University of Southern California Dissertations and Theses

Deriving traverse paths for scientific fieldwork with multicriteria evaluation and path modeling in a geographic information system

(USC Thesis Other)

Deriving traverse paths for scientific fieldwork with multicriteria evaluation and path modeling in a geographic information system

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

DERIVING TRAVERSE PATHS FOR SCIENTIFIC FIELDWORK WITH MULTICRITERIA
EVALUATION AND PATH MODELING IN A GEOGRAPHIC INFORMATION SYSTEM

by

Ryan Richardson Reeves

A Thesis Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF SCIENCE
(GEOGRAPHIC INFORMATION SCIENCE AND TECHNOLOGY)

May 2015

Copyright 2015 Ryan Richardson Reeves
ii

ACKNOWLEDGMENTS
I extend my gratitude to my thesis committee members, Dr. Karen Kemp, Dr. Travis Longcore,
Dr. Jennifer Swift, and Dr. Kyle House, for their expertise on various subjects and overall
contribution to my thesis work. I would like to extend a special thanks to Dr. Kyle House for
allowing me to assist him with a portion of his research along the lower Colorado River and the
valuable insight he provided during that time. I would like to thank Dr. Jordan Hastings for his
assistance with various aspects of my thesis work during its early stages and for his assistance
with presenting this work at the Digital Mapping Techniques 2014 Workshop. Also, I would like
to thank my wife and the rest of my family for their constant support.

iii

TABLE OF CONTENTS
ACKNOWLEDGMENTS .............................................................................................................. ii
LIST OF TABLES .......................................................................................................................... v
LIST OF FIGURES ....................................................................................................................... vi
ABSTRACT .................................................................................................................................. vii
CHAPTER ONE: INTRODUCTION ............................................................................................. 1
1.1 A Methodology to Explicitly Define Traverse Paths ............................................................ 2
1.2 Thesis Organization .............................................................................................................. 4
CHAPTER TWO: BACKGROUND .............................................................................................. 5
2.1 Fieldwork Planning and Technology .................................................................................... 5
2.2 Traverse Planning and Related Work ................................................................................... 6
2.2.1 Traverse Planning in Forestry ....................................................................................... 6
2.2.2 Traverse Planning in Geological Science...................................................................... 7
2.2.3 Traverse Planning in Planetary Science........................................................................ 8
2.3 Multicriteria Evaluation ...................................................................................................... 10
2.3.1 Problem Definition....................................................................................................... 11
2.3.2 Criterion Selection ....................................................................................................... 12
2.3.3 Standardization ............................................................................................................ 12
2.3.4 Allocation of Weights ................................................................................................... 16
2.3.5 Implementation of Aggregation Algorithm .................................................................. 18
2.3.6 Sensitivity Analysis and Analysis of Outcome ............................................................. 20
2.3.7 Using MCE Results to Make Decisions ....................................................................... 21
2.4 Path Modeling ..................................................................................................................... 21
CHAPTER THREE: METHODOLOGY ..................................................................................... 23
3.1 Identify Objectives and Criteria .......................................................................................... 24
3.2 Assemble Relevant Data ..................................................................................................... 25
3.3 Sketch and Derive Features of Interest Relevant to Meeting Objective(s) ......................... 26
3.4 Apply Necessary Manipulations or Analysis to Derived Features of Interest .................... 26
3.5 Standardization ................................................................................................................... 26
3.6 Allocation of Weights ......................................................................................................... 27
3.7 Implementation of Aggregation Algorithm ........................................................................ 27
3.8 Review Results of Multicriteria Evaluation ........................................................................ 27
3.9 Define Traverse Paths on Basis of Time Availability ........................................................ 27

iv

CHAPTER FOUR: DEMONSTRATION OF METHODOLOGY .............................................. 31
4.1 Identify Objectives and Criteria .......................................................................................... 32
4.2 Assemble Relevant Data ..................................................................................................... 36
4.3 Sketch and Derive Features Relevant to Meeting Objectives ............................................. 38
4.3.1 Derivation of Criterion Layers Representing Scientific Return .................................. 39
4.3.2 Derivation of Criterion Layers Representing Accessibility ......................................... 42
4.4 Standardization ................................................................................................................... 44
4.5 Establish Field Campaign Weights ..................................................................................... 49
4.6 Production of Weighted Overlay Layers ............................................................................ 52
4.7 Prioritize Target Paths for Fieldwork on Basis of Time Available .................................... 54
4.8 Derived traverse paths ........................................................................................................ 56
4.9 Analysis of Results and Sensitivity Analysis ..................................................................... 58
4.9.1 Analysis of Result from Aggregation ........................................................................... 59
4.9.2 Change in Weights for Scientific Return and Accessibility ......................................... 61
4.9.3 Change in Percent used to Derive Origin and Destination Points .............................. 63
CHAPTER FIVE: DISCUSSION AND CONCLUSION ............................................................ 66
5.1 Benefits Gained by Using this Methodology ...................................................................... 66
5.2 Limitations of Methodology and Suggestions for Improvement ........................................ 67
5.3 Conclusion and Opportunities for Future Work ................................................................. 70
REFERENCES ............................................................................................................................. 71

v

LIST OF TABLES
Table 1 - Summary of Standardization Techniques ...................................................................... 13
Table 2 - Summary of Weighting Techniques .............................................................................. 16
Table 3 - Intensity of Importance .................................................................................................. 17
Table 4 - Pair-Wise Comparison Example ................................................................................... 18
Table 5 - Summary of Aggregation Techniques ........................................................................... 19
Table 6 - Methodology Steps, Required Work, and the Related Layers ...................................... 24
Table 7 - Criteria and Procedural Assumptions Relevant to Geologic Fieldwork ....................... 35
Table 8 - Pair-Wise Comparison Results ...................................................................................... 49
Table 9 - Example AHP Matrix .................................................................................................... 50
Table 10 - Derived Weight. ......................................................................................................... 52
Table 11 - Statistical Summary of Traverse Path ......................................................................... 58

vi

LIST OF FIGURES
Figure 1 - Multicriteria Evaluation Flowchart. ............................................................................. 11
Figure 2 - Example Derivation of Traverse Paths. ....................................................................... 29
Figure 3 - Workflow to Create Traverse Path Segment. ............................................................... 30
Figure 4 - Hypothetical Geologic Map. ........................................................................................ 33
Figure 5 - Location Map for Demonstration. ................................................................................ 34
Figure 6 - Features of Interest Relevant to Scientific Return Objective ....................................... 42
Figure 7 - Access Features of Interest ........................................................................................... 43
Figure 8 - Slope Criterion ............................................................................................................. 44
Figure 9 - Standardization Function for Regions, Access, and Visibility Criteria. ...................... 46
Figure 10 - Standardization Function for Access Criterion. ......................................................... 47
Figure 11 - Derivation of Five Criteria ......................................................................................... 48
Figure 12 - Outputs from Aggregation Steps ................................................................................ 53
Figure 13 - Flowchart of GIS Work. ............................................................................................. 55
Figure 14 - Flowchart of Aggregation Steps................................................................................. 57
Figure 15 - Final Traverse Path. ................................................................................................... 58
Figure 16 - Sample Point Analysis. .............................................................................................. 60
Figure 17 - Change in Weights of Influence. ................................................................................ 62
Figure 18 - Variation in Centroids Due to Change in Weights of Influence. ............................... 63
Figure 19 - Change in Destination Location Resulting from Percent Favorability Used. ............ 65
vii

ABSTRACT
Field research is a necessary component of many realms of ecological and geoscientific practice
since it provides the primary data crucial to understand the characteristics of an object,
phenomenon, or process. Unlike work in an office or laboratory, fieldwork has additional cost
related to travel, lodging, and per diem expenses. Field scientists must therefore ensure they
make efficient and effective field navigational decisions that result in expedient execution of
field campaign objectives.
Technologies and analytical approaches such as decision analysis, path modeling, and
geographic information systems offer assistance to navigational decision making while in the
field as do the analytical techniques of weighted linear combination and analytical hierarchy
process. These tools are often underutilized, however. This thesis describes a methodology by
which these technologies and analytical procedures may assist field scientists with navigational
decision making. Specifically, the thesis documents development of a model that uses a spatial
multicriteria decision evaluation to derive favorability values. These values are then used to
determine the placement of traverse paths that are suggested routes to be taken by field
researchers. The thesis includes a description of concepts behind the methodology, a
demonstration of the methodology for a hypothetical geologic campaign, and an analysis of
resulting traverse paths.

1

CHAPTER ONE: INTRODUCTION
Field research is a necessary component of many realms of ecological and geoscientific practice
since it provides the primary data crucial to understand the characteristics of an object,
phenomenon, or process. While data may be acquired from existing maps and other publications
and through various remote sensing techniques, scientists still need to work in the field,
interacting directly with the entity in which they are interested.
An inherent component of this fieldwork is the need to navigate across a study area in
order to make observations or attain measurements and samples, a process referred to in some
domains as making a traverse. Scientists carefully record descriptions of the observations and
measurements they make while on a traverse and often use global navigation satellite systems
(GNSS) to record precise coordinates of where these data were collected. Scientists use these
primary data, and information derived from them, as evidence to explore a set of existing
hypotheses or to discover patterns. This information then allows scientists to constrain the set of
plausible hypotheses about the origin and distribution of features within the study area. This
knowledge may then be used to better understand a region's likelihood for the presence of a
hazard, species habitat, or resource, for its suitability for a human-made structure, and so on.
While in the field, scientists are faced with scientific and logistical objectives. The
primary scientific objective is to achieve a high scientific return. This is considered to occur
when efforts made result in substantial progress related to expanding the knowledge regarding an
object, phenomenon, or process. A high scientific return is often attained at sites that provide
opportunities for direct observation and measurement of materials that help test current working
hypotheses, exceptional views of features of interest, or broad views of the study area or entity of
interest.
2

The logistical objectives of fieldwork relate to issues of access, preparation, and time.
Issues of access may include constraints such as property ownership, waterways, slope of the
terrain, lack of access trails, and hazardous topography. Preparatory issues relate to such aspects
as weather, lodging, equipment, and so on.
Unlike work in an office or laboratory, fieldwork has costs related to travel, lodging, and
sometimes per diem allowances. These costs, both in time and in money, make it necessary to
conduct fieldwork in an efficient and effective manner. To do so, scientists are required to
strategically consider how they may fulfill their particular field campaign objectives. One way to
do this is to plan a cost- and time-effective traverse.
Technologies and analytical approaches such as decision analysis, path modeling, and
geographic information systems offer assistance to navigational decision making while in the
field as do the analytical techniques of weighted linear combination and analytical hierarchy
process. These tools are often underutilized, however, in these contexts. This thesis describes a
methodology by which these technologies and analytical procedures may assist scientists with
navigational decision making.
1.1 A Methodology to Explicitly Define Traverse Paths
Field scientists have long used published maps and remotely sensed imagery to
understand the scientific value and difficulty of traversing through particular terrain. Often these
tools are used in an informal manner, with decisions being made implicitly, without the use of
formally defined analytical models.
As this thesis shows, however, a supplemental approach is to make navigational
decisions more explicitly by using analytical techniques from decision analysis, path modeling,
and geographic information science. Such explicit navigational models are needed and have been
3

developed for geologic fieldwork in planetary science where all work is supervised remotely
(e.g., Carr et al. 2003, Hӧrz et al. 2010, Johnson 2010, Skinner and Fortezzo 2013). Examples
are also found outside the realm of geologic science. Store and Antikainen (2010) is a good
example from the domain of forestry.
Building on tools used in other domains, this thesis presents a methodology that makes
explicit the often ad hoc and implicit process of field navigation for scientific fieldwork. This
methodology is intended to provide scientists with a set of traverse paths that have been derived
using explicit tools and processes that can be used to supplement traditional traverse planning.
As will be demonstrated later, these traverse paths are designed to place scientists close to
locations where they may attain a high scientific return, using as much as possible existing routes
with low slope that provide increased accessibility.
This methodology collectively considers many of the criteria, objectives, and constraints
associated with fieldwork, while accounting for their relative importance, to suggest strategic
traverse paths to be used by field scientists. This methodology is first presented in a generic
sense, outlining the sequence of steps needed to derive the traverse paths. Then, it is
implemented for a hypothetical geologic field campaign in Hildago County, New Mexico. This
demonstration shows how the methodology might be applied using Esri ArcGIS version 10.2.2
to create and overlay the criteria layers and develop the traverse paths for this example field
campaign. All GIS work performed for this thesis was administered using Esri ArcGIS version
10.2.2. Lastly, an initial assessment of the validity of results from this methodology is performed
through analyses of the resulting traverse paths.

4

1.2 Thesis Organization
Chapter One has introduced the nature of fieldwork and outlined the objectives of the
present thesis. Chapter Two offers a background to the relevant aspects surrounding fieldwork,
multicriteria evaluation, and path modeling and describes other works related to deriving traverse
paths in the fields of planetary science, geological science, and forestry. Chapter Three outlines
the concepts underlying the traverse generating methodology. Chapter Four provides a
demonstration of its employment and examines the traverse paths generated from the
methodology. Chapter Five discusses the advantages and limitations of the methodology and
considers its usability and suggest potential improvements and future work.

5

CHAPTER TWO: BACKGROUND
This chapter provides background regarding the process of planning fieldwork as well as some of
the data and technologies that are available for its practice. Also, this chapter provides an
introduction to the decision analysis technique termed multicriteria evaluation (MCE), including
a brief description of its components of standardizing, weighting, and combining criteria, and the
technique of path modeling.
2.1 Fieldwork Planning and Technology
Scientists often use archival data such as government records, maps, and remotely sensed
imagery to inform fieldwork planning. For example, a team of soil scientists and plant ecologists
conducting a soil survey within Denali National Park and Preserve, Alaska used aerial
photography and satellite imagery to interpret landform and vegetative characteristics and
distribution (Clark and Duffy 2006). These interpretations allowed them to delineate
representative sites of the various physiographic regions within the park and thus identify the
locations where fieldwork should occur.
Geologists, too, use archival data to conduct preliminary inspections of study areas prior
to visitation (Compton 1985, Coe 2012). This may involve digital processing of satellite imagery
to characterize and distinguish varying rock types (e.g. Mars 2013), study of existing geologic
maps and reports, or personal communiqué from other field scientists. Scientists are able to use
this archival data prior to conducting their fieldwork to reduce the size of the sampling area
necessary to fully describe a study area.
The tools that may be used to plan a traverse and augment it while in the field are quickly
advancing. House et al. (2013) describe how technological advancements regarding geographic
information systems, light detection and ranging (LiDAR), virtual globes, mobile hardware and
6

software, and geocoded field data are changing the practice of geologic mapping. Nevertheless,
fieldwork is still an essential and costly activity, so the incorporation of such advanced
technology to improve the process of traverse planning is important.
2.2 Traverse Planning and Related Work
A few researchers have attempted to strategically define the traverse planning process.
Examples of some of these efforts can be found in the fields of planetary science (Carr et al.
2003, Johnson 2010, Hӧrz et al. 2013, Skinner and Fortezzo 2013) and forestry (Store and
Antikainen 2010). These are discussed in the following sections.
2.2.1 Traverse Planning in Forestry
To advance the effectiveness of field inventorying in Finnish forests, Store and
Antikainen (2010) demonstrate how to determine visitation sites and design optimal routes to
reach those sites. They combine decision analysis and path modeling techniques using a GIS
platform. Sites deemed most important for forest inventorying endeavors and areas that affect
traversability are selected through a multicriteria evaluation (MCE) process involving expert
knowledge modeling and the analytical hierarchy process (AHP). A custom path modeling tool
was developed to derive paths among a series of selected forest stands. This tool uses greedy
heuristics and a variation of the traveling salesmen problem to establish the optimal solution for
their traverse path, which they view as an orienteering problem, .
Store and Antikainen note that existing tools that may be used to determine optimal
traverse paths using least-cost path analysis were too inaccurate for their needs. The technique
they use relies on nodes at a raster's cell boundary as opposed to its center, noting that this
enables "a more accurate calculation of traverse path" (Store and Antikainen 2010, 156). In an
effort to mitigate elongation and deviation errors caused by the raster data model, as presented by
7

Goodchild (1977), their technique also uses a rectification procedure to give priority to bent-line
over straight-line segments.
The MCE process allows decision makers to assign weights to the various criteria and
objectives relevant to fulfilling a particular goal. These weights indicate the relative importance
of each criterion and objective. Store and Antikainen note that their work lacks a sensitivity
analysis step to determine the effects of variations in the values of these weights. Such an
analysis will determine if small adjustments to the input of a MCE results in significant changes
in its output. If significant changes do occur, it often means that the model is overly sensitive and
adjustments that mitigate these affects need to be performed.
2.2.2 Traverse Planning in Geological Science
Despite numerous advancements in remote sensing technologies (e.g., digital image
processing and geophysical analysis), there continues to be a need for geologists to visit the field.
These technologies may actually support the need for further surveys, as they often reveal gaps
and potential error in current knowledge. Compton (1985) explains that while the geologic
makeup of many areas has been mapped (e.g., the contiguous United States at a scale of
1:24000), advancements in geologic theory and techniques related to new remote imagery, field
techniques, and analysis create the need for these surveys to continue. Passchier and Exner
(2010) describe that many areas lack geologic information with sufficient detail and that new
understanding of many geologic phenomenon renders many older geologic maps obsolete. Ernst
(2006) describes how he simply uses geologic field mapping to acquire answers to specific
geologic questions.
Prior to arriving in the field, geologists may have outlined areas they intend to go and
perhaps even plan what appears to be the best way to get there. However, based upon new
8

observations made while in the field, the initial plans often change. As observations are made, a
geologist gradually becomes more aware of the geologic makeup of the study area, as well as
discovers its imposing obstacles. Also, geologists will often operate under multiple working
hypotheses (Chamberlin 1890) in contrast to the ruling hypothesis mentality dominating many
other facets of science. New observations may confirm or refute these hypotheses and thus
change where the geologist chooses to go.
This does not, however, undermine the advantages wrought through careful consideration
of the available archival data. Field geologists will often bring this archival data along with them
into the field and use it to supplement the primary data they collect there. Field geologists have
long used geologic maps, topographic maps, and remotely sensed imagery to understand the
scientific value and difficulty of traversing through particular terrain. Traditionally, these tools
have been used in an informal manner, with decisions being made implicitly, without the use of
formally defined analytical models (Riggs et al. 2009).
2.2.3 Traverse Planning in Planetary Science
A predicted increase in the amount of surface travel that will be conducted during future
planetary missions has led researchers to develop means by which the space crew may plan and
conduct traverses in an efficient and effective manner (Johnson et al. 2010). Many of the factors
affecting planetary fieldwork, such as concerns regarding thermoregulation, oxygen support,
depth perception, are not applicable to fieldwork on Earth. Like fieldwork on Earth, however,
planetary fieldwork contains many expenses not relevant to work in an office or laboratory. It is
therefore sensible that those concerned with extraterrestrial geologic fieldwork would be
concerned with matters affecting its efficiency and effectiveness.
9

Desert Research and Technologies Studies (DRATS) are analog planetary endeavors
carried out in northern Arizona to test various hardware and operations (Ross et al. 2013).
Skinner and Fortezzo (2013), in their work to assist the 2010 DRATS team, used photogeologic
mapping to gain an initial understanding of their study area and to identify key locations where
remaining research questions could be addressed. Their work involved first identifying the study
area’s various geologic materials based on imagery characteristics. Then, a geologic map was
constructed based upon these distinctions. Upon completion of the map, numerous questions
remained. Sites interpreted as locations where these questions could be deciphered became
visitation sites recommended for the DRATS team.
Horz et al. (2013), in similar work aimed at assisting the same 2010 DRATS team in their
analog mission, made a series of traverse paths intended to account for a series of technical and
operational constraints associated with planetary geologic campaigns. The work described in
their article was supported by the earlier work of Skinner and Fortezzo (2013). Sites given
visitation preference were determined based upon scientific return and logistical considerations
such as slope trafficable by rovers, road conditions, and fence and gate locations. The best
traverse paths were determined manually via group consensus as the result of an on-site
reconnaissance trip and workshop.
This work by Skinner and Fortezzo and Horz et al. demonstrates the process of
discerning what may be seen within satellite imagery and using these observations to derive and
test hypotheses remotely. These hypotheses may then be used to select key locations within a
study area where geologists are likely to attain a high scientific return.
Carr et al. (2003) and Johnson et al. (2010) used similar approaches to generate traverses
to be used for planetary extravehicular activities. Both developed MATLAB based tools and
10

assessed what effect factors such distance, time, energy, slope, and visibility have on potential
traverse paths. These tools provide geologists with information regarding metabolic cost,
visibility, mission compliance, and hazards once they have identified a series of waypoints.
2.3 Multicriteria Evaluation
MCE is set of analytical procedures that may be thought of as a sub-discipline of
multicriteria decision analysis (MCDA) or multicriteria decision making (MCDM) (Carver
2008). Many of the methods applied in MCE originated from the field of operations research in
the 1960s and 1970s (Carver 2008). They arose in response to critiques of early techniques in
decision making and site location analysis (Carver 1991). The MCE approach combines multiple
datasets representing various criteria and/or objectives, assigns them with a weight indicating
their relative importance, and produces a multi-valued output (e.g. a raster data model with a grid
of georeferenced cells with different values) indicating the degree to which an objective(s) has
been met.
The term criteria is often used generically to refer to concepts of both criterion attributes
and objectives. It is used here to refer to attributes of entities or phenomena that may be used to
measure the fulfillment of a certain objective, or various objectives. This process may be done
for geographic space by designing such an evaluation around spatial data. A GIS is often used
for this due to its ability to store, display, and analyze this data relevant to many decision
problems (Carver 1991).
Geographic information systems alone, while advantageous for working with various
types of spatial data in a wide variety of applications, were originally not designed to handle an
analysis involving a complex value structure consisting of conflicting objectives and varying
priorities (Malczewski 1999). In 1991, Carver described a GIS as a data management framework
11

for the spatial data used in a MCE. He noted that a MCE provides a GIS with the ability to
handle conflicting objectives that encompass multiple criteria and multiple decision makers.
Now, two decades later, most GIS do provide a means by which at least some MCE techniques
may be implemented directly within the GIS framework. By incorporating the technologies
associated with MCE and GIS, decision makers are able to confront spatial problems containing
multiple criteria, objectives, and decision makers.
Carver (2008) outlines the main steps involved in a multicriteria evaluation. These are
problem definition, criterion selection, standardization of criterion scores, allocation of weights,
and implementation of an aggregation algorithm. Additional steps such as a sensitivity analysis
and making decisions with the processed information may also be included. Problem definition
involves identifying the difference between existing and desired states of a system (Malczewski
1999). Once the problem has been identified, it can be determined that the achievement of a
certain objective(s) may bring the system closer to the desired state. Once the attribute values of
multiple criteria have been standardized, weighted, and aggregated, they may then be used to
determine the degree to which an objective(s) has been met. A sensitivity analysis is performed
to discover error or uncertainty that may be contained within the derived values. Once confident
that the values attained are of sufficient quality, they may be used to make decisions (Figure 3).

Figure 1 - Multicriteria Evaluation Flowchart: Typical flowchart of multicriteria evaluation
procedures.
2.3.1 Problem Definition
When defining a problem for the MCE process, considerations must be made as to
whether there are groups of people with different vested interests in the decision problem.
12

Groups, as opposed to individuals, are of concern as the decision making process will be affected
more by the amount of conflicting goals, preferences, and beliefs than by the number of those
involved (Malczewski 1999).
2.3.2 Criterion Selection
Criteria are selected based upon their ability to measure the degree to which an
objective(s) has been met. Criteria are often defined as either factors or constraints. Factors
describe criteria attributes that promote fulfillment of a given objective, while constraints
describe criteria attributes containing hard limitations to objective fulfillment. There are several
ways to select criteria including a survey of relevant literature, analytical studies, and an opinion
survey (Malczewski 1999).
2.3.3 Standardization
Data used as input to derive criteria layers are likely to contain varying value scales.
Input data may use nominal scales such as soil and rock types, or quantitative scales such as
slope and distance to a feature. They may also be based upon natural or constructed scales
(Keeney and Raiffa 1976). A natural scale is often considered objective meaning that it is
standard and may be measured, whereas a constructed scale is considered to be subjective in that
it is based on opinion.
Given the variety of scales used, data for each criterion must be standardized prior to
being compared. Any mathematical or logical function may be used to describe the relationship
between input data and the developed criterion layer. This relationship, however, should be
based on a defensible association (Bolstad 2012). The standardized values may also be seen as
having a direction (Voogd 1983). For instance, as explained by Malczewski (1999), when using
a floating point scale from zero to one, the criteria are seen as benefit criteria when favorable
13

characteristics are given a high score (e.g., one) and detrimental characteristics are given a low
score (e.g., zero). The criteria are seen as cost criteria when favorable characteristics are given a
low score (e.g., zero) and detrimental characteristics are given a high score (e.g., one). There are
numerous standardization techniques available, four of which have been summarized in Table 1.
Table 1 - Summary of Standardization Techniques: Summary of four common
standardization techniques (after Malczewski 1999)
Standardization Technique Description
Linear scale transformation
Divides the raw attribute values within a
given criterion layer by the layer’s maximum
value for this same attribute.
Value/Utility function approach
Uses input from decision makers to assist in
defining a function that identifies the
relationship between a non-standardized
criterion layer and a standardized criterion
layer.
Probability
Uses probability theory to determine the
likelihood of a given outcome, which is then
used to determine standardized values.
Fuzzy set membership
Process of assigning standardized values
based on a membership function.
Two forms (Equations 1 and 2) of a linear scale transformation technique, the score range
procedure (Malczewski 1999), are explained in more detail below. This is followed by an
explanation of the value/utility function approach.
A linear scale transformation technique, termed the score range procedure, assumes a
linear relationship between non-standardized and standardized criterion attribute values. If the
criteria layer to be standardized is of the benefit-criteria variety, then the following equation may
be used to transform each value x into the standardized value x':
14

(1)
where is a raw value of the non-standardized criterion layer for the attribute ,

is the
minimum value contained within layer a for the attribute b, and

is the maximum value
contained within all criteria layers containing attribute b. If the criteria layer to be standardized is
of the cost-criteria variety, then the following equation may be used to transform each value x
into the standardized value x':

(2)
where x is a raw value of the input layer a for the attribute b, x
b
min
is the minimum value
contained within input layer a for the attribute b, and x
b
max
is the maximum value contained
within all criteria layers containing attribute b. This standardization procedure, implemented with
the use one of these two equations, assumes that the raw input values contain a high value when
the criterion is favorable and low raw input values when the criterion is unfavorable. Thus, when
the reverse is true, the nomenclature defining the variety of criteria to be standardized is switched
making Equation 1 suitable for cost-criteria and Equation 2 suitable for benefit-criteria.
The value function approach to standardization defines the relationship between a non-
standardized criterion layer and a standardized criterion layer based upon the decision makers’
preference of worth (i.e. its value or utility). An example of this is the mid-value method. This
technique requires decision makers to define the value they think best describes the middle value
between two endpoints, where the endpoints are the maximum and minimum values of the non-
standardized criteria layer. Once this value is chosen, it becomes the abscissa of a point in a
15

curve. The ordinate of this same point is the median value of the desired standardized criterion
layer.
When creating a benefit-criterion layer, the maximum non-standardized criterion layer
value would be paired with the desired maximum standardized criterion layer value and the
minimum non-standardized criterion layer value would be paired with the desired minimum
standardized criterion layer value. The opposite is the case for a cost-criterion layer where the
maximum non-standardized criteria layer value would be paired with the desired minimum
standardized criterion layer value and the minimum non-standardized criteria layer value would
be paired with the desired maximum standardized criterion layer value. This work results in three
ordered pairs of numbers. The equation that simultaneously fits each of these points defined by
these pairs is the value function and may be used to standardize the remaining values of the
applicable non-standardized layers.
Standardization algorithms, such as those described above, may be applied to non-
standardized attribute values within a GIS by way of applying the appropriate algorithm to the
non-standardized attribute field within a criterion layer's attribute table (e.g., with the Field
Calculator function of ArcGIS). The ArcGIS Fuzzy Membership tool may also be used to define
the function describing the relationship between the non-standardized and standardized criterion
attribute values. This tool provides various options to describe membership types. One option,
termed fuzzy linear, transforms attribute values using a linear function, yet gives decision makers
the opportunity to define minimum and maximum threshold values. These threshold values may
be used to transform criterion attribute values beyond a certain range to either definitely a
member of an entity group or definitely not a member of an entity group.

16

2.3.4 Allocation of Weights
Following standardization, weights or priorities are derived to indicate the relative
importance of each criterion and/or objective. Voogd (1983) identifies weights as quantitative
values indicating the relative importance of given criterion layers and priorities as ordinal
expressions of their importance. The term weights is used to describe the importance of criteria
and objectives discussed here when ratio attribute values are used to make these comparisons.
There are numerous weighting techniques available (see Table 2). Malczewski (1999) identifies
pair-wise comparison (i.e. AHP) as more appropriate for analysis where accuracy and theoretical
foundations are a concern, whereas ranking and rating systems are appropriate where cost, ease,
and time are a concern.
Table 2 - Summary of Weighting Techniques: Summary of four common weighting
techniques (after Malczewski 1999)
Weighting Technique Description
Ranking
The weighting process of ranking requires the decision maker to
use their preference to place the set of chosen criteria in order
based on their relative importance. Then, numerical weights may
be derived by inserting these ordinal values into a mathematical
formula.
Rating
The weighting process of rating involves the decision maker's
estimate of criteria weights as they relate to a predetermined scale.
Then each criterion is allocated a number of points across a
predetermined scale with a set range, where the collective points
allocated equate to a set number.
Analytical Hierarchy
Process (AHP)
Use of pair-wise comparison to create a matrix, which is then
subject to calculations to derive the right eigenvector of the largest
eigenvalue of this matrix. It is the derived eigenvectors that
become the criterion and objective weights.
Trade-off analysis Assess trade-offs between pairs of alternatives.
The AHP begins by developing a matrix that records the relative importance of each
criterion for each objective. This pair-wise comparison is performed by using an intensity of
17

importance scale such as that developed by Saaty (1980) (Table 3). Every possible pair of criteria
is compared using such a scale to determine relative importance when considering each objective
individually.
Table 3 - Intensity of Importance: Reference used when determining the relative importance of
criteria during pair-wise comparison (after Saaty 1980)
Intensity of Importance Definition
1 Equal importance
2 Equal to moderate importance
3 Moderate importance
4 Moderate to strong importance
5 Strong importance
6 Strong to very strong importance
7 Very strong importance
8 Very to extremely strong importance
9 Extreme importance

The results of these comparisons may be recorded within a table for documentation or
inserted directly into the matrix M shown below.
Criterion
1
Criterion
2
... Criterion
n

= M
Criterion
1

...
.
.

Criterion
2

...
.
.

...
.
.
...
.
.
...
.
.
...
.
.
...
.
.
Criterion
n

...
.
.

where c represents the intensity of importance given to the corresponding criterion 1 through n.
For example, if a criterion, C 1, is compared to another criterion, C 2, where C 1 is moderately
important when compared to C 2, then the following table may be constructed:
18

Table 4 - Pair-Wise Comparison Example
Criteria
Intensity of
Importance
Criteria
Intensity of
Importance
C
1
3 C
2
1
Inserting these values into matrix M results in the value 3/1 for row one, column two (i.e.
C 1/C 2) and 1/3 for row two, column one (i.e. C 2/C 1).
Once a matrix is created for each objective, it is necessary to square the eigenvalues (i.e.
the matrix) until the eigenvectors begin to approach unity. The eigenvectors are the values
resulting from the normalization of the product of matrix multiplication. The approach towards
unity will be reflected in the decimal values of the eigenvectors. Decimal places of decreasing
value will begin to match the previously derived eigenvector. In order to derive weights accurate
to the second decimal place, the matrices must continue to be squared and normalized until the
second decimal place of the normalized eigenvectors remains unchanged. Unless one is using a
GIS that has built in AHP functionality, these steps must be performed outside of the GIS
environment (e.g., within Microsoft Excel) and the results reinserted. An example
implementation of this technique is demonstrated in section 4.6.
2.3.5 Implementation of Aggregation Algorithm
One of the most important components of a GIS is its ability to combine spatial data from
a variety of sources (O’Sullivan and Unwin 2010). The process of aggregation, within the
context of a MCE, produces a layer representing the degree to which the objective(s) has been
met. Several options are available to aggregate various layers. One common and deterministic
approach is a Boolean overlay, which utilizes binary true/false logic. This technique, formalized
by McHarg (1969), involves aggregating multiple layers, each containing values indicating
19

whether or not a particular characteristic is met, to determine what locations meet a set of desired
characteristics.
An alternative to the Boolean overlay, formalized by Malczewski (2000), is termed a
weighted linear combination. This technique differs from the binary logic imposed in the
Boolean overlay. The output values produced from this process indicate, on a graduated scale,
the degree to which a particular objective has been met. This approach is advantageous, as it
retains the metric information contained within the layers being overlain and avoids the
oftentimes illogical assumption that a given criterion no longer has an effect on an objective once
it boundaries have been crossed (O’Sullivan and Unwin 2010).
After all the relevant criteria factoring into a decision making process have been
standardized and once weights for these layers have been determined, these layers may then be
aggregated. There are numerous weighting techniques available, three of which are summarized
in Table 5.
Table 5 - Summary of Aggregation Techniques: Summary of three common aggregation
techniques (after Malczewski 1999)
Aggregation Technique Description
Weighted linear
combination
Takes predetermined weights and multiplies them by
normalized values given to criterion attributes and then
sums the products over all criteria.
Ideal Point methods
Derives values that represent amount of separation from
an ideal value.
Concordance methods
Based on a pair-wise comparison of alternatives and a
mathematical function applied to a matrix derived from
these comparisons. Differs from AHP in that criteria may
only be compared as having preference over another
criteria, but without indication of how much.
The weighted linear combination technique is represented by the following formula:
A = ∑
j
W
j
X
ij

(3)
20

where A is the aggregated layer, W
j
is the weight given to criteria layer j, and x
ij
is the
normalized value for the criteria layer j for attribute i. This process multiplies each criterion
value by its corresponding weight and then sums these new values.
For benefit criteria, the highest values in the aggregated layer represent the most
favorable conditions. When concerned with cost criteria, one may develop cost criteria during
the standardization process or they may subtract the product from the above equation from 1.
This is represented in the following formula:
A = 1 - (∑
j
W
j
X
ij
) (4)
Cost criteria values may have already been incorporated into the criteria layers during the
standardization step. Therefore care should be taken to ensure these values are not reversed
unnecessarily.
2.3.6 Sensitivity Analysis and Analysis of Outcome
The penultimate step of a MCE is an analysis of the outcome from the aggregation step
and/or sensitivity analysis. Such analyses are necessary to determine if errors contained in raw
input data or created during any of the MCE steps have been propagated into the outputs from
the evaluation. Errors may be derived from inaccuracies related to data collection or
manipulation strategies, misrepresentation of real world features, spatial autocorrelation,
modifiable aerial unit problem, scale, edge effects, or other factors.
The most basic approach is to analyze the range of output values from the aggregation
step to explore the reason they contain particular values. This will help build an understanding of
what factors are most important in determining the outcomes. An assessment of errors of
omission and commission (e.g., an error matrix) may be conducted if the error present can be
quantified. Also, one may make small changes to the analysis boundaries or input values and
21

determine if these changes have significant effects on the aggregation output (i.e., the weighted
surface). Additional sensitivity analysis methods are the Monte Carlo simulation and the
analytical error propagation method, which involve changing two or more criteria simultaneously
to assess the effect. If significant effects do occur, a more detailed analysis would need to be
conducted to determine the underlying issue and whether the results are sufficiently stable to be
meaningful.
2.3.7 Using MCE Results to Make Decisions
The end result of the MCE process is an assemblage of choice alternatives. As is the case
with a spatial MCE, these alternatives define locations that are preferred, that should be avoided,
or that act as hard constraints. In the case of a raster GIS environment, each alternative would be
represented as an individual raster cell. One must keep in mind, however, that there is no
“correct” or “best” alternative. As O’Sullivan and Unwin (2010) explain, the results provide one
or more solutions given the standardization, weighting, and aggregation techniques chosen. They
note that this process helps reveal the effects of the various criteria involved and may assist in
reducing the size of the area under investigation.
2.4 Path Modeling
Paths may be modeled to suggest the best way to move between locations (Mitchell
2012). There are two common types of paths that are generated. This includes those that follow a
predetermined network (e.g. transportation network), termed network paths, and those that model
a path between two points, termed overland paths. The former is performed in a vector GIS
environment while the later is performed in a raster GIS environment. The placements of paths
are often determined by associated costs. These costs may be expressed as money, time, distance,
22

and so on. Network costs are associated with edges, intersections, and turns, while overland costs
are associated with raster cells values (Mitchell 2012).
A cost-path analysis may be performed within ArcGIS using a raster to determine the
cost values associated with traveling across particular cells. Using an evaluation process such as
MCE is an example of how such a cost surface, or weighted surface, may be created. The
accumulative cost is calculated on a cell by cell basis by starting at the origin cell and traveling
towards a destination cell, sampling all of its adjacent cells, and recording the value associated
with each edge. Once the cost distance rasters have been generated, they may be used as inputs
to derive a path.

23

CHAPTER THREE: METHODOLOGY
The methodology described here derives traverse paths for fieldwork in a nine-step process.
Most of these steps follow the workflow of a MCE described in the previous chapter. Additional
steps cover the processes of data assembly and construction of evaluation criteria layers. The
final step involves the derivation of the origin and destination points and the traverse paths that
cross these points. The location of the origin and destination points and the traverse paths are
determined by the values contained within the final weighted surface derived by the MCE. These
steps, along with a summary of the work they require and the GIS layer relevant to each of them
for a typical scientific field campaign, are shown below in Table 6. There are many options
available with regard to the specific techniques that may be employed at each step during this
methodology. While a generic framework is presented in this chapter, one must determine which
of these techniques is best suited to the decision problem of a particular field campaign. Chapter
Four provides a demonstration of how this methodology may be performed for a specific, yet
hypothetical, geologic field campaign representing an attempt to develop a traverse for fieldwork
intended to determine whether an area contains ore.
24

Table 6 - Methodology Steps, Required Work, and the Related Layers
Steps Required Work Relevant Layers
1. Identify objectives and
criteria — Problem
definition and criteria
selection steps of MCE
Literature examination,
analytical study, or attainment
of expert opinions
not applicable
2. Assemble relevant data
Acquire data that may be used
to create criteria layers for
measure of objective(s)
fulfillment
e.g., remotely derived
imagery, digital elevation
model, topographic map, and
special-purpose map —
Termed raw data
3. Sketch and derive features
of interest relevant to
meeting objective(s)
Delineate and/or create criteria
layers and ensure they share a
common coordinate system
and extent
Points, lines, polygons, or
continuous data delineating
features of interest — Termed
features of interest
4. Apply necessary
manipulations or analysis to
derived features of interest
For example, apply distance
calculations
Layers containing
manipulation or analysis
results — Termed non-
standardized criteria layer
5. Standardize the non-
standardized criteria layers
— Standardization step of
MCE.
Transform non-standardized
criteria layer values to a
common scale
Termed standardized criteria
layers
6. Establish field campaign
priorities — Weighting step
of MCE.
Derive and assign weights to
each criteria and objective
not applicable
7. Produce weighted surface
layers of study area —
Aggregation step of MCE
Perform map overlay
(aggregation)
Termed weighted surface
8. Review results of MCE —
Sensitivity analysis step of
MCE
Perform sensitivity analyses
and/or analysis of results
not applicable
9. Define traverse paths on
basis of time availability
Derive origin and destination
points and traverse paths
Termed destination points,
and traverse paths

3.1 Identify Objectives and Criteria
This step of the methodology relates to the problem definition and criteria selection steps
of a MCE. As field research will often be conducted to attain data that is unavailable via remote
25

means, the problem facing scientists preparing to go in the field will often be determining what
data should be acquired and how. This problem should be divided into multiple objectives. These
objectives may be, for example, to attain a scientific return and to avoid obstacles that impede
travel across the study area. A group consensus, or individual decision, establishing the overall
decision problem and separating this problem into applicable objectives is required to proceed to
the subsequent steps.
This step also involves deciding which criteria may be used to measure the fulfillment of
the determined objectives. Research may be needed to determine which criteria affect a given
objective in order to establish a scientific foundation for the remaining steps. Analytical studies
or an opinion survey are additional options that may be used to make this decision regarding
appropriate criteria.
Once the set of criteria is determined, they should be separated into factors or constraints.
If a criterion is considered a constraint, the constraining attribute values should be noted. For
example, a criterion layer containing slope values may be considered a constraint where all
values are greater than some threshold, for example, slope values greater than fifty degrees.
It also should be determined whether a given criterion's attribute values will have a favorable or
unfavorable influence on meeting the objective to which it is applicable. This will assist in
determining an appropriate algorithm to use during standardization. While a criterion may
contain attribute values that are not favorable with regards to meeting a particular objective, they
may not necessarily act as a constraint (i.e., a hard limitation).
3.2 Assemble Relevant Data
Once it has been determined which criteria may be used to measure the degree to which a
particular objective(s) is being met, data are sought that may be used to represent these criteria.
26

Each data set used must share a common coordinate system and have positional accuracy
sufficient for the research at hand.
3.3 Sketch and Derive Features of Interest Relevant to Meeting Objective(s)
Raw data that provide information on criteria influencing the degree to which a given
objective(s) is being met will often not be suitable for the remaining steps of this methodology in
its raw form. That is, it may not provide a suitable representation of the criteria or objectives
being evaluated. Such data should be brought into a suitable form through various manipulations
or analyses. Such techniques may include the derivation of slope or visibility from a DEM, or
buffering of features. It may also include the manual delineation of various features of interest
based upon image or map interpretations.
3.4 Apply Necessary Manipulations or Analysis to Derived Features of Interest
The delineated features of interest may be further analyzed so that they contain
information more directly related to measuring the degree to which a particular objective(s) has
been met. An example of this is determining the distance from these delineated features of
interest to all other locations within the study area. This would be relevant to situations where a
scientist's proximity to various features of interest relate to the ability to attain a scientific return
from the features.
3.5 Standardization
This step relates to the standardization step of a MCE. Criteria layers that do not share a
common scale must be converted to a common scale before they may be aggregated. This is
done through the process of standardization. Various standardization techniques are shown in
section 2.4.3. During this process, each layer is transformed into a common scale containing
floating point values ranging from zero to one. The aim of this methodology is to create traverse
27

paths derived from a cost surface, based on cost-criteria. It is fitting therefore, to represent
favorable characteristics with a low value and unfavorable characteristics with a high value. Care
should be taken to ensure these values are not reversed erroneously in the subsequent steps.
3.6 Allocation of Weights
This step relates to the weighting step of a MCE. It must be determined which weight
assessment technique is appropriate for the given decision problem. Various weighting
techniques are shown in section 2.4.4.
3.7 Implementation of Aggregation Algorithm
This step relates to the aggregation step of a MCE. After all the relevant criteria factoring
into the navigational decision making process had been standardized and once weights for these
layers had been determined, these layers may then be aggregated. Various aggregation
techniques are shown in section 2.4.5. If more than one objective is necessary to assess the
decision problem, then multiple weighted surfaces will be created. These surfaces may be
combined to create one final weighted surface to be used during the subsequent steps.
3.8 Review Results of Multicriteria Evaluation
This step relates to the sensitivity analysis and analysis of outcome steps of a MCE. If
appropriate, an error propagation analysis or the construction of an error matrix may be
performed. Otherwise, one should inspect the values contained within the weighted surface to
ensure its values contain the desired meaning. Once confident that these values are reliable, one
may proceed to the subsequent steps.
3.9 Define Traverse Paths on Basis of Time Availability
This step involves deriving the set of points that the traverse path must visit during the
field campaign and the path to be followed between them. Note that this methodology does not
28

describe how to determine visitation sites, but produces suggested traverse paths that position
field scientists within close proximity to features they have deemed to be of interest. The
description below explains how this step may be performed within ArcGIS. In order to derive a
traverse path using this software, users must possess origin and destination points and a weighted
surface. These are used as inputs to the ArcGIS Cost Distance and Cost Path tools. The tools are
run once for each segment of the traverse path. Each iteration of these tools requires the
weighted surface, one origin point, and one destination point. All origin points also act as
destination points and will thus be referred to hereafter as destination points.
Since the lowest values in the weighted surface (i.e. the cost layer) indicate favorability
with regard to meeting a particular objective, this layer is used to determine the destination
points. If time is limited for a particular field campaign, a traverse path may be prioritized by
delineating only the most favorable locations (i.e. those with the lowest values on the weighted
surface). For example, only locations with the top five percent most favorable values of the
weighted layer may be used when a short time duration is available for fieldwork. Once it is
determined which percent to use, the weighted layer is reclassified so that these most favorable
values are represented as some value (e.g., one) while all greater values (i.e., less favorable) are
represented as NoData. This reclassified layer is then converted to a polygon feature class. The
centroids of these polygons are then derived and serve as the destination points. Derived points
that are in close proximity should be manually deleted to avoid excessive calculations that will
not greatly alter the location of the derived traverse path.
In order to use these derived points as individual destination points in the multiple
iterations of the Cost Distance and Cost Path tools, all points in the feature class created above
must be separated into separate feature classes. The sequence in which these points are used in
29

the iterations of the Cost Distance and Cost Path tools will determine the connectivity of the
destination points of the derived traverse path. Thus, these points should be manually ordered
such that the resulting path will contain a logical sequence. For example, in Figure 2, the
destination points have been arranged so that no segments of the traverse path cross. Rather, the
traverse path makes a loop around this portion of the study area. Figure 2 also illustrates that
each point acts as both an origin and destination point. These locations can either contain one
point feature class that acts as both and origin and destination point or contain two point feature
classes with one representing an origin and the other, a destination.

Figure 2 - Example Derivation of Traverse Paths: Example of how the origin and destination
points factor in to the derivation of the traverse paths. Note that the segments between points are
drawn here as simple straight lines, not as final derived traverse paths.
Once the destination points have been derived they may be used in combination with the
weighted surface to complete the traverse path derivation process using the ArcGIS tools
mentioned above. Finally, the ArcGIS Raster to Polyline tool is used to convert the raster output
of the Cost Path tool to a polyline feature class. This will reduce the size of the files representing
30

the traverse path and will convert it to a format that may be easier for use in the field. The former
will assist with bringing the traverse files into the field on a mobile computer. The latter will
enable the traverse paths to be seen more easily seen in combination with a raster (e.g., it may be
overlain on imagery or a topographic map).
Shown in Figure 3 is a portion of a model, built using ArcGIS ModelBuilder, that is used
to derive one segment of the traverse path. In order to make a traverse path with multiple
segments this series of tools will need to be run multiple times. Conducting this work within
ModelBuilder expedites this process.
Figure 3 - Workflow to Create Traverse Path Segment: ArcGIS ModelBuilder steps used to
derive traverse path segment.
In summary, this chapter has outlined a straightforward, generic methodology that may
be used by a field scientist to determine a traverse path in unfamiliar territory. The next chapter
demonstrates the use of the methodology to plan a simple hypothetical geologic field campaign.

31

CHAPTER FOUR: DEMONSTRATION OF METHODOLOGY
The hypothetical field campaign used to demonstrate the traverse generating methodology is
designed to depict preparatory efforts of an exploration geologist (who is arbitrarily considered
to be a woman) who is about to embark on a geologic field survey. Exploration geologists
identify and assess the landscape to determine the likelihood of economically extracting minerals
from a particular area. The geologist here has become aware of a new, more detailed geologic
map for a small area within Hildago County, New Mexico. (This “new map” has been created for
the purpose of demonstrating the methodology described here and does not necessarily represent
the true geology of the area (see Figure 4).) This map depicts an abundance of dikes and sills
(sheet like bodies of rock that cut or follow surrounding rock features) that share characteristics
with other dikes and sills within Hildago County that have been known to contain elevated
amounts of gold. This geologist has been allocated one day to carry out a survey to determine if a
4.5 km
2
portion of this new map area contains evidence that suggests the need for further
exploration (see Figure 5). Her work will consist of documenting the location and a description
of gold bearing rocks or any other potentially economic materials within the study area. In
addition, she will try to record the overall history and distribution of rocks in the study area in
order to supplement the information provided by the geologic map.
As the existence or precise distribution of the gold and its relation to the dikes and sills
are not known for this area, the geologist will attempt to better understand the nature of any gold
deposits and allow this to determine the exact locations she visits. Nevertheless, she must also
attempt to cover a significant portion of the study area in a short time and wants to take a more
analytical and systematic approach in determining how to do so. She has chosen to perform a
MCE in an attempt to place herself near the most advantageous and accessible regions within the
32

defined study area. The goal is to develop a traverse path using the aforementioned methodology
for an approximate ten hour day. This path will place her in close proximity to locations
containing a high scientific return and that are most accessible to foot travel. Described below is
the methodology used to develop this path.
4.1 Identify Objectives and Criteria
There are geologic and logistical objectives that have been determined to be applicable to
the hypothetical field campaign (Table 7). These have been categorized into their suitability to
measure the fulfillment of meeting these objectives through a MCE or path model. A category of
procedural assumptions has also been included, but no attempt is made here to measure their
effect on the target objectives. All of the criteria listed have been derived from a review of
Compton (1985) and Coe (2010).
33

Figure 4 - Hypothetical Geologic Map: Hypothetical geologic map identifying the geologic features that
exist within the study area.
34

Figure 5 - Location Map for Demonstration: Location map illustrating the analysis area used for the demonstration of
the developed methodology. This area includes a buffer around the actual study area and is approximately 5.4 km
2
.
35

Table 7 - Criteria and Procedural Assumptions Relevant to Geologic Fieldwork
Criteria Appropriate for a Multicriteria Evaluation
1. Regions: Regions of interest where opportunities for direct
observation and measurements exist. This criterion differs from
Exposures in that these areas correspond with testing existing
hypotheses.
2. Exposures: Areas where geologic materials are exceptionally exposed
due to mines, road cuts, wash cuts, etc. This criterion differs from
Regions in that this layer only represents locations where relatively
exceptional opportunities for direct observation, measurement, or
sampling exist, but not locations where hypotheses may necessarily
be tested.
3. Visibility: Areas where exceptional distant observation opportunities
exist.
4. Access: Areas containing roads and trails.
5. Slope: Change in elevation per change in distance.
Criteria Appropriate for Path Modeling
6. Time to conduct survey
7. Utility cost of travel across the surface
8. Distance
9. Survey scale
Procedural Assumptions
10. Data availability
11. Property ownership
12. Vegetation
13. Soil cover
14. Weather
15. Equipment
16. Lodging/Camping locations
Criteria one through three are used to measure scientific return, while criteria four
and five are used to measure accessibility. Scientific return, with regards to exploration
geology, is the likelihood of locating and defining potential economically mineable geologic
materials. Criteria one and two are often related, but have been separated here to distinguish
36

the anticipated amount of scientific return associated with each criterion. Time is
incorporated because it will control the number of way points through which the traverse
paths will cross. Cost is incorporated, in that the traverse paths between the selected points
will travel across a weighted cost surface. Of the remaining two criteria, distance is
implicitly considered, in that it is related to cost and scale is incorporated by conducting all
GIS work at a particular scale.
4.2 Assemble Relevant Data
The hypothetical case study presented here uses four authentic data sets and one
hypothetical data set. The process of using these data to derive and sketch the features relevant to
meeting the campaign objectives is described in the next section. These data sets include a New
Mexico Bureau of Geology and Mineral Resources (NMBGMR) state geologic map, aerial
imagery provided by the U.S. National Agriculture Imagery Program, a digital elevation model
from the USGS National Elevation Dataset, and a USGS topographic map. The hypothetical
geologic map was derived to depict a scenario in which a new, more detailed map has peaked the
interest of an exploration geologist.
The NMBGMR state geologic map was acquired from the USGS National Geological
Map Database website. This map was published by the New Mexico Bureau of Geology and
Mineral Resources in 2003 at a scale of 1:500,000. This map was not available for download
directly from the USGS website, but was available only by mail order from NMBGMR offices.
Rather than order this map, with realization that the scale was far more generalized than what
was needed for the case study, a screen shot was taken and then georeferenced using as control
points, state boundaries and roadways contained in the USGS aerial imagery. The total root mean
37

square error for the georeferencing process was approximately 0.84 meters, using an affine
coordinate transformation.
The National Agriculture Imagery Program aerial imagery was acquired from The
National Map Viewer and Download Platform website on 11 October 2014. This imagery has a
spatial resolution of 1.0 meter and was captured between 2011 and 2013.
The USGS National Elevation Dataset (NED) digital elevation model (DEM) was
acquired as a single 10,500 km
2
tile from the USGS National Map Viewer website. It has a
horizontal resolution of 1/3 arc second (which is approximately nine meters at this latitude). The
overall absolute vertical accuracy of the NED within the conterminous US is 1.55 meters and
while this value varies greatly across the US, it does not exceed twenty five meters when
compared to National Geodetic Survey benchmark elevations (Gesch et al. 2014). The various
light detection and ranging (LiDAR), radar, photogrammetric, and topographic data sources used
to create this DEM tile were acquired between 1 February 1999 and 1 November 2013. This
DEM tile was published in 2013.
The Doyle Peak, New Mexico topographic map was acquired from the The National Map
Viewer and Download Platform website on 20 August 2014. It was compiled from aerial
imagery taken in 1976, field checked in 1977, and contains a map scale of 1:24,000. This map
was published in 1982 as a provisional edition map.
Of these datasets, the aerial imagery was used to delineate features of interest pertaining
to Regions, Exposures, and Access. The DEM was used to delineate features of interest
pertaining to Visibility and Slope. The topographic map was used to provide additional support
for the delineation of features of interest pertaining to Exposures and Access. The geologic map
38

was used to support interpretations made from the aerial imagery regarding the delineation of
features pertaining to Regions.
4.3 Sketch and Derive Features Relevant to Meeting Objectives
The hypothetical case study presented here required both delineation of various features
of interest and additional analyses relevant to these features such as distance. These processes are
further described below. Prior to such analyses, however, it was necessary to ensure all raw data
were projected to a common coordinate system, North American Datum 1983 Universal
Transverse Mercator Zone 12. It was also necessary to ensure this data extended across all
portions of the study area. All analyses performed on the data were done for a buffer of
approximately 200 meters beyond the actual study area to mitigate any possible edge effects (a
nonuniformity problem). This area is termed the analysis area and is the rectangular boundary
bounding most figures shown in this thesis. The derivation of slope is the only process that
incorporates neighboring cells into its computation (i.e. bilinear interpolation) and likely the only
layer that may be affected by such a non-uniformity problem. It is assumed that by running each
analysis in this methodology for the extent of the analysis area, and not the study area, any
possible edge effects that may have occurred have been prevented.
This methodology also required that each criteria layer have the same spatial resolution.
With the exception of the NMBGMR geologic map, the elevation data required for the Visibility
and Slope criterion layers had the coarsest resolution at approximately 9.26 meters. The
NMBGMR map, published at a scale of 1:500,000, was only used to gain a broad understanding
of the study area's geology and was only subjected to a brief and visual analysis. Thus, it was not
used to determine the resolution for the subsequent work. A resolution of ten meters was selected
for these analyses due to its similarity to the DEM resolution, its applicability to a 4.5 km
2
39

survey, and for the computer processing requirements involved. Provided below is a description
of how each criterion layer was created.
4.3.1 Derivation of Criterion Layers Representing Scientific Return
The criteria chosen to measure scientific return are Regions, Exposures, and Visibility.
The layers used to represent these criteria not only delineate features of interest, but also the
Euclidean distance, measured in meters, from these features to all other locations within the
analysis area. Where the cells are coincident with the layer’s feature of interest, this value is
zero. As the distance away from these features increases, its value increases linearly. This design
was used in order to account for the advantage of being close to these features of interest
considered to be aids to achieving high scientific return. This is an example of some functionality
that would be unavailable when using the Boolean logic of sieve mapping.
The Regions non-standardized criterion layer represents the distance to areas where the
geologist may test one or more of her current working hypotheses (i.e., those related to
understanding of the distribution and history of the geologic materials within the study area, with
emphasis placed on those materials containing economic interest). The features of interest from
which a measure of distance was calculated were manually delineated based upon information
provided by the aerial imagery and the geologic maps. While sophisticated digital image analysis
is possible and would produce much more precise results, for the purposes of this demonstration,
it was decided that a classic manual interpretation approach was sufficient. Characteristics of
tone, texture, pattern, and shape within the aerial imagery were used to interpret where changing
rock units were likely to occur, what types of rocks were contained in these units, and to
determine structural aspects of the study area. Also, the geologic maps provided a further
explanation of what rock types could be expected to exist in certain portions of the analysis area.
40

This information provided context for determining which locations might provide opportunity to
test hypotheses.
Once the Region’s features of interest had been delineated, the Euclidean distance from
these features was derived. This was done using the ArcGIS Euclidean Distance tool. The input
for this tool was the Regions polygon feature class and the output was a raster with the resolution
set to ten meters. This output raster was used later as an input during the standardization process.
The Exposures criterion represents areas containing features that provide an exceptional
view of the study area rocks. These features include locations such as road cuts, mines, deep
canyons, and perhaps, in non-arid lands where intact rocks are limited, rock outcrops. The only
mines discovered around the study area were approximately 150 meters beyond the analysis area
boundary. Also, no road cuts or deep canyons were found within the study area. To demonstrate
the effect of Exposures on the traverse paths generated, however, three Exposures point features
have been manually added. Once these Exposure’s feature of interest had been delineated, the
Euclidean distance from these features was derived. This was done using the ArcGIS Euclidean
Distance tool. The input for this tool was the Exposures point feature class and the output was a
raster with the resolution set to ten meters. This output raster was used later as an input during
the standardization process.
The Visibility criterion represents areas that provide opportunities to view large portions
of the study area. Visibility was determined using the ArcGIS Visibility tool. This tool derives
visibility by calculating the elevation change between an observation point and the local horizon.
If a center of a raster cell making up a portion of the local horizon is positioned so that there are
no obstructions, or cells with higher elevation values, between it and an observation cell, it is
considered to be visible by that specific observation point. Each cell within the input raster is
41

considered when determining its relation to the observation points. The output from this analysis
is a raster layer representing either the number of times that each cell of the input raster can be
seen by an observer point (the frequency analysis type) or the number of observation points that
can be seen by each cell in the input raster (observation analysis type).
The Visibility analysis used for the methodology described here was of the frequency
analysis type. However, the geologist’s concern was not to determine visibility for just a few
observation points, but from every point across the entire study area. Thus, the centroids for each
raster cell making up the study area were derived and these centroids then became the
observation points used within the visibility analysis. The resulting output raster contained a
value field defining how many cells within the study area could be seen from each observation
point, or from every cell, within the study area. The observation points containing the top thirty
percent of the value results was considered the high visibility area and delineated as such. A
layer containing only the cells with these top values was created by reclassifying the layer. Those
cells with values in the top thirty percent were reclassified to a value of one and the remaining
cells were reclassified to NoData. The Euclidean distance from the visibility features making up
this new reclassified layer was then calculated. This output raster was used later as an input
during the standardization process. Figure 6 depicts these features that were delineated as
Scientific Return criteria in this analysis. An inset in this figure shows the raw values derived
from the Visibility tool.
42

Figure 6 - Features of Interest Relevant to Scientific Return Objective
4.3.2 Derivation of Criterion Layers Representing Accessibility
The criteria chosen to measure impedance to foot travel are Access and Slope. The layer
used to represent the Access criterion not only represents the access to features of interest (i.e.
roads and trails), but also values derived from a distance function indicating a decreasing
favorability value as the distance away from these features increases. The layer used to represent
the Slope criterion contains values of slope in degrees.
43

Aerial imagery and USGS topography maps have been used to identify access routes.
Roads were discovered on the northern and southern edges of the analysis area. The same roads
were identified in both the topographic map and aerial imagery. No trails of significant size were
found within the analysis area. Figure 7 shows the routes identified in this step. Once the Access
features of interest had been delineated, the Euclidean distance from these features was derived.
The input for this tool was the Access polyline feature class and the output was a raster with the
resolution set to ten meters. This output raster was used as an input during the standardization
process.

Figure 7 - Access Features of Interest: Access features of interest relevant to the Accessibility
objective
The Slope criterion represents the slope across the analysis area. The slope within the
analysis area was derived by running the ArcGIS Slope tool. Unlike the other four criteria,
44

Euclidean distance was not involved, as a value of slope was derived for each cell. Slope values
greater than forty nine degrees were consider a constraint. Thus, the slope layer representing this
criterion was reclassified so that all values greater than forty nine degrees were converted to a
value of NoData. This process ensured that the traverse paths generated would not intersect these
areas. This reclassified raster containing the slope and NoData values was used as input during
the standardization process. This layer is shown in Figure 8 below.

Figure 8 - Slope Criterion: Slope values relevant to the Accessibility objective
4.4 Standardization
The non-standardized criteria layers must be converted to a common scale (i.e.,
standardized) before they may be aggregated. The attribute values defining the five non-
standardized criteria layers contain measures of distance and slope. In the standardization step,
45

these layers are transformed into a common scale containing floating point values in a range
from zero to one. The aim of this methodology is to create traverse paths derived from a cost
surface, based on cost-criteria. It is fitting therefore, to represent favorable characteristics with a
low value and unfavorable characteristics with a high value.
The Regions, Exposures, and Visibility non-standardized criteria layers are entirely
composed of distance values. These non-standardized criterion layers are named after the
delineated feature of interest they represent. The distance value within the boundary of these
delineated features of interest is zero. The values of the remaining cells within the non-
standardized criterion layers represent its distance from the given feature of interest within that
layer.
Because these three layers must be compared to layers representing very different values
(i.e. utility for the Access criteria and slope for the Slope criteria), they must be standardized.
There is no information available that quantifies the relationship between the values defined here
for proximity to Regions, Exposures, and Visibility and their ability to assist in meeting the
objective of attaining a high scientific return. Thus, choosing a simple, straightforward approach,
it is assumed here that there is a negative linear relationship between distance from the given
feature of interest contained within these layers and likelihood of attaining a high scientific
return from these features. As the distance from the given layer’s feature of interest increases, its
favorability, or value representing its ability to help meet the objective of scientific return,
decreases.
Equation 1 (Chapter Two) was used to standardize the non-standardized criteria layers of
Regions, Exposures, and Visibility. This equation was run separately for each of these three
layers and calculated for each ten square meter cell within each layer. This function is shown
46

graphically in Figure 9. The minimum distance value (i.e. zero) within these layers was assigned
a standardized value of zero. The greatest possible distance value given the defined analysis area
(i.e. approximately 3319 meters) was assigned a standardized value of one. The intermediate
distance values increased linearly between these two endpoints.

Figure 9 - Standardization Function for Regions, Access, and Visibility Criteria: Graphical
form of the function used to describe the relationship between distance to Regions, Exposures,
and Visibility features of interest and their anticipated contribution to meeting the Scientific
Return Objective.
The Access criterion has been standardized using the value function approach described
in Section 2.3.3. This differs from the simple linear scale transformation in that it is not based
entirely upon a linear relationship between the non-standardized and standardized criteria layers.
No research quantifies the relationship between distance to access routes and its ability to
support accessibility. Thus, consideration of time it would take to travel to a delineated access
route was used to determine the linear function used to standardize this layer. At 3.2 kilometers
per hour it would take five minutes to walk approximately 268 meters. This was assumed to be
the distance at which it no longer makes sense to travel to an access route rather than directly to
the next destination. A decay function is used to represent this relationship. This function is
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 664 1328 1991 2655 3319
Standardized Value
Approximate Distance (meters)
47

shown graphically in Figure 10. Similar to the value function for the Regions, Exposures, and
Visibility layers, the minimum distance value (i.e. zero meters) within the Access layer was
assigned a standardized value of zero. The standardized values increased linearly as distance
values increased, until a distance value of 268 meters was reached. All distance values between
268 meters and the maximum possible distance value (i.e. 3319 meters) were assigned a
standardized value of one. The maximum value, however, was set to 268 meters rather than 3319
meters. This resulted in all values above this maximum to receive a standardized value of one. It
is important to note that since locations beyond the 268 m distance may still be incorporated into
the path if other criteria make it worthwhile to go there, they have not been set to NoData as is
described below for slope.

Figure 10 - Standardization Function for Access Criterion: Graphical form of the function
used to describe the relationship between the location of Access features of interest and their
anticipated contribution to meeting the Accessibility Objective.
There is no information available that quantifies the relationship between slope and its
detriment to accessibility. Thus, once again, it is assumed that there is a negative linear
relationship between an increase in slope and accessibility for foot travel. This occurs until a
48

value of fifty degrees is reached, at which point all higher slope values are represented as
NoData. As the slope increases, its value representing its ability to help meet the objective of
accessibility decreases. Equation 1 defines this relationship and was used to derive the
standardized criterion layer for Slope. Figure 11 shows the standardized layers for each of the
five criteria.

Figure 11 - Derivation of Five Criteria: Illustration and description of the formation of the
standardized criteria layers where lighter shades of green indicate a greater favorability with
regard to meeting the objective of either attaining a high scientific return or increasing
accessibility
49

4.5 Establish Field Campaign Weights
The analytical hierarchy process was chosen to determine the weights for the hypothetical
field campaign. As described in Section 2.3.4, pair-wise comparisons may be made among all
criteria affecting a given objective and between objectives. For the purposes of this
demonstration, three illustrative pair-wise comparisons were proposed using the intensity of
importance scale shown in Table 3 above. Results of this comparison are shown in Table 8
below. To assign these values, the less important item in each pair is given a weight of one and
the other is assigned a relative importance value determined by assessing the level of greater
importance.
Table 8 - Pair-Wise Comparison Results: Results of the pair-wise comparison for each criteria
using the intensity of importance definitions outlined in Table 2. Truncated intensity of
importance definitions are 1 = Equal Importance, 3 = Moderate importance, 5 = Strong
Importance, 7 = Very strong importance, and 9 = Extreme Importance
Comparison of Criteria Affecting Scientific Return
6 Regions over Exposures 1
7 Regions over Visibility 1
4 Exposures over Visibility 1
Comparison of Criteria Affecting Accessibility
4 Slope over Access 1
Comparison of Objectives
8 Scientific Return over Accessibility 1

As shown in Table 8, Regions is considered to have a strong to very strong importance
when compared to Exposures. Regions is considered to have a very strong importance when
compared to Visibility, and so on. Once all comparisons had been made, their values were
inserted into matrix M, which, for example, led to the derivation of the following matrix for the
Scientific Return objective shown in Table 9.
50

Table 9 - Example AHP Matrix
Regions Exposure Visibility
Regions

Exposures

Visibility

Once the matrices had been created, it was necessary to derive the eigenvectors of the
largest eigenvalues. This was done by squaring each of the three matrices until the second
significant decimal point of the normalized eigenvectors remained unchanged. This allowed for
the derivation of a weighted value accurate up to two decimal points. For example, the work used
to derive the right eigenvector of the largest eigenvalue for the Scientific Return objective is as
follows:

Adding the sums of each of the product rows and normalizing produces the approximate
eigenvector of:

The new eigenvalues are squared again and produce the approximate values shown below:

Adding the sum of each of the new product rows and normalizing produces the approximate
eigenvector of:
51

Note the values of this eigenvector differs slightly from that shown in the first eigenvector (i.e.
the second decimal place in top and bottom rows and the third decimal place in the middle row).
The eigenvectors will approach unity as their eigenvalues continue to be squared. Again, the
newest eigenvalues are squared and produce the approximate values shown below:

Adding the sum of each of the new product rows and normalizing produces the approximate
eigenvector of:

It can be seen that the second decimal place in this last eigenvector remains unchanged. A
weight, up to two decimal places, may now be used. The first row of the original matrix
represents the Regions layer, the second represents the Exposures layers, and the third the
Visibility layer. Thus, the values extracted from this process is 0.74 for the Regions layer, 0.19
for the Exposures layer (rounded up), and 0.07 for the Visibility layer. This same process was
performed for the Accessibility and Combination objectives. These results are shown in Table
10.
52

Table 10 - Derived Weights: Weights for each criterion and objective resulting from the
analytical hierarchy process.
Criteria for Scientific Return Weight
Region 0.74
Exposures 0.19
Visibility 0.070
Criteria for Accessibility Weight
Slope 0.80
Access 0.20
Objective Weight
Scientific Return 0.89
Accessibility 0.11
Unless one is using a GIS that has built in AHP functionality, these steps must be
performed outside of the GIS environment and the results reinserted. The AHP work conducted
to derive the weights for the hypothetical field campaign criteria and objectives was performed
using Microsoft Excel.
4.6 Production of Weighted Overlay Layers
After all the relevant criteria factoring into the navigational decision making process had
been standardized and once weights for these layers had been determined, these layers were then
aggregated. As the cost criteria values were incorporated into the criteria layers during the
standardization step, Equation 5 has been used for the analysis presented here. Had the cost not
been incorporated up to this point, Equation 6 could be used to do so. Care should be taken,
however, to ensure these values are not reversed unnecessarily.
Three weighted overlay layers were created to represent: the Scientific Return objective,
the Accessibility objective, and the Combination objective that represents the combination of
these two objectives, or the ultimate goal of the field campaign. The two former maps were
created by aggregating the appropriate criteria and the later by aggregating the appropriate
53

objectives. A weighted linear combination was used to aggregate these layers with the weights
determined through the AHP. These results are shown in Figure 12.

Figure 12 - Outputs from Aggregation Steps: The outputs from the three aggregation steps.
Notice the NoData cells contained within the Accessibility objective layer carried through to the
Combination layer and that the Combination layer has characteristics of both the Scientific
Return and Accessibility layers. Also, note that lightest shaded of green represent the most
favorable locations for a traverse path.
54

4.7 Prioritize Target Paths for Fieldwork on Basis of Time Available
For the hypothetical case study, those cells from the aggregation step of the MCE with
the most favorable ten percent value were used to determine the origin and destination points for
the day's traverse. As described in section 3.9, these points, along with a weighted surface,
determine the placement of traverse paths.
These points were created by reclassifying the MCE output so that all but the most
favorable ten percent values contained a value of NoData. This reclassified raster was then
converted to a polygon feature class, thus making polygons in those areas containing the top ten
percent favorability values. From this new layer, the polygon centroids were derived. In order to
avoid the extra work of creating more origin and destination points and small traverse path
segments, centroids that were clumped within groups less than seventy five meters apart had all
but one centroid manually deleted, a task that was relatively simple and did not require the use of
an automated tool. Lastly, new point feature classes were created to act as origins and
destinations from these centroid locations. Each centroid is both an origin and destination feature
class, as each of these points acts as an origin and destination point for multiple traverse path
segments. Determining the order in which these segments connected was done manually with an
attempt create segments that would decrease the overall length of the traverse.
Once all of the origin and destination points were derived, they were used as inputs to the
ArcGIS Cost Distance and Cost Path tools. As the output of the Cost Path tool is a raster, the
Raster to Polyline tool was used to create a polyline feature class. This series of tools was run
fourteen times within ModelBuilder. One for each segment within the traverse path and while
using different origin and destination points.
55

A flowchart of the entire process is shown in Figure 13. By following the paths derived
from the output of the aggregation step of the MCE, geologists will be positioned in locations
where they are close to features containing a high scientific return while avoiding locations that
contain impedance to foot travel. The decisions made at each step up to this point determine the
derived traverse paths. These include the decisions made with regard to criteria selection,
standardization technique, weighting method, aggregation algorithm, and point derivation.

Figure 13 - Flowchart of GIS Work: Shown here is a complete flowchart of the implemented
methodology. Layers are represented by rectangles, GIS operations by green ellipses, and order
of operations by arrows.
56

4.8 Derived traverse paths
There are four controls determining the placement of the generated traverse paths. These
include the cell values contained within the Combination weighted surface resulting from the
MCE, the origin and destination points that have been derived based upon the areas containing
the ten percent most favorable values, the order in which the origin and destination points are
created, and the determination the point at which the traverse path begins. A flow chart
summarizing the development of the traverse paths is shown in Figure 14. The resulting traverse
paths for the hypothetical field campaign are shown in Figure 15. A statistical summary of the
traverse path is provided in Table 11.
57

Figure 14 - Flowchart of Aggregation Steps: Illustrated flowchart of the aggregation steps
involved in the developed methodology. Equation 4 is used to aggregate the three criteria
representing the Scientific Return objective and to aggregate the two objectives representing the
Accessibility objective. These two objectives are then aggregated using the same equation to
derive the final weighted surface used to derive the traverse paths.
58

Figure 15 - Final Traverse Path: Final traverse path overlain on aerial imagery, criteria
affecting the Scientific Return objective, and destination points, and the polygons making up the
areas containing the top most favorable ten percent values of the Combination weighted surface.
Table 11 - Statistical Summary of Traverse Path
Traverse Path and Return Path combined
Distance 7.39 kilometers
Maximum slope 23.8 degrees
Average slope 2.70 degrees
4.9 Analysis of Results and Sensitivity Analysis
As outlined in the background chapter, a sensitivity analysis may involve making small
adjustments to the values of the MCE input data and analyzing what effects it has on the derived
output. A sensitivity analysis may also involve an analysis of the output data without making any
59

adjustments. Both of these analyses have been performed here to assess what effect the MCE
based methodology employed here has had on the final recommended traverse paths.
4.9.1 Analysis of Result from Aggregation
To assess what determines the cell values of the Combination weighted surface, its values
at four point locations have been analyzed. The locations of these four sample points were
chosen to include some of the highest and lowest values contained within the Combination
weighted surface. An assessment of these values reveals why the Combination weighted surface
contains the values it contains. This, in turn, demonstrates what has determined the placement of
the traverse path. The cell values of the Combination weighted surface have a maximum of 0.24,
minimum of 0.024, mean of 0.095, and median of 0.13. Figure 16 shows the locations of theses
sample points within the analysis area.
60

Figure 16 - Sample Point Analysis: Shown here are the sample analysis points in relation to the
five criteria features of interest that are the most favorable ten percent areas from the
Combination weighted surface that have been selected to derive the destination points. Elevation
contours have been shown in replacement of the slope raster data model.
Location A is within a location where the weighted surface contained some of the least
favorable values (i.e., approximately 0.22). Location B is within a location that contained some
of the most favorable values (i.e., around approximately 0.039). While location A is located in an
area with almost no slope, it is distant from Regions, Exposures, and Visibility features of
interest. Location B is within a Regions feature of interest, less than 350 meters from two
Visibility features of interest, and within an area of approximately three degrees slope. The
61

generated traverse path, if followed, would put geologists within close proximity to locations
such as Location B. Thus, putting them in close proximity to areas interpreted as providing a
high scientific return and that contain increased accessibility.
A comparison can also be made between locations C and D. Location C has a score of
approximately 0.062 and D of approximately 0.036. They both are within a Regions boundary
and have a Slope value of one degree, yet the distances to Exposures, Visibility, and Access
features are all less for D than they are for C. This is the reason D is within the top five percent
and C is not. The reveals that the placement of the traverse path will be closer to locations such
as Location C, rather Location D.
A comparison can also be made between locations C and D. Location C has a score of
approximately 0.062 and D of approximately 0.036. They both are within a Regions boundary
and have a Slope value of one degree, yet the distances to Exposures, Visibility, and Access
features are all less for D than they are for C. This is the reason D is within the top five percent
and C is not. The reveals that the placement of the traverse path will be closer to locations such
as Location C, rather Location D.
4.9.2 Change in Weights for Scientific Return and Accessibility
Figure 17 and Figure 18 show the variation resulting from altering the weights between
the Scientific Return and Accessibility objectives. Figure 17 shows a contour map of the
weighted surface given three different relative weightings. The weights for the two objectives
have been altered up and down by two percent around the weights determined for these
objectives by the AHP. It can be seen that this weight change results in a change in the cell value
distribution, although less so around the perimeter of the analysis area. Figure 18 shows the
various origin and destination points derived from the point extraction process with the same
62

slightly altered weights. The positioning of these points change due to change in the location of
the centroids when the boundaries of the polygons with the top ten percent favorability values
are delineated.

Figure 17 - Change in Weights of Influence: Variation resulting from changing the weights
above and below that derived by the analytical hierarchy process (i.e. 0.89 for Scientific Return
and 0.11 for Accessibility). Each weighted surface was classified into five identical classes and
the boundaries of these classified areas (which are similar to contour lines) were overlain.
63

Figure 18 - Variation in Centroids Due to Change in Weights of Influence: Variation in
centroids used to determine origin and destination points derived from the same weighted
surfaces described in Figure 17.
4.9.3 Change in Percent used to Derive Origin and Destination Points
The traverse paths outlined above travel among points defined by the most favorable 10
percent cell values of the weighted overlay produced by the MCE. These points are the centroids
of the polygons formed by the areas with these most favorable values. To assess the affects of
changing the value used to derive the centroids, the most favorable 15 percent cell values have
been used to derive centroids. This allows comparison of the origin and destination points that
64

would be derived based upon a variation in the percent of favorability values used to derive
polygons and their centroids. Figure 19 shows the results of this comparison. The cutoff value
for the top ten percent is approximately 0.046. Examined here, is the affect of changing this
cutoff value to the most favorable fifteen percent, which is approximately 0.056.
It can be seen that this five percent change produces a different number of centroids and
positions them at different locations. There are fourteen centroids created when using the ten
percent most favorable values and sixteen created when this value is raised to fifteen percent.
Increasing the amount of favorability values used to derive origin and destination points extends
the reach of the traverse path. The percent favorability values used to derive origin and
destination points are based on time. This enables the length of the generated traverse to
accommodate for the time available for a geologic survey. However, the number of polygons
will not necessarily continue to increase as the percent of favorable cell values from the weighted
surface is increased. Eventually large polygons are created and the number of centroids actually
decreases. This will occur to the point where only one centroid is created and thus no traverse
paths may be derived. Applying a limit on to the percent favorability values used may mitigate
this occurrence.
65

Figure 19 - Change in Destination Location Resulting from Percent Favorability Used:
Areas representing the most favorable ten and fifteen percent values from the Combination
weighted surface, along with centroids to be used for the origin and destination points. Note that
all polygons do not contain centroids, as some may have been deleted due to their proximity to
other centroids. Also, the representative center of a polygon feature may lie outside the polygon
perimeter.

In summary, this chapter has demonstrated the steps involved in developing a traverse
path for a mineral exploration field campaign using the methodology described in Chapter Three.
This chapter has also presented a traverse path that has been produced as the result of following
these steps. Lastly, aspects of the methodology determining the location of the destination points
and the traverse path segments between these points has been analyzed.
66

CHAPTER FIVE: DISCUSSION AND CONCLUSION
This chapter discusses some of this methodology’s benefits, limitations, and suggestions for its
improvement. It then concludes with a summary of this work.
5.1 Benefits Gained by Using this Methodology
The methodology described in this thesis allows field scientists to obtain benefits of the
MCE and path modeling processes. It also utilizes a GIS’s ability to manage, manipulate, and
analyze spatial data. Of these three, a GIS is the only tool that has been extensively adopted by
the field research community.
Voogd (1983) explains that a MCE provides an opportunity to classify a problem and
allows an examination of the form, controls, and cost of a decision making process. Field
scientists are driven by a variety of objectives, but all are faced with the problem of determining
where to go in the field. The MCE process divides this problem into its various components.
These components are then assessed to determine their form and control on the decision making
process. Classifying the problem has been the first component of the MCE process demonstrated
here. For the hypothetical field campaign, the decision problem was to determine where to go
while in the field in order to fulfill the objectives of attaining a high scientific return while
avoiding obstacles. Five criteria were used to measure how well these objectives were being met.
The process of characterizing these criteria required each to be considered both individually (i.e.
with the analysis of digital data and derivation of criteria layers) and collectively (i.e. with the
pair-wise comparison and aggregation step). The developed methodology also made possible an
examination of form, control, and cost of the decision making process by explicitly defining each
criteria and objective, considering their relative importance, and assessing their overall
contribution to meeting the campaign objectives.
67

The advantages of using path modeling include its suitability to analytically address the
problem of field navigation. The final weighted surface produced by the MCE process explicitly
indicates which areas within the analysis area will best account for a campaign's objectives. The
traditional fieldwork approach would involve a more intuitive and iterative determination of
these areas. Path modeling has also provided a means to incorporate considerations of distance
and time into the field navigational decision making process. The largest control on the traverse
path placement are the values within the final weighted surface produced during the MCE. Also,
the path modeling process accounts for distance and, by proxy, time by determining direct routes
in between origin and destination points.
Lastly, this methodology requires scientists to thoroughly investigate a study area prior to
its visitation. The steps required to complete this methodology leads scientists through work that
has the potential to elicit valuable new information. This information may then be used as an
advantage when the actual fieldwork begins.
5.2 Limitations of Methodology and Suggestions for Improvement
While there are many advantages to the methodology described here, there are several
limitations that may impede its usefulness. Many of the disadvantages of the MCE process relate
to its complexity and the lack of a simplified framework for its implementation. While the
process of deriving destination points and traverse paths is quite simple, it too lacks a framework
that allows a quick and easy implementation. ArcGIS ModelBuilder was used to expedite many
of the techniques performed while developing the traverse paths. The development of a program
to automate the entire process, however, would make this methodology better suited to
widespread use. This methodology also requires the repetition of many of its components once
68

new parameters are introduced. For example, introducing new objectives, changing feature
boundaries, or changing weights would all result in the need to repeat many steps.
This methodology also lacks the ability to quickly incorporate the information provided
by new observations made while in the field. An ultimate goal is the development of this
methodology so that it may be employed while in the field to quickly generate new traverses
once new observations related to a campaign's objectives have been made. Thus, the traverse
would continually adapt to a scientist's understanding of a study area. This, of course, would
require the development of a streamlined program that would allow the quick and easy
incorporation of new information into this model. The methodology developed here provides an
example of a workflow that may be developed into such a model or into a mobile application.
During the standardization steps of the methodology demonstration, many assumptions
were made regarding the relationship between the non-standardized and standardized attributes
values of these criteria. While quantification of these relationships may be impractical and expert
opinion may prove to define an accurate association, misinterpretations made during this step
will be carried through the subsequent steps. This may result in the development of an inferior
traverse path. Research that quantifies the impact various criteria have on meeting a field
campaign's objectives will likely increase the chances that a useful traverse path is generated.
An additional limitation relates to the assumptions made when using the Visibility
criterion during the hypothetical demonstration. While the amount of study area observable from
any given location has been determined, this view may not have provided much geologic insight.
Also, the ArcGIS Visibility tool assumes that the observer will be able to see the landscape being
viewed even if some portions of the study area are so distant from the observer that a beneficial
view cannot be obtained. Improvements to this component may be made by developing a means
69

to quantify the distance threshold from which beneficial observations may be made and
determine whether a particular view displays science rich or monotonous scenes of a landscape.
The methodology described here only accounts for travel on foot. Many field scientists
will carry out their work using a combination of foot and vehicular travel. As these differing
modes of transportation will have differing criteria providing a measure of their effect on
accessibility, using other forms of transportation will require a modified methodology. Thus,
modifications to this methodology may involve devising a traverse-generating technique that
compensates for a combination of foot and vehicular travel.
During the hypothetical demonstration another assumption was made regarding the
selection of the top ten percent most favorable cell values to determine the destination points.
Determining what this value should be, given the time available for a particular field campaign,
will require further research. Also, an inherent characteristic of increasing the percent of
favorable cell values used is that eventually the polygons representing these areas will coalesce
into one large polygon. The next step of deriving the centroid of this polygon will result in one
destination point. This will not suffice in allowing the derivation of traverse paths. Further
development is needed regarding the means by which destination points are derived. One
potential solution is to use a MCE-derived weighted surface to suggest a traverse path between a
series of predetermined visitation sites. These sites could be manually determined by geologists
and chosen based on their anticipated ability to assist meeting the given campaign objectives.
70

5.3 Conclusion and Opportunities for Future Work
This thesis has demonstrated how technologies and analytical approaches such as
decision analysis, path modeling, and geographic information systems may offer assistance to
navigational decision making while conducting scientific fieldwork. It demonstrates an
alternative approach to planning traditional fieldwork and makes explicit key aspects of the
navigational decision making process. While the intuitive and artistic aspect of field research will
likely always remain, this work demonstrates the value of utilizing technologies that can provide
meaningful assistance to its practice.

71

REFERENCES
Bolstad, P. 2012. GIS Fundamentals: A First Text on Geographic Information Systems. White
Bear Lake, MN: Eider Press.

Carr, C. E., K. V. Hodges, D. J. Newman. 2003. Geologic Traverse Planning for Planetary EVA.
AIAA and SAE International Conference on Environmental Systems (ICES 2003)
Vancouver , B.C., Canada, July 2003.

Carver, S. J. 1991. Integrating multi-criteria evaluation with geographical information systems.
International Journal of Geographical Information Systems. 5, 321–339.

2008. Multicriteria Evaluation. In Encyclopedia of Geographic Information Science,
ed. K. K. Kemp. Los Angeles, CA: SAGE Publications, 291-294.

Chamberlin, T. C., 1890. The method of multiple working hypotheses. Science. 15 (366): 92–96

Clark, M. H., and M. S. Duffy. 2006. Soil Survey of Denali National Park Area, Alaska. Natural
Resources Conservation Service and United States Department of Agriculture.

Compton, R. R. 1985. Geology in the field. New York, NY: John Wiley & Sons.

Coe, A. L. 2010. Geological Field Techniques. Hoboken, NJ: Wiley-Blackwell.

Ernst, W. G. 2006. Geologic mapping-Where the rubber meets the road. Geological Society of
America Special Papers, 413, 13-28.

Gesch, D. B., M.J. Oimoen, and G.A. Evans. 2014. Accuracy assessment of the U.S. Geological
Survey National Elevation Dataset, and comparison with other large-area elevation
datasets—SRTM and ASTER. U.S. Geological Survey Open-File Report 2014–1008.

Goodchild, M. F. 1977. An evaluation of lattice solutions to the problem of corridor location.
Environment and Planning. 9: 727-738. Great Britain: Pion.

Hӧrz, F., G. E. Lofgren, J. E. Gruener, D. B. Eppler, J. A. Skinner, C. M. Fortezzo, J. S. Graf, W.
J. Bluethmann, M. A. Seibert, and E. R. Bell. 2013. The traverse planning process for D-
RATS. 2010. Acta Astronautica. 90 (2): 254-267.

House, P. K., R. Clark, and J. Kopera. 2013. Overcoming the momentum of anachronism:
American geologic mapping in a twenty-first-century world. In Geological Society of
America Special Papers, ed. V. R. Baker. 502, 103-125.

Johnson, A., J. Hoffman, D. Newman, E. Mazarico, and M. Zuber. 2010. An Integrated Traverse
Planner and Analysis Tool for Planetary Exploration. In AIAA SPACE 2010 Conference
& Exposition, 30 August - 2 September 2010, Anaheim, California. 1-28. Reston, VA:
American Institute of Aeronautics and Astronautics.
72

Keeney, R. L., and H. Raiffa. 1976. Decisions with Multiple Objectives: Preferences and Value
Tradeoffs. New York, NY: Wiley.

Mars, J. C. 2013. Hydrothermal alteration maps of the central and southern Basin and Range
province of the United States compiled from Advanced Spaceborne Thermal Emission
and Reflection Radiometer (ASTER) data (ver. 1.1, April 8, 2014). U.S. Geological
Survey Open-File Report 2013–1139.

Malczewski, J. 1999. GIS and Multicriteria Decision Analysis. New York, NY: Wiley & Sons.

2000. On the use of weighted linear combination method in GIS: common and
best practice approaches. Transactions in GIS, 4 (1):5–22.

McHarg, I. 1969. Design with Nature. New York, NY: Natural History Press.

Mitchell, A. 2012. The Esri guide to GIS analysis: Modeling Suitability, Movement, and
Interaction. Redlands, CA: ESRI Press.

O'Sullivan, D., and D. Unwin. 2010. Geographic Information Analysis. Hoboken, NJ: Wiley &
Sons.

Passchier, C.W., and U. Exner. 2010. Digital mapping in structural geology - Examples from
Namibia and Greece. Journal of the Geological Society of India. 75 (1): 32-42.

Riggs, E. M., C. C. Lieder, and R. Balliet. 2009. Geologic Problem Solving in the Field:
Analysis of Field Navigation and Mapping by Advanced Undergraduates. Journal of
Geoscience Education. 57 (1): 48-63.

Ross, A., J. Kosmo, and B. Janoiko. 2013. Historical Synopses of Desert RATS 1997-2010 and a
Preview of Desert RATS 2011. Acta Astronautica. 90 (2): 182-202.

Saaty, T. 1980. The analytic hierarchy process. New York, NY: McGraw-Hill.

Skinner, J.A., and C.M. Fortezzo. 2013. The Role of Photogeologic Mapping in Traverse
Planning: Lessons from DRATS 2010 Activities. Acta Astronautica. 90 (2): 242-253.

Store, R., and H.Antikainen. 2010. Using GIS-Based Multicriteria Evaluation and Path
Optimization for Effective Forest Field Inventory. Computers, Environment and Urban
Systems. 34 (2): 153-161.

Voogd, H. 1983. Multicriteria Evaluation for Urban and Regional Planning. London, England:
Pion.

Abstract (if available)

Abstract Field research is a necessary component of many realms of ecological and geoscientific practice since it provides the primary data crucial to understand the characteristics of an object, phenomenon, or process. Unlike work in an office or laboratory, fieldwork has additional cost related to travel, lodging, and per diem expenses. Field scientists must therefore ensure they make efficient and effective field navigational decisions that result in expedient execution of field campaign objectives. ❧ Technologies and analytical approaches such as decision analysis, path modeling, and geographic information systems offer assistance to navigational decision making while in the field as do the analytical techniques of weighted linear combination and analytical hierarchy process. These tools are often underutilized, however. This thesis describes a methodology by which these technologies and analytical procedures may assist field scientists with navigational decision making. Specifically, the thesis documents development of a model that uses a spatial multicriteria decision evaluation to derive favorability values. These values are then used to determine the placement of traverse paths that are suggested routes to be taken by field researchers. The thesis includes a description of concepts behind the methodology, a demonstration of the methodology for a hypothetical geologic campaign, and an analysis of resulting traverse paths.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Geographic information systems eelgrass (Zostera marina) habitat restoration suitability model, Long Island Sound, USA - a ‘sound-wide’ model

PDF

A comparison of GLM, GAM, and GWR modeling of fish distribution and abundance in Lake Ontario

PDF

Using volunteered geographic information to model blue whale foraging habitat, Southern California Bight

PDF

Modeling prehistoric paths in Bronze Age Northeast England

PDF

A threat-based least-cost path decision support model for national security resource allocation along the US-Mexico border

PDF

Modeling geopolitics in Tikal through least cost paths

PDF

Spatial distribution of the Nile crocodile (Crocodylus niloticus) in the Mariarano River system, Northwestern Madagascar

PDF

Testing LANDIS-II to stochastically model spatially abstract vegetation trends in the contiguous United States

PDF

Building better species distribution models with machine learning: assessing the role of covariate scale and tuning in Maxent models

PDF

Using pattern oriented modeling to design and validate spatial models: a case study in agent-based modeling

PDF

Investigation and analysis of land use/land cover to determine the impact of policy developments in cities

PDF

Validation of volunteered geographic information quality components for incidents of law enforcement use of force

PDF

Out-of-school suspensions by home neighborhood: a spatial analysis of student suspensions in the San Bernardino City Unified School District

PDF

Modeling burn probability: a Maxent approach to estimating California's wildfire potential

PDF

Operational optimization model for Hungry Marketplace using geographic information systems

PDF

Applying least cost path analysis to search and rescue data: a case study in Yosemite National Park

PDF

Identifying areas of high risk for avian mortality by performing a least accumulated-cost analysis

PDF

Effect of spatial patterns on sampling design performance in a vegetation map accuracy assessment

PDF

Generating bicyclist counts using volunteered and professional geographic information through a mobile application

PDF

Generating trail conditions using user contributed data through a web application

Asset Metadata

Creator Reeves, Ryan Richardson (author)

Core Title Deriving traverse paths for scientific fieldwork with multicriteria evaluation and path modeling in a geographic information system

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Publication Date 01/27/2015

Defense Date 01/26/2015

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag field research,fieldwork,geographic information system,multicriteria evaluation,OAI-PMH Harvest,path modeling,traverse path

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Kemp, Karen K. (committee chair), House, Kyle (committee member), Longcore, Travis R. (committee member), Swift, Jennifer N. (committee member)

Creator Email rrreeves@usc.edu,ryan-reeves@hotmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-524912

Unique identifier UC11298293

Identifier etd-ReevesRyan-3135.pdf (filename),usctheses-c3-524912 (legacy record id)

Legacy Identifier etd-ReevesRyan-3135.pdf

Dmrecord 524912

Document Type Thesis

Format application/pdf (imt)

Rights Reeves, Ryan Richardson

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA