As an example, a water temperature simulation model can be executed with weather datasets (e.g. air temperature,
solar radiation, cloud cover, relative humidity, and wind speed) in the form of
raster inputs. Model output can be exported to a gridded file, which in turn can be
238
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
imported into raster systems for further classification and display (Kapetsky and
Nath, 1997). Similarly, raster GIS can be used to model effluents discharged by
aquaculture operations into natural water bodies by applying appropriate equations
to cells in the spatial grid (Perez-Martinez, 1997). Another advantage of raster GIS
is that remote sensed data (which are usually stored in raster format) can often be
directly imported into the software and immediately become available for use
(Burrough, 1986).
2.3. Analytical scope, reporting and 6isualization
One of the powerful features of both vector and raster GIS packages is that
statistical summaries of layers/coverages, model stages or outcomes can easily be
obtained. Statistical data can include area, perimeter and other quantitative estimates, including reports of variance and comparison among images. A further
powerful analytical tool that aids understanding of outcomes is visualization of
outcomes through graphical representation in the form of 2D and 3D maps. For
example, entire landscapes and watersheds can be viewed in three dimensions,
which is very valuable in terms of evaluating spatial impacts of alternative decisions. Techniques have also been developed to integrate GIS with additional tools
such as group support systems, that allow interactive scenario development and
evaluation, and support communication among stakeholders via a local area
network (LAN; Faber et al., 1997). There is also currently rapid development and
deployment of Internet-enabled GIS tools, which allow a wider community of
decision makers to have instant access to spatial data. All of these tools are
constantly being added to GIS packages and are of great value if appropriately
used. Presently available tools include Arc/Explorer and Internet Map Server (from
ESRI, Inc) and MapXtreme (from MapInfo Corporation). Because GIS technology
is evolving at a very rapid rate, we have chosen not to evaluate specific products.
Information about GIS and related tools are, however, available from various Web
sites including: http://www.utexas.edu/depts/grg/virtdept/resources/vendors/vendors.htm and http://www.gislinx.com/. Guidelines for selecting GIS tools are
available in: Burrough (1986), Meaden and Kapetsky (1991) and Burrough and
McDonnell (1998).
3. GIS methodology
Ideally, any GIS study will consist of seven phases: identifying project requirements, formulating specifications, developing the analytical framework, locating
data sources, organizing and manipulating data for input, analyzing data and
verifying outcomes, and evaluating outputs (Fig. 3). In practice, it may be necessary
to iterate within the overall process, particularly among the first four phases. The
figure also indicates the degree to which different types of project personnel namely
end users, subject matter experts, and GIS analysts can be expected to be involved.
The degree of personnel involvement within each phase will vary according to the
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
239
organization(s) and/or types of personnel, as well as according to the requirements
of specific projects.
With regard to the first phase of any GIS project (basically a conceptualization
and planning step that precedes actual implementation), our opinion is that it has
traditionally been somewhat neglected both in the GIS literature as well as in
practice. This is true despite the fact that it will likely determine the extent to which
information generated by the use of GIS is used in real world decision making. The
latter criterion, of course, is the ultimate yardstick by which success of spatial
methods and technology will be measured over the long-term.
The phases involved in any GIS study (Fig. 3; described in detail below) occur
iteratively in the sense that project personnel may often conduct a pilot-scale study
Fig. 3. Schematic representation of the phases in a GIS project. In practice, most of the iteration within
the overall process is to be found within the first four phases. Involvement of end users (), subject
matter specialists ( ), and GIS analysts (
) within each of the phases is also indicated, with symbol
sizes reflecting the importance of their respective roles.
240
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
with available information, and then successively enhance and/or refine the analysis
until a satisfactory end point is reached. Alternately, subsequent phases in the
project cycle may result in new information that needs to be incorporated in the
preceding phase(s). For example, it may become evident during the development of
the analytical framework that requirements documented during the first iteration of
the project cycle (Fig. 3) may need to be revised.
Recognizing the iterative nature of a GIS study allows the entire project team to
develop an improved understanding of the issue on hand, and how spatial methods
and technology can be best be used to address it. Once project personnel explicitly
recognize that GIS work is iterative in nature and begin to document what is learnt
during each phase, it becomes easier to track how user requirements are explicitly
addressed during implementation (often referred to as traceability in the software
literature; Ambler, 1998), and to deliver timely as well as meaningful project
updates (e.g. GIS outputs and reports).
3.1. Identifying project requirements
The process of identifying requirements for a GIS is essentially a multiple
stakeholder decision-making situation. This is because such work is invariably
executed by a group of subject matter experts and analysts, and because results of
the analyses are potentially useful to a range of decision-makers. Individuals in
these three broad groups of stakeholders are likely to have different perspectives
and expectations with regard to the capabilities and limitations of GIS. It is
important that these perspectives and expectations are allowed to surface during the
early stages of GIS work so that an enabling environment can be created within
which decision support needs of end users are discussed and project goals
formulated.
Once project personnel, particularly end users, have had an opportunity to
present their spatial decision support needs, discussions can begin to examine how
GIS tools can address these needs, and clarify the limitations of such tools (e.g.
spatial data availability and quality, software/hardware resources that may be
needed, cost and time constraints, etc.). At this stage, it is usually not necessary or
even productive to invest an inordinate amount of time in discussing these issues.
Instead, the intent is to develop common understanding about decision support
needs, specify project goals, clarify GIS functionality and develop a listing of
general requirements. More importantly, such discussions can facilitate the development of a creative and productive project team, wherein each participant has an
opportunity to identify their own roles and responsibilities, and those of their
partners.
The above observations are consistent with the work by Campbell (1994) (see
also Campbell and Masser, 1995), who suggests that the following factors are
important in successful GIS implementation and its subsequent use in decisionmaking:
an information management strategy in which the needs of users are clarified,
and resources and values of the organization(s) are accounted for;
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
241
identification of simple applications that generate information fundamental to
the work of potential users;
implementation that is user-directed, and which has the commitment and participation of staff throughout the organization(s) involved; and
the ability for the organization(s) to create sufficient stability or to be able to
cope with change, both in terms of evolving specifications and rapidly changing
technology.
In our own experience, there has been a tendency among some of the technical
assistance organizations with which we collaborate to invest a considerable amount
of time, effort and financial resources in developing GIS capabilities and acquiring
expensive spatial datasets. However, it is often the case that this investment is made
without a clear understanding about what they wish to accomplish with such
capabilities, and the stakeholder groups (both within and outside the organization)
whose decision support needs they are attempting to fulfil. More often than not,
GIS capabilities of such organizations are primarily used as a tool for generating
and displaying maps. The current state of spatial methods and technology, on the
other hand, clearly indicates that GIS capabilities go beyond data management and
visualization alone.
For instance, process-based models and GIS (especially raster systems) are very
complementary tools in that the former can be used to generate new output layers
from multiple base maps already in GIS format. The derived layers often contain
information that is more useful to decision-makers. Moreover, it is now possible to
directly integrate within GIS models that have distributed (i.e. spatially explicit)
parameters, and to seamlessly run alternate scenarios within a single software
package. Prior to the availability of such techniques, it was necessary to execute
multiple model runs outside of a GIS environment, and to import the output into
GIS for visualization. Such indirect coupling often requires additional time and
technical expertise to handle model management and data transfer requirements,
besides limiting the range of scenarios that end users may wish to analyze. Direct
integration of models within GIS, however, has the potential to enable such users
in the aquaculture domain to explore alternate scenarios in a truly interactive way,
simultaneously reducing the need for the continued presence of analysts or at least
using their expertise in a manner that is more cost effective. The trend towards
interactive GIS use is already evident in both in the business world and in some
natural resource management domains such as conservation efforts.
Another observation to be made in the current context is that if the target
decision support audience is not adequately consulted to assess their needs, the
possibility exists that considerable effort may be expended by organizations to
generate information that is rarely used by the intended audience. The fundamental
message here is that if end users are involved early on in the planning process for
GIS projects and their needs better understood, experts and analysts are more likely
to generate strategies that better address these needs. Moreover, they are also more
likely to make optimal use of spatial methods and technology within any organizational constraints that may exist. It should be noted that our observations about
adopting a stakeholder-driven approach to project design and management, and
242
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
specific techniques to manage the process are not unique to GIS applications, but
are increasingly recognized within a number of domains including organizational
development (Adams, 1986; Senge, 1990), concurrent engineering design (Voland,
1992), and information technology (Ambler, 1998).
3.2. Formulating specifications
The goals and requirements of the project articulated in the first phase above are
by necessity of a general nature. This is largely because it is rarely productive to
embark upon an exercise to specify project components without a broad understanding of decision support needs. Furthermore, if team members begin to focus
too much on project specifics early on in the overall process, they run the risk of
not being able to adequately accommodate additional needs of end users (given that
such needs usually tend to evolve over time), and of already confining themselves to
a certain mindset about implementation strategies.
However, once an overall understanding of project requirements has been
developed among team members, it is helpful to develop a listing of more
functional specifications corresponding to each of the requirements that have been
identified. For instance, if the project requires that the final GIS be interactive
(implying that the end user can explore alternate scenarios on their own), one or
more of the following are representative of possible functional specifications:
capability of generating and editing different thematic maps;
features for querying attributes in spatial databases;
support for adding new themes and/or updating current information in the GIS
over time;
enabling application of the analytical approach to different geographical regions;
and
capability to modify existing models and/or link new models to the GIS.
The process of identifying functional specifications involves an in-depth analysis
of each of the project requirements. In practice, it is usually beneficial for the
subject matter experts and GIS analysts to first formulate project specifications
jointly, and to then share these with the end users. This approach tends to result in
time (and therefore cost) savings. Moreover, the need for communicating (potentially) difficult technical concepts is minimized because end users are likely to be
more interested in the actual specifications, rather than in the process by which they
were determined.
The benefits of formulating specifications for a GIS project far outweigh the
initial investment in terms of resource and time commitments. Some of the benefits
include an increased probability of implementing a system that does indeed meet
end user needs, laying the foundation for a work breakdown structure with
associated timelines and responsibilities (for different project personnel), and an
improved likelihood of building a GIS with the appropriate types of hardware and
software tools. Experience from the software industry also suggests that when
project specifications are formally identified, there is an improved probability of
minimizing cost over-runs and time delays. On the other hand, as noted by the
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
243
same author, it is safe to assume that no project can ever be fully specified
(particularly during the initial phases of development) and a judicious end point for
this phase must be identified to ensure that timely progress towards the overall
goals is maintained. It is also often the case that when functional specifications are
identified, project personnel may identify new requirements, which need to be
appropriately documented with existing information from the first phase of the
overall project (i.e. identifying project requirements; Fig. 3).
3.3. De6eloping the analytical framework
The previous two phases deal primarily with aspects of what is to be accomplished (i.e. project needs). Development of the analytical framework for a GIS
project, on the other hand, addresses issues of how these needs will be addressed.
This phase largely involves subject matter specialists and GIS analysts (Fig. 3).
However, consultations with end users are recommended to ensure that the project
will indeed address their needs, to accommodate new needs that may arise, and to
foster an improved understanding of analytical methods that may be used. An
understanding of the methods used, as well as their limitations, is critical in terms
of appropriate application of GIS outcomes for decision support.
Several methods have been used, either singly or in combination, by GIS
practitioners to integrate spatial information into a useful format for analysis and
decision making. The following represent analytical methods that have already been
used in aquaculture GIS or have potential for use in the future. Their inclusion here
is intended to provide readers with an overview of the range of analytical methods
that may come up for discussion during this phase of a GIS project.
3.3.1. Arithmetic operators
A large number of arithmetic operators can be used in GIS (Burrough, 1986).
For instance, a useful precursor is the scalar operation in which a constant term is
to be applied to spatial data (e.g. to estimate the potential annual consumption of
fish in a region, the population size of towns can be multiplied by a constant
representing the annual fish consumption per capita). Another example would be
the subtraction of monthly evapotranspiration and seepage layers from a precipitation layer to estimate mean monthly water requirements for ponds (Aguilar-Manjarrez and Nath, 1998).
3.3.2. Classification
It is almost always the case that the source data, whether in real or integer
format, will need to be further classified before further use. Classification is an
essential part of any data reduction process, whereby complex sets of observations
are made understandable. Although any classification process involves some loss of
information, a good scheme not only aims to minimize this loss, but by identifying
natural groups that have common properties, provides a convenient means of
information handling and transfer (Burrough, 1986). Further, in any classification
process, care must be taken to preserve the appropriate level of detail needed for
244
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
sensible decision making at a later stage (Burrough, 1986; Aguilar-Manjarrez, 1996;
Ross, 1998).
Aguilar-Manjarrez (1996) provides an exhaustive review of five methods that
have been explored to classify data on land types for various uses:
1. The FAO land evaluation methodology which assesses land suitability in terms
of an attribute set corresponding to different activities.
2. The limitation method in which each land characteristic is evaluated on a
relative scale of limitations.
3. The parametric method in which limitation levels for each characteristic are
rated on a scale of 0 to 1, from which a land index (%) is calculated as the
product of the individual rating values of all characteristics.
4. The Boolean method which assumes that all questions related to land use
suitability can be answered in a binary fashion, and that all important changes
occur at a defined class boundary.
5. The fuzzy set method in which an explicit weight is used to assess the impact of
each land characteristic. Fuzzy techniques are then used to combine the evaluation of each land characteristic into a final suitability index. Apart from a
dominant suitability class, the fuzzy set method equally provides information on
the extent to which a certain land unit belongs to each of the suitability classes
discerned.
For GIS applications, any of the above methods can be used to classify source
data into a four- or five-point scale of suitability (with one being the least suitable).
However, the choice among classification methods is dependent on the type of data
and intended uses of the output information. Classification allows normalization of
all data layers, an essential pre-requisite for further modeling. An example is the
grouping of soil data into four classes based on a number of properties important
for pond construction (Kapetsky and Nath, 1997).
3.3.3. Simple o6erlay
The most common technique in geographic information processing is that of
topological overlaying in which multiple data layers are overlain in a vertical
manner. Topological overlays may be broadly classified into simple and weighted
(see below) methods. After reclassification of thematic maps in terms of suitability
for an activity has been accomplished, simple overlay can be accomplished by
applying mathematical or logical operators to the layers. This can often generate
very valuable outcomes. Simple overlays operations assume that all layers have
equal importance. Outcomes of the operation can be further refined if they are
overlain with constraint layers by subtraction. A more common practice, however,
is to represent constraint layers in Boolean format, with disallowed and allowed
areas defined as 0 and 1, respectively. In this way, when the constraint layer is
multiplied by one containing suitability data, the former is automatically excluded
from the outcome. A series of such simple reclassifications and overlays can be
assembled as a tree of sequential choices (i.e. a decision tree) to provide a rational
outcome to a spatial problem. Examples of constraint layers are large water bodies
and forested lands that are unavailable for aquaculture (Kapetsky and Nath, 1997).
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
245
3.3.4. Weighted o6erlay
In decision making, it is usually the case that different attributes under consideration do not have the same level of importance. This calls for a weighted overlay
approach, in which each layer is assigned a weight that is proportional to its
importance. To accomplish this, each source layer is first reclassified onto a
common scale and then multiplied by a weighting factor. The resulting values are
used in further overlay operations to obtain an outcome (applying Boolean
constraints in the final stage if needed).
The derivation of weightings for source attributes is often assumed to be an
objective scientific matter. However, it is clear from a range of studies that even
expert opinions on ranking of attributes may differ substantially. Aguilar-Manjarrez (1996) showed that, from a list of attributes, a group of experts with similar
backgrounds generally agree upon which are the most relevant to use for any given
decision process. However, the ranking assigned to the attributes by these experts
can differ. In general, Aguilar-Manjarrez (1996) showed that from a list of ten
selected attributes, different experts would be in agreement on the priority of the
first three to four and the last three. However, attributes with mid-level priority are
often dealt with quite differently. Moreover, subject matter experts with different
backgrounds (e.g. aquaculturist, coastal zone planner, conservationist, etc.) bring
differing agendas to the same problem, resulting in a range of outcomes. At worst,
it is clear that without guidance, a range of prioritizations can be obtained which
are cumulatively meaningless. GIS analysts therefore need to ensure that weightings
applied personally are as objective as possible and that where expert group input is
used, the basis for this input is fully explained and carefully analyzed before its use
for decision making. Summary tables that transparently document factors used in
the evaluation and the weights assigned to them by different experts are a useful aid
in this regard (Aguilar-Manjarrez, 1996; Aguilar-Manjarrez and Nath, 1998).
As the number of layers in a model increases, the logic of processing them can
become confusing. Consequently, special routines to assist with this process have
been provided in certain GIS packages. For example, the decision support module
in IDRISI (Eastman, 1993) supports multi-criteria evaluation (MCE), which is
based on the analytical hierarchy process (AHP) described by Saaty (1980). This
module enables a structured, logical approach to weighted overlay with built-in
crosschecks at some stages of the weighting process to ensure consistency of logic
by the operator. AHP falls into the broader category of pairwise comparison
techniques in which attributes are ranked against each other to assess their relative
importance. Although AHP or other pairwise comparison variants do provide
opportunity for increased objectivity in the ranking of factors, they do not
adequately address the issue of different weightings that may be assigned by diverse
subject matter experts. For instance, different weightings were assigned to factors
important for pond aquaculture by different experts in continental-scale GIS
studies (Kapetsky and Nath, 1997; Aguilar-Manjarrez and Nath, 1998). The AHP
process was useful in these studies in terms of providing a logical framework for
comparing the factors. However, in both instances, the authors ultimately relied on
their own ranking schemes (which had similar patterns to those of the experts) to
246
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
assign weights to the factors. In general, it is unlikely that technical methods such
as AHP or other multi-criteria decision making tools (see review by Merkhofer,
1998) can be a complete replacement to a well-facilitated exercise in which
stakeholders (i.e. subject matter specialists, GIS analysts and end users in the
current context) are encouraged to seek consensus on the relative importance of
attributes for use in GIS. Instead, such tools should be viewed as providing
additional support to the consensus building process.
3.3.5. Neighborhood analysis
A key function provided by GIS, and one which cannot be easily addressed by
the use of any other decision support tool, is its capability to allow evaluation of
the characteristics of an area that surrounds a specific location, which is referred to
as neighborhood analysis. GIS software provides a range of neighborhood functions including point interpolation techniques in which unknown values are estimated from known values of neighboring locations. These techniques are typically
used to convert point coverages into grids or to generate elevation data either in
triangular irregular network (TIN) format for vector GIS or as a digital elevation
model (DEM) for raster GIS. DEMs have not as yet found widespread use in
aquaculture, but are often at the heart of environmental GIS that are targeted
towards analyses of entire landscapes. For instance, the DEM for a watershed of
interest is a key layer in many spatially explicit hydrological models. In the future,
such tools are likely to find applications in aquaculture from the perspectives of
assessing water availability and modeling environmental impacts of existing and/or
proposed aquaculture operations. A common type of neighborhood operation is
buffering, which allows the creation of distance buffers (or buffer zones) around
selected features. An example where this operation may be useful in aquaculture is
when there is a need to determine how many streams are available (or to be
avoided) within a specified distance of a proposed facility.
3.3.6. Connecti6ity analysis
This type of analysis is characterized by an accumulation of values over the area
that is being traversed. Support for connectivity analysis in commercially available
GIS software is not as standardized in comparison to the operators described
above. Network analysis is one type of connectivity operation, which is characterized by the use of feature networks such as hydrographic hierarchies (i.e. stream
networks) and transportation networks. As an example of the use of network
analysis in aquaculture, Kapetsky and Nath (1997) estimated the shortest transportation path from a given site to the nearest market.
The most sophisticated type of connectivity operation is three dimensional
analysis which is helpful in terms of visualizing impacts of alternate decisions on
entire landscapes (e.g. watersheds). Such advanced features of GIS are likely to
have applications in aquaculture but have apparently yet to be used in the systems
that have been developed in this domain.
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
247
3.3.7. Hierarchical models
In practice, comprehensive representation of the real world in a GIS often
involves the use of a large number of source variables, which when processed into
GIS using one or more of the analytical methods described above, may result in
considerable complexity. Experience suggests that when the number of layers
exceeds about 10, MCE becomes difficult, even to the experienced modeler. Such
situations warrant the development of a hierarchical modeling scheme.
In this approach, naturally grouped variables are first considered together to
produce ‘sub-model’ outcomes such as water needs, soil suitability, input availability, farm gate sales, and markets (Kapetsky and Nath, 1997). It is often the case
that a source variable or processed layer will be used in more than one sub-model
and that the layer may need to be transformed depending on the intended purpose.
Each of these sub-models may, in turn, be derived from lower-level models which
pre-process variable data into useful factors. Another example from aquaculture is
the estimation of wave height from wind velocity and fetch (Ross et al., 1993).
It is, of course, not necessary to embed all of the models and/or sub-models
within a single system per se. This is because commercial GIS can export and
import information quite seamlessly with external modules. For instance, growth
estimates derived from an external bioenergetic model (linked to spatial weather
and water temperature datasets) have been used to predict fish yields in continentalscale GIS (Kapetsky and Nath, 1997; Aguilar-Manjarrez and Nath, 1998). In a
different context, external rule-based systems are being used in conjunction with
GIS to generate future agricultural land use patterns for assessing policy alternatives in the Pacific Northwest region of the United States (Bolte, unpublished
information).
3.3.8. Multi-objecti6e land allocation
Once an activity has been modeled and quantified, it will invariably have some
potential to conflict with other uses of the space or resources. For example,
practices such as agriculture, forestry, and aquaculture may compete for available
land and associated resources. This calls for trade-off decisions to be made so that
activities can coexist. These decisions typically require consideration of economic,
environmental and social ramifications of alternative land use practices. Decision
support tools are available in some GIS packages to facilitate this process. The
multi-objective land allocation (MOLA) and multi-dimensional decision space
(MDCHOICE) tools in IDRISI are good examples. MOLA was used by AguilarManjarrez and Ross (1995) to identify areas suitable for agriculture and aquaculture across the Sinaloa state of Mexico (see section on case studies below). Inputs
for the MOLA technique originate from the results of weighted overlays. The
approach allows GIS analysts to set limits on areas required for different land uses
and assign requirements for these uses. An iterative process is then used to
successively reassign ranked cells to alternate activities depending upon how closely
they match the associated requirements.
248
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
3.4. Identifying data sources
Once the analytical framework has been developed, it is necessary to identify
data sources to be used in the project. This phase is largely restricted to GIS
analysts (Fig. 3), although subject matter specialists often provide helpful advice
(e.g. identifying non-spatial datasets). Information for spatial decision-making and
analysis is varied, and will usually consist of data describing the biophysical,
economic, social and infrastructural environments. These data can come from a
variety of sources ranging from primary data gathered in the field or satellite scenes
to all forms of secondary data, including textual databases and reports.
It is generally both costly and time consuming to collect field (primary) data first
hand. Therefore, all GIS practitioners attempt to locate the data they need from
existing secondary sources, either in paper or digital form. The initial consideration
is identifying what data are needed for the overall analysis. This is followed by
attempts to source the data, and to assess their age, scale, quality and relative cost.
Different data types are often developed in different geographic coordinate systems,
and must be re-projected onto a common coordinate system. Other common issues
are ensuring that features common across multiple layers are spatially coincident,
and understanding the resolution, constraints and uncertainty associated with the
data. For these reasons, data collection and preprocessing are typically the most
expensive and time-consuming component of an analysis.
Digital database availability is increasing at a rapid rate throughout the world.
These databases contain information ranging from natural resources (e.g. maps of
soils, water resources and temperature distributions), to population census and
cadastral (i.e. property ownership and associated boundaries) data. In many cases,
such data may only be available in hard copy reports, although information is also
available on CD-ROMs, from which databases for GIS use can be extracted.
A further source of data is the WWW, from which a wide range of mapped and
other spatial data can be found by carefully executed searches. Spatial information
at low resolution can often be obtained in this way, good examples being the
Global 30 arc second elevation (http://edcwww.cr.usgs.gov/landdaac/gtopo30/
gtopo30.html), and the 1-km Global Land Cover datasets. Other online datasets
include information on drainage basins, and weather, which can be downloaded
from Web sites at some universities and international agencies. In this regard, a
useful site that provides metadata search capabilities and access to a range of data
resources is the one maintained by the Center for International Earth Science
Information Network (CIESIN) at http://www.ciesin.org. In some countries such as
the USA, GIS clearinghouses have been established on the Web and offer access to
consolidated datasets from multiple agencies (see http://www.usgs.gov). Many of
these free spatial databases are of immense value and should not be overlooked.
The ‘lingua franca’ of GIS is the thematic map in which particular attributes of
the geographical region are represented digitally or on a specially prepared hard
copy map. Clearly, where such maps are available in digital form, they are directly
usable in a GIS. Hard copy maps can also be digitized, and the information
imported into GIS. The digitizing process, however, is very time consuming and
requires many hours of data editing to ensure superior quality.
S.S. Nath et al. / Aquacultural Engineering 23 (2000) 233–278
249
Thematic data can be in the form of a choropleth (areas of equal value separated
by boundaries), isopleth (lines which connect points of equal value) or point maps
(Burrough, 1986), all of which are often valuable, a good example in terms of
aquaculture being the hydrographic chart. The more common topographic map
frequently contains many thematic data (including elevations, water bodies, roads,
cities, and woodlands) which are of value to the GIS analyst. These themes can be
extracted at the stage of digitization, and established as separate layers in the
spatial database.
3.5. Organizing and manipulating data
After datasets have been collected, it is necessary to organize and manipulate
them for use in the target GIS. This phase is also largely restricted to GIS analysts
(Fig. 3), although depending on the type of application, occasional interaction with
subject matter specialists may be warranted. Some of the key activities that occur
in this phase include verification of data quality, data consolidation and reformatting, creation of proxy data and database construction. Proxy data refer to
information that is derived from another data source, for which established
relationships may exist. Examples include estimation of water temperature from air
temperature (Kapetsky, 1994), extraction of semi-quantitative texture from FAO
soil distribution maps (Kapetsky and Nath, 1997) or calculation of maximum wave
heights from wind speed and fetch (Ross et al., 1993).
In terms of verification of data quality, the reliability of thematic maps worldwide is variable, as is the currency of their content. These aspects must be
considered where critical decisions are based upon such material. Although digital
maps are often quite up to date, the spatial accuracy of some printed material can
be suspect. Critical assessment and verification of source data quality is very
important. The value of such verification for all input data cannot be over-stressed,
before digitizing as well as after the fact. It is usually the case that at least some of
the layers required in a GIS database will not be of a high enough standard, and
verification on paper may very well need to become verification by survey, where
warranted. A detailed overview of technical methods to address data quality issues
is available in Burrough (1986).
Certain data sources such as satellite data are already in digital form, but all
others may requir