University of Southern California Dissertations and Theses

A comparison of weighted and fuzzy overlays in mapping landslide susceptibility, south-central front range, Colorado

(USC Thesis Other)

A comparison of weighted and fuzzy overlays in mapping landslide susceptibility, south-central front range, Colorado

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

A Comparison of Weighted and Fuzzy Overlays in Mapping Landslide Susceptibility,
South-Central Front Range, Colorado

by

Patricia L. Lee

A Thesis Presented to the
FACULTY OF THE USC DORNSIFE COLLEGE OF LETTERS, ARTS AND SCIENCES
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF SCIENCE
(GEOGRAPHIC INFORMATION SCIENCE AND TECHNOLOGY)

May 2023

Copyright © 2023 Patricia L. Lee

ii

Dedication
To my family, who have always had my best interests at heart

iii

Acknowledgements
My gratitude goes to Dr. Sedano for ensuring I did not stray too far into the weeds and for
guidance when I needed it. I am also grateful to my work wife Dr. Caekaert for bouncing ideas
with me whenever I need a sounding board and giving me sound advice. To my dearest friends
and my classmates turned friends, thank you for your help and support.

iv

Table of Contents
Dedication ....................................................................................................................................... ii
Acknowledgements ........................................................................................................................ iii
List of Tables ................................................................................................................................ vii
List of Figures .............................................................................................................................. viii
Abbreviations .................................................................................................................................. x
Abstract ......................................................................................................................................... xii
Chapter 1 Introduction .................................................................................................................... 1
1.1. Background .........................................................................................................................1
1.2. Study Area ..........................................................................................................................2
1.3. Motivation ...........................................................................................................................4
1.4. Importance ..........................................................................................................................5
1.5. Overview .............................................................................................................................7
Chapter 2 Related Work.................................................................................................................. 8
2.1. Regional Setting ..................................................................................................................8
2.2. Landslide Overview ..........................................................................................................10
2.2.1. Landslide Anatomy ..................................................................................................11
2.2.2. Landslide Triggers and Characterization .................................................................11
2.3. Criteria Employed in Landslide Prediction .......................................................................13
2.3.1. Topography ..............................................................................................................20
2.3.2. Hydrology ................................................................................................................24
2.3.3. Subsurface ................................................................................................................27
2.3.4. Surface .....................................................................................................................30
2.4. Modeling Methods ............................................................................................................33
2.4.1. Multiple-Criteria Decision Analysis ........................................................................34

v

2.4.2. Logistic Regression ..................................................................................................37
2.4.3. Other Methods .........................................................................................................38
Chapter 3 Methodology ................................................................................................................ 40
3.1. Research Design................................................................................................................40
3.2. Choice of Study Area ........................................................................................................42
3.3. Criteria Selection and Data Preparation ............................................................................44
3.3.1. Elevation ..................................................................................................................45
3.3.2. Slope ........................................................................................................................47
3.3.3. Precipitation .............................................................................................................48
3.3.4. Drainage Proximity ..................................................................................................49
3.3.5. Drainage Density .....................................................................................................50
3.3.6. Lithology ..................................................................................................................51
3.3.7. Lineament Proximity ...............................................................................................52
3.3.8. Road Proximity ........................................................................................................53
3.3.9. Unselected Criteria ...................................................................................................55
3.4. Weighted Overlay .............................................................................................................56
3.4.1. Reclassification of Criteria ......................................................................................56
3.4.2. Weighting of Criteria ...............................................................................................68
3.5. Fuzzy Overlay ...................................................................................................................75
3.5.1. Fuzzy Membership Layers .......................................................................................75
3.5.2. Selection of Fuzzy Overlay Method ........................................................................84
Chapter 4 Results .......................................................................................................................... 87
4.1. Weighted Overlay Result ..................................................................................................88
4.2. Fuzzy Overlay Result ........................................................................................................96
Chapter 5 Discussion and Conclusion ........................................................................................ 104

vi

5.1. Bias and Limitations .......................................................................................................104
5.2. Societal Impacts ..............................................................................................................108
5.3. Future Application and Work .........................................................................................114
References ................................................................................................................................... 116

vii

List of Tables
Table 1 Literature references ........................................................................................................ 14
Table 2 Landslide susceptibility criteria in the literature.............................................................. 19
Table 3 Weighting schemes in other studies................................................................................. 39
Table 4 Variables and raw data ..................................................................................................... 45
Table 5 Elevation reclassification ................................................................................................. 56
Table 6 Slope reclassification ....................................................................................................... 58
Table 7 Precipitation reclassification ............................................................................................ 59
Table 8 Drainage proximity reclassification ................................................................................. 61
Table 9 Drainage density reclassification ..................................................................................... 62
Table 10 Lithology reclassification .............................................................................................. 64
Table 11 Lineament reclassification ............................................................................................. 66
Table 12 Road proximity reclassification ..................................................................................... 67
Table 13 Criteria used for reclassification .................................................................................... 69
Table 14 Calculation of normalizing percentages ........................................................................ 71
Table 15 Normalization of percentages ........................................................................................ 72
Table 16 Missing percent addition ................................................................................................ 72
Table 17 Finalized reclassification ............................................................................................... 73
Table 18 Ranking for slope derived from frequency .................................................................... 74
Table 19 Slope categorization by frequency ................................................................................. 75
Table 20 Fuzzy membership for each criterion ............................................................................ 76

viii

List of Figures
Figure 1 Denver Basin Geometry ................................................................................................... 3
Figure 2 Area of interest location ................................................................................................... 4
Figure 3 Denver Basin Cross-section ............................................................................................. 9
Figure 4 Landslide-rainfall index.................................................................................................. 33
Figure 5 Project Workflow ........................................................................................................... 41
Figure 6 Buffers on Colorado county boundaries ......................................................................... 44
Figure 7 Elevation in meters ......................................................................................................... 46
Figure 8 Slope in degree ............................................................................................................... 47
Figure 9 Precipitation in inches .................................................................................................... 49
Figure 10 Drainage proximity in meters ....................................................................................... 50
Figure 11 Drainage density in meters/meters squared .................................................................. 51
Figure 12 Lithologic classifications .............................................................................................. 52
Figure 13 Lineament proximity in meters .................................................................................... 53
Figure 14 Road proximity in meters ............................................................................................. 54
Figure 15 Reclassified elevation ................................................................................................... 57
Figure 16 Reclassified slope ......................................................................................................... 58
Figure 17 Reclassified precipitation ............................................................................................. 60
Figure 18 Reclassified drainage proximity ................................................................................... 61
Figure 19 Reclassified drainage density ....................................................................................... 63
Figure 20 Reclassified lithology ................................................................................................... 65
Figure 21 Reclassified lineament proximity ................................................................................. 66
Figure 22 Reclassified road proximity .......................................................................................... 67
Figure 23 Fuzzy Gaussian data distribution.................................................................................. 76
Figure 24 Fuzzy large elevation .................................................................................................... 77

ix

Figure 25 Fuzzy Gaussian slope ................................................................................................... 78
Figure 26 Fuzzy large precipitation .............................................................................................. 79
Figure 27 Fuzzy small drainage proximity ................................................................................... 80
Figure 28 Fuzzy large drainage density ........................................................................................ 81
Figure 29 Fuzzy large lithology .................................................................................................... 82
Figure 30 Fuzzy small lineament proximity ................................................................................. 83
Figure 31 Fuzzy small road proximity .......................................................................................... 84
Figure 32 Weighted overlay result ................................................................................................ 89
Figure 33 Comparison of weighted overlay and landslide inventory ........................................... 91
Figure 34 Slope and weighted overlay comparison ...................................................................... 93
Figure 35 Lithology and weighted overlay comparison ............................................................... 89
Figure 36 Fuzzy overlay result ..................................................................................................... 97
Figure 37 Comparison of fuzzy overlay and landslide inventory ................................................. 99
Figure 38 Slope and fuzzy overlay comparison .......................................................................... 101
Figure 39 Drainage systems and fuzzy overlay comparison ...................................................... 103
Figure 40 Detail of weighted overalay ........................................................................................ 111
Figure 41 Detail of fuzzy overlay ............................................................................................... 113

x

Abbreviations
AHP Analytical hierarchy process
AOI Area of interest
DEM Digital elevation model
ELECTRE ELimination Et Choix TRaduisant la REalité
GIS Geographic information systems
GISci Geographic information science
GIS-MCDA Geographic information science-multiple criteria decision analysis
LiDAR Light Detecting And Ranging
LR Logarithmic Regression
LULC Land use/land cover
MAUT Multi-attribute utility theory
MCDA Multiple criteria decision analysis
MS Mean and standard deviation
MSL Mean sea level
NDVI Normalized difference vegetation index
NOAA National Oceanic and Atmospheric Administration
PROMETHEE Preference Ranking Organization METHod for Enrichment Evaluations
SPI Stream power index
STI Sediment transportation index
SWR Soil water retention
TPI Topographic position index
TWI Topographic wetness index

xi

VC Vegetation coverage
US United States
USDA United States Department of Agriculture
USGS United States Geological Survey

xii

Abstract
Landslide susceptibility mapping incorporates variables such as slope, precipitation, and
lithology, among others, alongside a wide range of different methodologies in order to generate
maps that may aid in landslide prediction. Criteria in the literature is expansive and varied, and
the weighting methods used equally so. Weighted overlay and fuzzy overlay were chosen and
compared using a select number of criteria as a means of testing which method would yield a
better, more accurate result. Between the two, fuzzy overlay appears to be the more accurate of
the two methods after evaluating the outputs, and this is due to the ways in which the two
methods classify criteria. Of the eight criteria used, slope has been the most influential criterion
for both methods with lithology coming in as a surprisingly strong factor for the weighted
overlay and drainage systems as a strong influence for the fuzzy overlay. This influence is
reflected in the locations of areas of higher landslide susceptibility and reveal that weighting and
bias have definite effects on the outputs. There then exists a circular influence between the
outputs shaping decisions that may affect large numbers of people and decisionmakers’ opinions
affecting criteria emphasis. Of the two methods used, fuzzy overlay produced less biased results
than weighted overlay, as the emphasis used in weighted overlay are highly subjective and
influenced by the user.

1

Chapter 1 Introduction
Landslides are natural phenomena that cause extensive damage to property and loss of life across
the globe (AMERICAN GEOSCIENCES INSTITUTE n.d.; USGS n.d.; Wayllace et al. 2019;
Wieczorek and Leahy 2008). They are not easy to predict, as they have various triggers and
mechanisms of movement (Highland 2008; Korup 2017). Researchers use geographic
information science (GIS) to not only image the before and after of a landslide event (Petley
2012) but also to create predictive maps and models on landslide severity and frequency (Lai and
Tsai 2019; Qiu and Mitani 2017; Saha et al. 2005; Tan et al. 2020; Zhou et al. 2021). The
purpose of this project is to compare landslide susceptibility results of two methodologies,
weighted overlay and fuzzy overlay, to understand how differences in weighting methods affects
the results. An overview of the different criteria used in the literature is covered, as well as what
variables were included in the analyses. The results of the two methods are then compared along
with limitations of this study as well as how projects like this impact society.
1.1. Background
The combination of geographical information science and geology makes logical sense,
as both fields convey information primarily through the use of maps. Studies viewing landslides
through the combined lenses of geology and GIS have been done before, and research and
modelling within the geoscience realm has prioritized the effects of precipitation and
groundwater on landslide initiation and movement (Bogaard et al. 2007; Shou and Chen 2021;
Wayllace et al. 2019). Many studies on various geologic properties of landslides that are not
strictly “spatial” and do not employ GIS exist, such as studies of water saturation and pore
pressure (De Maio et al. 2020; Schulz et al. 2009; Viesca and Rice 2012; Wang and Sassa 2009),

2

soil composition and geology (Blońska et al 2018; Donnarumma et al. 2013; Park 2015), and
slope angle (Çellek 2020; Coe et al. 2004; Iwahashi et al. 2002; Zakaria et al. 2017).
A variety of models have been used for landslide hazard and prediction ranging from
heuristic fuzzy approach to machine learning (Ercanoglu and Gokceoglu 2001; Feizizadeh and
Blaschke 2012; Feizizadeh and Blaschke 2014; Francioni et al. 2019; Kavzoglu et al. 2013; Lai
and Tsai 2019; Palcic and Lalic 2009; Stanley and Kirschbaum 2021; Zhou et al. 2021). Multiple
criteria decision analysis (MCDA) has been the most frequently used method to determine
locations of highest landslide risk. There are multiple methods within MCDA to weight criteria,
and each method has its strengths and weaknesses. Here, the weighted overlay method was used
to rank and weigh criteria and a fuzzy overlay was used as comparison.
This project utilized a number of criteria that have been used for landslide susceptibility
mapping and evaluation. A digital elevation model (DEM) is an important dataset for this
project, as from this one piece of data, slope may also be derived. As far as properties of the
natural environment go, geology, precipitation, and rivers have been shown to be important
factors that were used (Ayalew and Yamagishi 2004; Chen and Li 2020; Du et al. 2014; Erener
et al. 2016; Feizizadeh and Blaschke 2013; Feizizadeh and Blaschke 2014; Lai and Tsai 2019;
Mallick et al. 2018; Patil et al. 2020; Roccati et al. 2021; Shou and Chen 2021). Data on the built
environment includes roads, to analyze proximity of landslide risk to areas of human habitation
and activity.
1.2. Study Area
The study area is a 10,947 square mile portion of Colorado that covers most of the Front
Range just to the west of the urbanized metroplex of Denver, Boulder, and Fort Collins. This
area was chosen because the presence of the Rocky Mountains renders it particularly prone to

3

landslides. The Front Range is a smaller mountain belt within the Rocky Mountains that lies
directly west of the Denver area (Figure 1).

Figure 1 Schematic diagram of the geometry of the Denver Basin
Its boundary demarcates a five-mile buffer around three counties that encompass the
majority of the predicted high-hazard locations: Larimer, Boulder, and Jefferson Counties.
Figure 1 shows the locations of predicted landslide hazards as generated by the United States
Geological Survey (USGS), along with population centers within Colorado borders. The
predicted locations tend to coincide with areas of relatively significant elevation change (Figure
2).

4

Figure 2 Area of interest location and detail
The goal of this project is to produce detailed analyses of potential landslide
susceptibility locations using both weighted and fuzzy overlays. These results are compared not
only to each other, but also to the landslide susceptibility inventory generated by the USGS.
1.3. Motivation
Landslides are significant natural hazards around the globe, and the US has its fair share
of landslides (Regmi et al. 2013). According to the USGS, landslides account for 25-50 deaths in
the US each year (USGS, n.d.; Wieczorek and Leahy 2008) and between $2-4 billion in annual
losses (AMERICAN GEOSCIENCES INSTITUTE, n.d.; Wayllace et al. 2019). The ability to
predict landslide events, like volcanic eruptions and earthquakes, can save numerous lives as

5

well as decrease damage to property and infrastructure (Winter et al. 2019). Other natural
disasters, such as wildfires, earthquakes, tsunamis, and volcanic eruptions (Korup 2012), or rapid
changes in weather (Huggel et al. 2012) also have a tendency to affect landslide frequency.
Therefore, it is important to have a baseline understanding of topographies that are more
susceptible to landslides in different environments.
Prediction of landslide occurrences may be improved upon with improved integration of
GISci and geology. The field of geology covers a sizable range of topics of study, among them
the wholesale study of the various kinds of landslides, including composition, failure
mechanisms, and movement mechanics (Çellak, 2020; Çellak, 2021; Cerri et al. 2020; Chen et
al. 2016; Donati et al. 2019; Donnarumma et al. 2013; Hu and Bürgmann 2020; Di Maio et al.
2020; Clague and Stead 2012; Glade et al. 2005; Highland 2008; Martel 2004). Many of these
studies observe landslides strictly through the lens of geology, though the integration of GIS into
geologic studies has increased as technology has developed (e.g., Ali et al. 2021; Bragagnolo et
al. 2020; Chen and Li 2020; Feizizadeh and Blaschke 2013; Feizizadeh and Blaschke 2014;
Francioni et al. 2019; Kavzoglu and Colkesen 2012; Mallick et al 2018; Saha et al. 2005; Zhou et
al. 2021). Many of these studies have incorporated geology into spatial modeling to further
landslide susceptibility mapping. However, they may still be lacking in the comparison between
the two weighting methods, as well as an analysis on how bias affects the results.
1.4. Importance
There are many variables that go into the process that both affected and are affected by
the opinions of the decisionmakers. Variables are weighted according to what are believed to be
more or less important, and this relative subjectivity is integrated into the results. The availability

6

of data, as well as the quality of data, also affects results. These results then may be incorporated
into land surveys for future infrastructure and urban expansion.
The utility of landslide susceptibility mapping cannot be overstated, particularly in
regions prone to landslides. The ability to predict where landslides might occur is an area of
study that is continually growing, and the environmental impacts of these natural occurrences
can be felt for years after an event. This project aims to look at the differences between two
variations on MCDA: weighted overlay and fuzzy overlay. These two methods were chosen due
to their widespread usage, and this project aims to compare the results within the AOI with
various criteria.
Research into landslide susceptibility mapping has yielded a number of different criteria
depending on the location, data available, and if there is a specific aspect of landslide
susceptibility that is being focused on. Bias in criteria selection or methodological decisions must
be considered when building a project and generating results, and these results may affect people
who live and work within the study area in question. The number of criteria reviewed for this
project far exceeds the number used. Some of the reviewed criteria are location-specific to the
studies conducted, and ultimately the results of this project are pertinent to the AOI alone. This
project does not aim to test models that address global landslide susceptibility mapping.
Many factors had to be considered over the course of this project. How this project would
contribute to both the scientific community as well as the wider general populace was a key
component in determining the subject matter and study area. The results of this project were
meant to be a practical and useful piece of information to aid in more accurate and precise
potential landslide location determinations.

7

1.5. Overview
The remainder of this thesis is split into four chapters. Chapter 2 details previous work
done that is related to this project, delving into more depth with regards to landslides, landslide
susceptibility mapping, utilized criteria, and criteria weighting. Chapter 3 discusses the
methodology used in this project, including data acquisition, research design, data preparation,
and data processing. Chapter 4 focuses on the material results from the project. And finally,
Chapter 5 dives into the analyses of the project results, as well as its impact on the scientific
community.

8

Chapter 2 Related Work
Landslides are hazards that have increasingly posed potential threats to mankind (Glade and
Crozier 2005). With ever-expanding urbanization encroaching on areas of unpopulated
wilderness, increasing numbers of people are living in environments prone to landslides
(AMERICAN GEOSCIENCES INSTITUTE, n.d.; USGS, n.d.; Winter et al. 2019), and
therefore it becomes ever more important to both understand and forecast how and where
landslides form. Geoscientists have reconstructed historical landslides in order to get a better
grasp on landslide susceptibility prediction (Ebertardt 2012; Regmi et al. 2014; Reid et al. 2012),
and the USGS maintains an extensive inventory of historic landslides in the US. Predicting
landslide locations is done through the use of models constructed with specific criteria in mind –
conditions that are necessary for landslide initiation. Geoscientists, therefore, have generated a
number of various model types used for landslide susceptibility mapping. Understanding the
overall processes that initiate landslides and what conditions are necessary for initiation is a
primary point of interest for this project.
2.1. Regional Setting
The study area covers a small central-western portion of the Denver Basin that lies
directly to the east of the Front Range. The present-day basin formed as a result of the uplift of
the Front Range, with eroded sediment from the mountain belt loading down the western edge of
the basin (Figure 3). The sediment weathered and eroded from the Front Range is deposited at
the base of the mountains, which in turn causes that side of the basin to sink—hence the fact that
the Denver Basin is what geologists would call an asymmetric basin. The Front Range consists
primarily of uplifted formations that range from the Precambrian to the present – the
Precambrian basement is predominately granite, the Paleozoic and Mesozoic layers consist

9

primarily of alternating siliciclastic and carbonate rocks, and the Cenozoic rocks are composed
of mixed-source sedimentary rocks (Knepper 2002).

Figure 3 Cross-section of the Denver Basin (Nelson and Santus 2011)
The uplift of the modern Rocky Mountains, the mountain belt of which the Front Range
is a part of, began during the late Pennsylvanian period, around 300 million years ago. The
Denver basin was a shallow basin for most of its lifespan up until the Paleogene (65-45 Ma),
when the Laramide orogeny uplifted the modern Rocky Mountains, causing sediment from the
mountains to deposit and depress the western end of the basin (Nelson and Santus 2011). The
Front Range is one of many smaller mountain belts that make up the greater Rocky Mountains.
The metropolitan areas of Denver, Boulder, and Fort Collins sit just to the east of the
Front Range, near the axis of the basin. The locations of the cities themselves do not directly
suffer from the effects of landslides, though their corresponding suburbs to the west –
particularly where they press into the foothills of the Front Range – tend to feel said effects more
frequently due to proximity to potential landslide hazard locations.

10

Ott (2020) and Suchet et al. (2003) indicate that intrusive igneous rocks, metamorphic
rocks, and siliciclastic sedimentary rocks weather the slowest, while carbonate sedimentary rocks
weather the fastest. In between these endmembers are extrusive igneous rocks and rocks
consisting of a mix of igneous, metamorphic, and sedimentary rocks. Unconsolidated lithology
consists of material that has already been eroded and mixed with other sediments that, size-wise,
are more likely to be easily transportable. As the Front Range consists of Precambrian granite –
an intrusive igneous rock – at its core with Paleozoic and Mesozoic siliciclastic and carbonate
rocks above, the softer carbonates likely eroded before the siliciclastic rock and granite.
2.2. Landslide Overview
The term “landslide” encompasses a wide range of mass movement, be it slope failure or
otherwise due to gravity (Highland 2008; Korup 2012; Yamagishi 2017). Highland (2008) and
Yamagishi (2017) break down the term landslide into different classifications based on
movement type and involved material. Different terms to describe displacement of land
masses/material include avalanche, fall, flow, slide, slump, spread, and topple (Clague and
Roberts 2012; Highland 2008; Yamagishi 2017), and each of these terms have their own set of
criteria that differentiates how they are named. Landslides have a variety of triggering
mechanisms, some of which include earthquakes, heavy precipitation, snowmelt, and erosion
(Korup 2012). For purposes of this project, the term “landslide” refers primarily to mass
movements of earthen material that is categorized as a flow (spatially continuous movement of
material where surface of shear is short-lived, closely-spaced, usually not preserved), slide
(cohesive, relatively undeformed material moves over surface of rupture or relatively thin zones
of intense shear strain), or fall (detachment of rock and soil from a steep slope along a surface
with little no shear displacement) (Highland 2008).

11

2.2.1. Landslide Anatomy
A failure surface, or sliding surface, is the plane on which the bulk of a landslide’s
material travels over (Highland 2008; Martel 2004). The surface itself is a plane of weakness
where gravity is able to overcome the shear frictional forces keeping the material above it
immobile and may be planar or curved (Martel 2004). These failure surfaces may be visible on
the surface and are known as rupture surfaces (Highland 2008).
A landslide consists of multiple features, though terminology can differ even when
referring to the same feature. The surface of rupture, also known as the sliding surface, is the
main surface on which the loose debris travels upon. The head of a landslide is the term used to
describe the area furthest upslope where material has shifted, while the toe of a landslide
represents the material that has traveled furthest downslope. Scarps form at or near the head of
the landslide due to tension, and tension fractures may form on the flanks of the landslide.
Material that has traveled downslope may buckle if there is a decrease in the travel velocity,
forming ripples or ridges on the surface.
2.2.2. Landslide Triggers and Characterization
Landslides occur due to a variety of triggering mechanisms, both natural and artificial
(Korup 2012). In places with high amounts of precipitation, rainwater can saturate the soil and
turn it into mud, which then slumps downhill due to decreased structural integrity (Highland
2008). Volcanic eruptions, along with earthquakes that may or may not be associated with said
eruptions, can destabilize the uppermost layer of earthen material due to the excessive vibrations
caused by such events (Korup 2012). Time and erosion may initiate slope failure simply due to
gravity (Clague and Roberts 2012). Manmade structures, such as roads, may also cause
landslides both during the construction process – destabilization of the surface due to

12

construction equipment and explosives, if bedrock is needed to be cleared – and post
construction – the resultant excavated road is a surface of high change in relief (Highland 2008).
Regmi et al. (2014) performed a study that focused on characterizing landslides in
Colorado. In their paper, they described two different types: smaller, surficial landslides and
larger, deep-seated landslides. According to the authors, these two kinds of landslides have a
temporal aspect to size: evidence of shallow landslides are modern-aged and small- to large-
sized, while evidence of large, deep-seated landslides are much older in age (hundreds to
thousands of years old). Shallow small- to medium-sized landslides tend to be located in areas of
steeper slope and are dominated by sedimentary rock close to rivers, while large-sized landslides
trend along areas of flatter slope (Cruden and Varnes 1996; Regmi et al. 2014; Wieczorek and
Leahy 2008). Shallow landslides have a sliding base that generally lies at the boundary between
soil and bedrock, and pore water plays a major role in the initiation of a landslide (Regmi et al.
2014). Deep-seated landslides have been determined to be older due to the fact that they tend to
have dense vegetation cover (Regmi et al. 2014). Shou and Chen (2021) imply that collapse and
movement mechanics tend to dictate the type of landside the occurs.
Of the landslides that have occurred in modern times in the state of Colorado, the
Slumgullion earthflow in southwestern Colorado has been the most heavily studied (Amitrano et
al. 2019; Gomberg et al. 1995; Gomberg et al. 2011; Madison et al. 2019). This nearly four-
kilometer-long landslide has been in motion for approximately 350 years at a rate of up to two
centimeters per day (Amitrano et al. 2019; Madison et al. 2019), giving scientists a natural
laboratory on landslide kinematics. The lithology and hydrology of the Slumgullion area has
been extensively studied, and the scale of the landslide has given rise to various “sections”
within the total length that act semi-independently of the whole, therefore allowing geoscientists

13

to study temporal and mechanical differences within the Slumgullion (Gomberg et al. 1995;
Gomberg et al. 2011; Madison et al. 2019). This research has tremendously increased knowledge
on landslide mechanics, and that translates into generalized landslide susceptibility mapping.
2.3. Criteria Employed in Landslide Prediction
For this project, a spread of forty-two papers (Table 1) were used to determine what
criteria were most frequently used in landslide susceptibility efforts. These studies varied widely
in terms of location: the studies were spread out within fourteen different countries across the
globe.

14

Table 1 Literature references (in alphabetical order)
Number Reference
1 Ali et al. 2020
2 Ayalew and Yamagishi 2004
3 Bragagnolo et al. 2020
4 Chen and Chen 2021
5 Chen and Li 2020
6 Chen et al. 2015
7 Du et al. 2017
8 Ercanoglu and Gokceoglu 2015
9 Erener et al. 2016
10 Feizizadeh and Blaschke 2013
11 Feizizadeh et al. 2014
12 Ghorbanzadeh et al. 2019
13 Huggel et al. 2012
14 Kavzoglu et al. 2013
15 Korup 2012
16 Lehmann et al. 2019
17 Lombardo and Mai 2018
18 Mallick et al. 2018
19 Nandi and Shakoor 2009
20 Nohani et al. 2019
21 Patil et al. 2019
22 Pawluszek and Borkowski 2016
23 Pham et al. 2020
24 Pourghasemi et al. 2012
25 Pourghasemi et al. 2020
26 Regmi et al. 2014
27 Rengers et al. 2016
28 Roccati et al. 2021
29 Roodposhti et al. 2016
30 Roy and Saha 2019
31 Roy et al. 2019
32 Saha et al. 2005
33 Saito et al. 2009
34 Schulz et al. 2009
35 Shou and Chen 2021
36 Vahidnia et al. 2010
37 Vakhshoori and Zare 2016
38 Vojtekova and Vojtek
39 Wayllace et al. 2019
40 Zhao et al. 2017
41 Zhou et al. 2021
42 Zhu et al. 2014

15

A multitude of various criteria were used in landslide susceptibility modeling, ranging
from atmospheric to geologic and everything in between. These criteria were used depending on
the purpose of the model being used. Landslide susceptibility is dependent on a number of
factors, though lithology, slope, and groundwater level are three of the most heavily studied
parameters (Regmi et al. 2014; Schulz et al. 2009; Wayllace et al. 2019; Zhou et al. 2021),
though they are certainly not the only parameters taken into consideration. Attributes such as
elevation, aspect, soil composition, land use, stream power index (SPI), and topographic wetness
index (TWI) have also been used in landslide susceptibility studies (Highland 2008; Kavzoglu et
al. 2013; Martel 2004; Saha et al. 2005; Schulz et al. 2009). Natural disasters and changes in
weather (Huggel et al. 2012; Korup 2012) have also been studied with regards to landslide
initiation frequency, and vegetation cover is one of the lesser-known factors studied in landslide
frequency (Lehmann 2019; Rengers et al. 2016).
The criteria authors working in landslide susceptibility mapping chose include various
attributes specific to not only their study areas, but to certain aspects of their study that they
either wanted to emphasize or had particular access to. For example, Chen et al. (2015) utilized
evaporation rates and soil water retention in their study – a criterion that no other study
referenced in this thesis has used. Roccati et al. (2021) used terracing and existing landslide
proximity as criterion for their own study – factors which influence slope gradient and surface
stability. The individual criteria, in the order listed in Table 2, are detailed in the following
sections, organized by criteria category.

16

Tally
36
32
30
12
6
2
1
1
1
1
1
1

References
1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 14, 16, 17, 18, 19, 20, 21,
22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 35, 36, 37,
38, 40, 42
1, 2, 3, 4, 5, 7, 8, 9, 10, 11, 12, 14, 17, 18, 19, 20, 21,
22, 23, 24, 25, 27, 28, 30, 31, 32, 33, 35, 37, 38, 40, 42
1, 2, 3, 4, 5, 7, 8, 9, 10, 11, 14, 17, 18, 20, 22, 23, 24,
25, 26, , 30, 31, 32, 33, 35, 36, 37, 38, 40, 42
1, 3, 4, 5, 12, 17, 20, 23, 24, 25, 33, 40
18, 21, 22, 26, 31, 40
24, 26
22
22
22
33
33
42
Table 2 Landslide susceptibility criteria used in the literature
Description
Vertical distance from a given baseline, usually
mean sea level (MSL)
Direction a slope faces, usually measured in
degrees from north
Angle of inclination of a surface
Derivative of elevation that calculates curvature
of a surface
Classification of landforms with regards to
erosional properties (with respect to landslide
studies)
Distance of a consistent angle of inclination of a
surface
Slope derivation via moving standard deviation
filter
A hypothetical illumination value of the surface
Calculated difference between cell elevation and
mean elevation of neighboring cells
Depth from the summit level (summit level
minus elevation)
Height above river level (elevation minus river
level)
Measure for the shape of each pixel
Criteria
Elevation
Aspect
Slope
Curvature
Geomorphology
Slope Length
Roughness
Shaded Relief
Topographic Position
Index (TPI)
Desiccation Height
Undesiccation Height
Slope Shape
Category
Topography

17

Tally
29
18
10
7
5
4
2
1
1
References
1, 4, 5, 6, 7, 9, 10, 11, 12, 13, 15, 18, 19, 20, 21, 22, 23,
24, 25, 26, 28, 29, 30, 31, 35, 36, 37, 38, 40
1, 5, 6, 10, 11, 13, 14, 15, 16, 18, 19, 21, 31, 34, 37, 39,
40, 41
1, 3, 4, 5, 14, 17, 22, 24, 25, 31
1, 4, 5, 17, 22, 24, 31
6, 11, 14, 29, 32
1, 4, 5, 31
1, 22
6
8
Description
Proximity to a fluvial network
Rainfall in a given area
A function of the slope and upstream
contributing area per unit width orthogonal to
flow direction
Potential for flow erosion at a given point on the
surface
Calculated density of fluvial network in a given
area
A function of slope and upstream catchment area
Radiant energy for a specific location and date
Sum of evaporation from water, soil, and plants
A classification on whether or not slip surfaces
are wet
Criteria
Drainage Proximity
Precipitation
Topographic Wetness
Index (TWI)
Stream Power Index
(SPI)
Drainage Density
Sediment
Transportation Index
(STI)
Solar Radiation
Evaporation
Water Conditions
Category
Hydrology

18

Tally
30
23
9
4
5
2
1
1
1
References
1, 2, 3, 4, 5, 7, 9, 10, 13, 14, 15, 17, 18, 21, 22, 23, 24,
25, 26, 28, 29, 30, 31, 32, 33, 34, 36, 38, 39, 40
1, 2, 5, 7, 9, 10, 11, 12, 17, 18, 20, 21, 23, 24, 25, 29, 30,
31, 32, 35, 36, 37, 38
4, 5, 9, 16, 18, 25, 30, 31, 34
8, 9, 23, 39
13, 15, 21, 31, 41
21, 32
6
19
35
Description
Geologic composition of the underlying bedrock
Proximity to faults or fractures in the underlying
bedrock
Composition of the top soil
Depth of the top soil
Categorization on whether or not area is in a
seismically active zone
Proximity to major geologic structural features
Ability of soil to absorb and retain free flowing
water
Index describing water content required to
liquify soil
Index describing topographic surfaces which
slope in the same direction
Criteria
Lithology
Lineament Proximity
Soil Composition
Depth of Soil
Seismic Zone
Major Structure
Proximity
Soil Water Retention
(SWR)
Soil Liquidity Index
Dip Slope Index
Category
Subsurface

19

Tally
28
19
15
3
3
3
1
1
References
1, 3, 4, 5, 6, 8, 9, 10, 11, 12, 13, 14, 17, 18, 19, 20, 21,
22, 24, 25, 28, 29, 30, 31, 32, 36, 37, 38
1, 2, 4, 5, 9, 10, 11, 13, 14, 20, 21, 22, 23, 24, 25, 28, 31,
35, 37
4, 5, 7, 13, 15, 16, 17, 18, 20, 21, 25, 30, 31, 35, 37
9, 18, 19
13, 21, 28
15, 26, 28
28
35
Description
Categorization of how surface is utilized
Proximity to a road
Percentage of vegetation coverage
Ability of top soil to be eroded
Proximity to an urbanized environment
Proximity to a previous landslide location
Presence of terraced surfaces
Correlation between cumulative rainfall and
rainfall intensity at different landslide locations
Criteria
Land Use/Land Cover
(LULC)
Road Proximity
Vegetation Coverage
Erodibility
Settlement Proximity
Existing Landslides
Proximity
Terracing
Landslide-Rainfall
Index
Category
Surface

20

2.3.1. Topography
Topography forms the basis of all landslide modeling. Without surface data, landslide
susceptibility modeling would not be possible. Different types of topographic data were utilized
in the various studies, with some forms being derivatives from others. A number of topographic
data are usually derived from DEMs (Ali et al. 2021; Ayalew and Yamagishi 2004; Chen and
Chen 2021; Chen and Li 2020; Chen et al. 2015; Mallick et al. 2018; Nohani et al, 2019;
Pawluszek and Borkowski 2016; Pham et al. 2020; Pourghasemi et al. 2020; Roodposhti et al.
2016; Roy and Saha 2019; Saito et al. 2009; Vahidnia et al. 2010; Vojtekova and Zare 2016;
Zhao et al. 2017) and are discussed below.
2.3.1.1. Elevation
Elevation is the vertical distance from a given baseline, usually mean sea level (Dempsey
2020), and elevation is an interpolated surface representation of DEMs (Du et al. 2014; Erener et
al. 2014). This vertical distance changes depending on surface features, such as mountains and
valleys above sea level. Chen et al. (2015) indicate that regions with low relative elevation
throughout are more prone to inundation, and that areas with a steeper topography have a lower
probability of flooding since water can be drained downslope.
2.3.1.2. Slope
Certain conditions need to be met in order for a landslide to occur, the least of which is a
slope gradient. Difference in elevation has a negative correlation with slope stability (Çellak,
2020; Kayastha 2015), and this factor is one of the most commonly used in landslide
susceptibility modeling, as reported by Çellak (2013), Dağ (2007), and Hasekioğullari (2011).
Çellak (2020) also mentions how slope plays an important role in lithological and soil properties,
such as permeability, cohesion, strain, and shear and normal stress, as well as hydrological
properties dealing with groundwater flow and saturation. Coe et al. (2004) reported that around

21

ninety-six percent of landslides within their study area had slopes between sixteen to forty-four
degrees. Patil et al. (2020) found in order of highest to lowest frequency of landslides occurring
on slope angle ranges in their study area: 30° – 40°, 20° – 30°, 10° – 20°, 40° – 50°, < 10°, 50° –
60°, 60° – 70°, > 70°. It is with their analysis that the ranking of slopes was determined for this
project.
2.3.1.3. Aspect
Erener et al. (2016) define aspect as a slope’s orientation using compass degrees, i.e., 0°
through 360°, with 0° and 360° both being due north, a property that may contribute to landslide
susceptibility modeling by implying which slopes are more likely to be affected by atmospheric
conditions, such as wind and precipitation, as well as amount of sunshine received (Pourghasemi
et al. 2012). Ghorbanzadeh et al. (2019) goes so far as to say that aspect is one of the most
important topographical features that can be used in landslide susceptibility studies. Aspect is
calculated based on the derived slope values.
2.3.1.4. Curvature
Curvature, as defined by Ghorbanzadeh et al. (2019) and Pourghasemi et al. (2012), is a
slope or aspect’s rate of change with respect to a particular direction. This criterion is particularly
useful in landslide susceptibility mapping because curvature defines topographic features as
concave, convex, or flat. Chen and Chen (2021) split curvature into five groups; Chen and Li
(2020), Lombardo and Mai (2018), Pourghasemi et al. (2020), and Saito et al. (2005) broke up
curvature into two categories for further distinction.
2.3.1.5. Geomorphology
Geomorphology refers to the landscape of an area that includes not only static surface
features, but also temporal aspects to the landscape, including seasonal changes (Mallick et al.

22

2018). Roy et al. (2019) use geomorphology to classify surface features into regions that not only
include lithological and soil composition, but also hydrological components.
2.3.1.6. Slope length
Slope length was a criterion that Pourghasemi et al. (2012) utilized, wherein this
parameter was a measurement of slope steepness and length. It was used as a means to measure
soil loss and sediment transport capacity overland through the use of a fluid.
2.3.1.7. Roughness
Pawluszek and Borkowski (2017) utilize roughness as a criterion. Roughness, as defined
by the authors, is a derivative from a slope map and applies a moving standard deviation filter
with a defined kernel size. Typically, rougher areas tend to indicate areas affected by landslides,
and the degree of roughness may correlate with specific types of landslide activities. Roughness
can be used to generate landslide inventory maps (McKean and Roering 2004).
2.3.1.8. Shaded relief
Shaded relief is, as defined by Pawluszek and Borkowski (2017), a hypothetical surface
illumination that works by visualizing shaded relief from eight different sun directions. Similar
in function to solar radiation, this data may be derived from DEMs.
2.3.1.9. Topographic position index
Topographic position index (TPI) is the calculated difference between the elevation of
one raster cell and the mean elevation of its surrounding cells (Pawluszek and Borkowski 2017).
This criterion identifies different topographic features based on the sharpness of edges, such as
ridges and valleys.

23

2.3.1.10. Desiccation height
Saito et al. (2009) use this criterion as a means of approximating ideal erosion volumes
and heights for both the past and the future. They define desiccation height as summit level
minus elevation (Dis), or depth from the summit level.
2.3.1.11. Undesiccation height
Undesiccation height, as defined by Saito et al. (2009) is the height above the river level,
or elevation minus river level. This criterion, like desiccation height, is used to approximate ideal
past and future erosion volume or height. Saito et al. (2009) further mention a correlation
between relief and slope angles to the standard deviation of undesiccation height. High relief and
steep slopes tend to have high standard deviation while low relief and flatter slopes correlate to
low standard deviation. The formulas to calculate the average undesiccation height in order to
get the standard deviation of undesiccation height are:

𝑈𝑑𝑖𝑠 𝑎𝑣
=
1
𝑁 ∑ 𝑈𝑑𝑖𝑠 𝑑 ( 1 )

𝑈𝑑𝑖𝑠 𝑠𝑑
=
1
𝑁 ∑ ( ( 𝑈𝑑𝑖𝑠 − 𝑈𝑑𝑖𝑠 𝑎𝑣
)
2
)
1
2
𝑖 ( 2 )
where Udis is measured undesiccation height, Udisav is average undesiccation height, Udissd is
the standard deviation of undesiccation height, and N is the population size.
2.3.1.12. Slope shape
Zhu et al. (2014) describe slope shape in terms of curvature: whether or not slopes are
flat, straight, convex, concave, or a combination thereof. They stated that areas with upper
convex, lower concave slopes are more likely to have landslide activity. The authors decided
upon slope shape instead of curvature because slope shape takes the shape of the entire slope into
account, whereas curvature measures the shape of a slope of a single pixel independently of its
neighbors.

24

2.3.2. Hydrology
Hydrology, for the purposes of this project, covers everything that pertains to water with
respect to landslide susceptibility. This includes, atmospheric, terrestrial, and subterranean water,
as water in all locations affect landslide initiation and propagation – water is generally the
lubricant that allows a landslide to travel on a slip surface (Wayllace et al. 2019). It can also be
the cause of slope failure – either by, again, due the fact that it can act as a lubricant, or by
weight should a soil absorb enough (Rotaru et al. 2007). Either way, reaching a critical saturation
point can tip a slope from being stable to unstable, thus giving rise to slope failure and
consequently, a landslide.
2.3.2.1. Drainage proximity
Many studies have used drainage proximity as one of their criterion (Chen et al. 2016; Du
et al. 2017; Erener et al. 2016; Feizizadeh et al. 2014; Ghorbanzade et al. 2019; Mallick et al.
2018; Pawluszek and Borkowski 2016; Pourghasemi et al. 2012; Roodposhti et al. 2016; Roy
and Saha 2019; Roy et al. 2019; Vahidnia et al. 2010; Vakhshoori and Zare 2016; Zhao et al.
2017). Weighting was done based on buffered distances to the nearest river or stream. Du et al.
(2016) point out how slope instability can develop as a result of river incision, hence the increase
of landslide susceptibility with drainage proximity.
2.3.2.2. Precipitation
A number of studies utilized precipitation as their means of gauging rainfall in their study
area. Chen et al. (2015) used daily rainfall, while Ali et al. (2021), Chen and Li (2020),
Feizizadeh et al. (2014), Vakhshoori and Zare (2016), and Zhao et al. (2017) used average
annual rainfall. Feizizadeh and Blachke (2013) and Mallick et al. (2018) used a 30-year
meteorological data, while Nandi and Shakoor (2009) used annual cumulative rainfall.

25

Regarding precipitation patterns, it has been noted that rainfall has a tendency to increase with
elevation, a trend that has been termed the orographic effect (Daly et al. 1993).
2.3.2.3. Drainage density
Chen et al. (2015), Feizizadeh et al. (2014), and Roodposhti et al. (2016) utilized
drainage density in their studies. Chen et al. (2015) defined drainage density as the length of
rivers for a given area. As mentioned by Du et al. (2016), the presence of a river can influence
slope stability, and therefore knowing what the drainage density is within a study area is crucial.
Kavzoglu et al. (2012) list the formula for drainage density as:

𝐷 𝑦 = ∑ 𝐿 𝐴 ⁄ ( 3 )
where Dy is the drainage density, L is stream length, and A is the catchment area.
2.3.2.4. Stream power index
Stream power index (SPI) is defined as flow erosion potential at a given surface point
(Pawluszek and Borkowski 2017; Roy et al. 2019) and is calculated with a formula Pourghasemi
et al. (2012) cited in their study:

𝑆𝑃𝐼 = 𝐴 𝑠 × tan 𝛽 ( 4 )
where As is catchment area and β the local slope in degrees.
2.3.2.5. Topographic wetness index
Pourghasemi et al. (2012), Pawluszek and Borkowski (2017), and Roy et al. (2019)
define TWI as a factor used to quantify topographic control on hydrologic processes. TWI is a
function of the slope and upstream contributing area per unit width orthogonal to flow direction:

𝑇𝑊𝐼 = ln
𝐴𝑠
tan 𝛽 ( 5 )

26

where As is the area that is drained through a certain point, and β the slope at the point of
drainage.
2.3.2.6. Sediment transportation index
Sediment transport index (STI) is a factor used to measure how an area directly
contributes to sediment discharge – it quantifies the process of erosion and deposition. Roy et al.
(2019) include the equation for STI in their study:

𝑆𝑇𝐼 = ( 𝑚 + 1)× (
𝐴𝑠
22.13
)
𝑚 × sin (
𝐵 0.0896
)
𝑛
( 6 )
where As is the catchment area, B the local slope in degrees, the contributing area exponent m is
usually set to 0.4, and the slope exponent n to 0.0896.
2.3.2.7. Solar radiation
Pawluszek and Borkowski (2017) define solar radiation – more specifically, area solar
radiation (ASR) – as a derivative of slope and aspect. This criterion combines radiant energy
from the sun with the sun angle and direction for a given location. Ali et al. (2021) describe
higher amounts of solar radiation being indicative of greater availability of soil and rock pore
space – hence a lower probability of landslide occurrence. Both Pawluszek and Borkowski
(2017) and Ali et al. (2021) essentially use solar radiation as a means of measuring evaporation.
2.3.2.8. Evaporation
Evaporation, as defined by Chen et al. (2015), is the sum of water, soil, and plant
evaporation. This criterion takes into account evapotranspiration from plants and evaporation
from surface water, though Chen et al. (2015) point out that this parameter is more critical during
the summer when evapotranspiration rates are higher due to seasonal rainfall and longer hours of
sunlight.

27

2.3.2.9. Water condition
Ercanoglu and Gokceoglu (2002) factored water condition – the amount of moisture
found on a given surface – into their study. They used a simple classification for assessing water
condition, as dense vegetation and mountainous terrain sometimes prevented direct observation.
Four categories were used: landslide susceptibility is high if water condition is wet, moderate if
water condition is dripping or flowing, low if water condition is damp, and a non-issue if water
condition is dry.
They caveated their observations by stating that water conditions in the same area change
with the seasons, and that the observations they used were what had been noted specifically at
the time the study was being conducted.
2.3.3. Subsurface
The subsurface category consists of any criteria that deals with data beneath the Earth’s
surface. These data include lithology, lineaments, various soil properties, and seismic activity, to
name a few. Subsurface data is primarily collected by geologists, geophysicists, hydrologists,
soil scientists, though data collection is not restricted to these professions. In terms of landslide
susceptibility mapping, subsurface data is important because a large part of why landslides occur
and where they occur is dictated by subsurface properties.
2.3.3.1. Lithology
Lithology is a key component of landslide susceptibility mapping, as geology affects not
only the subsurface, but the surface as well. The lithological composition of an area tends to
dictate not only subsurface properties such as porosity, permeability and fluid composition
(Schulz et al. 2009; Wayllace et al. 2019), but it also affects surficial properties such as
topography, rates of weathering and erosion, and soil composition (Ott 2020). Because geology
is such a key component of landslide susceptibility mapping, a significant number of the authors

28

referenced in this study have used lithology as one of their key criteria. Vojtekova and Vojtek
(2020) went one step further and used geology as a proxy for permeability.
2.3.3.2. Lineament proximity
Lineaments, in the geological and geographical sense, comprise surface features that are
generally indicative of subsurface structures, such as elongated hills and valleys. Florinsky
(2016) states that lineaments are usually associated with linear subsurface features such as faults
and fractures, and to a lesser extent, mechanical deformation – fracturing or folding – or zones of
higher permeability. Lineaments indicate planes of weakness in the subsurface, areas that may
slip with enough pressure or lubrication, resulting in landslides if they are close to the surface, or
earthquakes if they are deep within the Earth’s crust. Landslides may also result as an event
secondary to an earthquake, thus further reinforcing the correlation between landslide activity
and lineament proximity.
2.3.3.3. Soil composition
Soil composition is largely a result of the lithology of a region, as the soil’s minerals are
primarily sourced from their parent rocks. Different soil compositions have different properties,
such as soil depth, land use type, and level of erosion, and certain soil compositions correlate
more readily to landslide frequency (Erener et al. 2016; Pourghasemi et al. 2020). Pourghasemi
et al. (2020) also indicate that variation in soil composition changes the permeability and
strength of a slope surface.
2.3.3.4. Soil water retention
Chen et al. (2015) identify soil water retention (SWR) as the amount of water a soil can
store after some sort of precipitation or inundation event. Dependent on the soil composition, as
well as types of vegetation on the surface, SWR may change depending on the season. Chen et
al. (2015) include the formula for calculating SWR:

29

𝑆𝑊𝑅 𝑖 = 𝑆𝑊𝑅 0
(
100
𝐶𝑁
𝑖 − 1) ( 7 )
where CNi is an integer between 0 to 100 that is determined by hydrologic soil properties and
ground cover conditions, and SWR0 is a scaling factor dependent on the unit of measurement.
2.3.3.5. Depth of soil
Erener et al. (2016) utilizes depth of soil in their study by classifying soil depth into four
categories: very deep, deep, shallow, and very shallow. The authors of this paper have
discovered that for their study area in northwestern Turkey, landslides occurred most frequently
in the very deep category, in which soil depth was over 90 cm and the occurrence rate was 78%.
2.3.3.6. Seismic zone
Roy et al. (2019) used seismic zones as a criterion – seismic zones meaning areas where
there is ongoing tectonic activity, primarily in the form of earthquakes. Given that their study
area was in western Bengal in the foothills of the Himalaya Mountains, earthquakes are not
uncommon to the region and therefore likely serve as triggers to landslides that occur in the area.
2.3.3.7. Major structure proximity
Patil et al. (2020) used the proximity to major structures as a criterion, their chief focus
being thrust faults. Saha et al. (2005) also denote importance to proximity to major tectonic
structures, most notably thrust faults. Both Patil et al. (2020) and Saha et al. (2005) single out
thrust faulting in particular because of the fact that their study areas are in the Himalaya
Mountains, where active thrust faults routinely generate earthquakes.
2.3.3.8. Soil liquidity index
Soil liquidity index was used by Nandi and Shakoor (2009) quantifies the amount of
water needed in a soil to change it from a solid state to a plastic based on soil composition. The
equation for soil liquidity is:

30

𝐿𝐼 = ( 𝑊𝑛 − 𝑃𝐿 ) /( 𝐿𝐿 − 𝑃𝐿 ) ( 8 )
where LI is the liquidity index, Wn is water content, PL plastic limit, and LL liquid limit.
2.3.4. Surface
Surface data plays a large role in landslide susceptibility mapping. Of the data used in
landslide susceptibility studies, surface information changes the most frequently. This is due to
the fact that surface attributes are and have been anthropomorphically shaped in timescales that
can easily fit within an average human’s lifespan. The changes wrought on the surface therefore
heavily affect surface properties that, in turn, affect the probability of a landslide occurring.
2.3.4.1. Land use/land cover
Land use/land cover (LULC) were a significant component of many studies. Pourghasemi
et al. (2020) mention how land use can affect hydrological and mechanical slope stability
properties. Human activity is generally the cause of multiple triggers that contribute to climate
change, and changes in LULC is a means of mapping how anthropogenic activity affects natural
process, landslides included (Mallick et al. 2018).
2.3.4.2. Road proximity
Studies have revealed a correlation between landslide activity and the presence of roads
in mountainous regions (Erener et al. 2016; Feizizadeh et al. 2014; Pawluszek and Borkowski
2017; Pourghasemi et al. 2012; Roy et al. 2019). This is due to the fact that the construction of
roads requires the excavation of rock and soil, which subsequently weakens slope integrity
(Pawluszek and Borkowski 2017). Nohani et al. (2019) point out that road construction on slopes
more than 10 degrees are more prone to landslide activity.

31

2.3.4.3. Vegetation coverage
Du et al. (2017) and Mallick et al. (2018) both use vegetation coverage (VC) in their
studies, as it was noted that increased vegetation cover negatively affected landslide frequency.
Mallick et al. (2018) calculated VC from Landsat-8 satellite imagery. Du et al. (2017)’s VC data
was derived from Landsat ETM+ imagery, and both Du et al. (2017) and Roy et al. (2019) used
normalized difference vegetation index (NDVI) to calculate VC as follows:

𝑁𝐷𝑉𝐼 =
𝐼𝑅 − 𝑅 𝐼𝑅 + 𝑅 ( 9 )

𝑉𝐶 =
𝑁𝑉𝐷𝐼 − 𝑁𝑉𝐷𝐼 𝑠𝑜𝑖𝑙 𝑁𝑉𝐷𝐼 𝑣𝑒𝑔 − 𝑁𝑉𝐷𝐼 𝑠𝑜𝑖𝑙
( 10 )
where NVDI is the normalized difference vegetation index, IR the infrared portion of the
electromagnetic spectrum, R the red portion of the electromagnetic spectrum, NVDIsoil the NVDI
of uncovered soil, and NVDIveg the NVDI for pure vegetation.
2.3.4.4. Erodibility
Erener et al. (2016) and Mallick et al. (2018) measure the ability of sediment to be eroded
from the soil. This erosion includes consideration of ease of soil transport due to infiltration and
runoff. The equation Mallick et al. (2018) used for this calculation is:

𝐾 = 0.0293( 0.65 − 𝐷 𝐺 + 0.24𝐷 𝐺 2
) ( 11 )
𝑒𝑥𝑝 {−0.0021(
𝑂𝑀
𝑓 𝑐𝑙𝑎𝑦 ) − 0.00037(
𝑂𝑀
𝑓 𝑐𝑙𝑎𝑦 )
2
− 4.02𝑓 𝑐𝑙𝑎𝑦 + 1.72𝑓 𝑐𝑙𝑎𝑦 2
}

𝐷 𝐺 = −3.5𝑓 𝑐𝑙𝑎𝑦 − 2.0𝑓 𝑠𝑖𝑙𝑡 − 0.5𝑓 𝑠𝑎𝑛𝑑 ( 12 )
where K is the soil erodibility factor, DG is geometric mean radius, OM is the percentage of
organic matter, fsand the percentage of sand, fsilt the percentage of silt, and fclay the percentage of
clay.

32

2.3.4.5. Settlement proximity
Patil et al. (2020) used settlement proximity by weighting inside a Landslide Numerical
Risk Factor geospatial model, though no detail as to how they weighted settlement proximity
was given. Roccati et al. (2021) used a buffer distance of ten meters and broke down
“settlement” into further categories: buildings, other manufacts, and retaining walls.
2.3.4.6. Existing landslide proximity
Roccati et al. (2021) also considered the existence of pre-existing landslide deposits,
which may be indicative of higher slope instability. The authors separated previous landslides
into four categories: active/reactivated/suspended landslides, dormant landslides,
inactive/stabilized landslides, and area affected by widespread shallow landslides.
2.3.4.7. Terracing
Roccati et al. (2021) take terraced surfaces into account with regards to slope stability.
They discuss how terracing both improves and worsens slope stability, depending on amounts of
rainfall and runoff, as well as vegetation growth. Vegetation increases slope stability, while
rainfall and runoff decreases slope stability.
2.3.4.8. Landslide-rainfall index
Shou and Chen (2021) define this criterion as the correlation between cumulative rainfall
and rainfall intensity at different landslide locations, and the dataset is used to predict the
landslide-rainfall index at a chosen location. As shown in Figure 4, L1 and L2 indicate the upper
and lower thresholds for a dataset, which is then used to generate a rectangular bound for the
data. Midpoints are then used to determine the slopes for L1 and L2, and a line perpendicular to
L1 and L2’s slopes is used to determine d1 and d2, which are the distances between a chosen
point and L1 and L2, respectively. The equation for the landslide-rainfall index (Id) is:

33

𝐼 𝑑 = 𝑑 2
/( 𝑑 1
+ 𝑑 2
) ( 13 )
where d1 is the distance from L1 and d2 is the distance from L2.

Figure 4 How landslide-rainfall index is graphically calculated (Shou and Chen 2021).
2.4. Modeling Methods
Landslide susceptibility mapping is a generalized term used to denote spatial analysis
regarding landslide susceptibility, and a number of studies have utilized various methods for
generating landslide susceptibility models. Criteria and methodologies have differed among
these studies, but several specific criteria and susceptibility analysis methods have been utilized
more frequently than others.
Multiple-criteria decision analysis (MCDA) and logistic regression (LR) were some of
the more commonly used modeling methods used. Other methods are briefly covered, though
their usage within landslide susceptibility mapping is much less frequent than that of MCDA and
LR.

34

2.4.1. Multiple-Criteria Decision Analysis
MCDA is a decision-making model that utilizes weights to determine the importance of
criteria relative to one another in order to achieve a result that best fits the criteria used, and it is
this method that was used for this project. The MCDA process requires several key components:
the decision maker, evaluation criteria, and decision alternatives (Malczewski and Rinner 2015).
MCDA is a frequent choice for environmental suitability studies, as it provides a systematic
means of incorporating decision maker priorities and various criteria and outputs various
alternatives which the decision maker can then select from (Huang et al. 2011; Jankowski 1995).
Different studies utilized different methods for generating landslide susceptibility models and
maps. MCDA, logistic regression, and machine learning were some of the modeling methods
used. MCDA, as mentioned before, stands for multi-criteria decision analysis, and this method is
dependent upon using a weighting scheme to determine the importance of one criterion against
another.
MCDA with regards to GIS can be a powerful tool if properly utilized. GIS on its own
analyzes and visualizes spatial data, while MCDA provides a structure and weighted criteria for
decision making. The combination of the two (commonly referred to as GIS-MCDA)
complement each other, as the combination allows for decision-making to occur while taking
into consideration spatial data (Feizizadeh and Blaschke 2013). Malczewski and Rinner (2015)
argue that the purpose of using GIS-MCDA is to provide options with geographical input to aid
decision makers in developing and making better-informed solutions to the problem that required
an MCDA in the first place, as opposed to yielding a single solution. GIS-MCDA is therefore a
powerful, systematic tool that offers options to assistance in solving a problem with spatial
information that supports the decision maker’s ability to come to a decision that not only
incorporates spatial data, but also their own value judgments.

35

Under the umbrella of MCDA are a number of different methods to weight criteria. Of
these, analytical hierarchy process (AHP) is the most widely used method (see, e.g., Ahmed
2015; Chen et al. 2015; Feizizadeh and Blaschke 2013; Feizizadeh and Blaschke 2014;
Pawluszek and Borkowski 2017; Pourghasemi et al. 2012; Roccati et al. 2021; Roy and Saha
2019; Vojtekova and Vojtek 2020), though de Montis et al. (2005), Huang et al. (2011),
Malczewski (2004), Malczewski and Rinner (2015), and Triantaphyllou and Baig (2005) discuss
other weighting methods as well. Of these papers that discuss different weighting methodologies,
AHP was discussed in all five papers. ELimination Et Choix TRaduisant la REalité (ELECTRE)
was mentioned in three out of five, and multi-attribute utility theory (MAUT) and Preference
Ranking Organization METHod for Enrichment Evaluations (PROMETHEE) were each
discussed in two of the five papers listed. These methods are described in the following
paragraphs.
AHP was developed by Saaty in 1980. A hierarchy of criteria is created before paired
comparison ratios are used to determine how criteria are weighted (de Montis et al. 2005; Huang
et al. 2011; Malczewski 2004; Malczewski and Rinner 2015; Triantaphyllou and Baig 2005).
AHP has found wide usage in suitability analyses and conflict resolution (Saaty 1987).
ELimination Et Choix TRaduisant la REalité (ELECTRE) was developed in 1968 by Roy (de
Montis et al. 2005; Huang et al. 2011; Malczewski and Rinner 2015). This method compares the
concordance and discordance of paired alternatives, wherein if one criterion is determined to be
better than the one its being compared to, it receives a higher rank, or weight. MAUT, developed
by Churchman, Ackoff, and Arnoff in 1957, allows for multiple objectives, qualitative data, and
intangible factors to be considered in the weighting process (de Montis et al. 2005; Huang et al.
2011). It allows for the comparison of risky outcomes through computed expected utility.

36

PROMETHEE is similar to ELECTRE in that it also uses a ranking scheme and paired
alternatives. Unlike ELECTRE, PROMETHEE ranks the paired alternatives based on criterion
type and threshold values (Huang et al. 2011; Malczewski and Rinner 2015).
This project intends to use two methods that fall under the MCDA umbrella: weighted
overlay and fuzzy overlay. The two overlays utilize multiple criteria in which each criterion is
weighted or ranked in terms of importance. Weighted overlay works by breaking criteria into
sub-criteria and reclassifying them according to importance before then weighing the criterion
itself in relation to other criteria. Roslee et al. (2017) and Hassan et al. (2020) utilize weighted
overlay in spatial suitability analyses with the former focused on landslides in Pahang, Malaysia
and the latter on agricultural land in Pakistan. Both studies use weighted overlay to determine
spatial susceptibility or suitability based on the rankings they assigned to criteria. Fuzzy overlay,
unlike weighted overlay, substitutes assigned ranks with fuzzy memberships. There are seven
different memberships: fuzzy Large, fuzzy Small, fuzzy Linear, fuzzy Near, fuzzy Gaussian,
fuzzy mean and standard deviation (MS) Large, and fuzzy MS Small. Each membership
reclassifies a criterion based on what sub-criteria are considered more important than others.
Once each criterion has an associated fuzzy membership, a fuzzy overlay method is chosen.
There are five methods: fuzzy And, fuzzy Or, fuzzy Product, fuzzy Sum, and fuzzy Gamma, and
each one emphasizes specific aspects of the resultant combination of fuzzy memberships.
Hasanloo et al. (2019) and Baidya et al. (2014) use fuzzy overlays to analyze flood risk and land
resources.
With how many varied weighting methods there are, both weighted overlay and fuzzy
overlay seem to be ideal candidate methods for decision making with multiple criteria.
Malczewski (2004) points out that data – particularly spatial data – has inherent inaccuracy and

37

imprecision due to ambiguity in inputted data, be it from scaling or from user preferences. The
fact that data formatting is not standardized means that different data sources have different
levels of accuracy and precision (Malczewski 2004). De Montis et al. (2005) indicates that the
choice of weighting method by the decision maker may or may not be the best fit for whatever
problem they are trying to solve, and Steele et al. (2009) furthers this argument by suggesting
that weighting individual criterion is subjective because ranks are defined by the decision maker.
While MCDA in general is a popular and widely used method for decision-making that involves
multiple criteria, there are certainly other methods that exist to generate similar outputs.
2.4.2. Logistic Regression
Logistic regression (LR) is a multiple criteria regressive analysis in which the dependent
variable may be neither continuous nor quantitative, and the relationship with several
independent variables is explored (Lee 2005). It is a method that is used in predictive analysis,
utilizes binary dependent variables, and generates nonlinear models (Kavzoglu et al. 2012; Lee
2005). This method was used by Du et al. (2014), Erener et al. (2016), Kavzoglu et al. (2012),
and Lee (2005) as a means of comparing different methodologies. Ayalew and Yamagishi (2004)
and Lombardo and Mai (2018), in contrast, use LR as their sole means of analyzing landslide
susceptibility.
LR is calculated based on a general linear model equation:

𝑃 =
1
( 1 + 𝑒 𝑍 )
( 14 )
where P is event probability, e is the base of the natural logarithm, and Z is a value that ranges
from -∞ to +∞, and Z is defined by the following equation:

𝐷 𝐺 = −3.5𝑓 𝑐𝑙𝑎𝑦 − 2.0𝑓 𝑠𝑖𝑙𝑡 − 0.5𝑓 𝑠𝑎𝑛𝑑 ( 15 )

38

𝑍 = 𝐵 0
+ 𝐵 1
𝑋 1
+ 𝐵 2
𝑋 2
+ ⋯ + 𝐵 𝑛 𝑋 𝑛 ( 16 )
where B is the model’s intercept, n is the number of independent variables, and Bn is the
coefficient that measures Xn, which is the contribution of an independent variable. The dependent
variable in LR is expressed as:

𝐷 𝐺 = −3.5𝑓 𝑐𝑙𝑎𝑦 − 2.0𝑓 𝑠𝑖𝑙𝑡 − 0.5𝑓 𝑠𝑎𝑛𝑑
( 17 )

𝐿𝑜𝑔𝑖𝑡 ( 𝑝 ) = ln (
𝑝 1 − 𝑝 ) = 1 1 ⁄ + 𝑒 𝐵 0
+ 𝐵 1
𝑋 1
+ 𝐵 2
𝑋 2
+⋯+ 𝐵 𝑛 𝑋 𝑛
( 18 )
where p is the dependent variable probability and 𝑝 ( 1 − 𝑝 ) ⁄ is the likelihood ratio.
The advantage of using LR in susceptibility analysis is that the dependent variable only
outputs as one of two values: 0 or 1, and the results can be interpreted as a probability that ranges
from 0 to 1. If the result is closer to 0, then it has a lower probability of success, whereas if the
result is closer to 1, the odds of the result occurring is higher (Kavzoglu et al. 2012).
2.4.3. Other Methods
Aside from MCDA and LR, a number of other modeling methods exist. Of the studies
referenced for this project, over twenty different modeling methods were utilized for landslide
susceptibility mapping. These other methods have been summarized in Table 3.

39

Table 2 Weighting schemes used in other studies
Method Study
Artificial neural network Vahidnia et al. 2010
Analytic network process Ali et al. 2021
Association rule mining Erener et al. 2016
Convolutional neural network Ghorbanzadeh et al. 2019; Li 2020
Frequency ratio Vakhshoori and Zare 2016
Fuzzy logic
Ercanoglu and Gokceoglu 2002; Roy and Saha 2019;
Vakhshoori and Zare 2016
Fuzzy membership function Roodposhti et al. 2016
Heuristic fuzzy approach Stanley and Kirschbaum 2017
Fuzzy interference system Vahidnia et al. 2010
Information value method Du et al. 2014; Saha et al. 2005
Landslide nominal susceptibility
factor
Saha et al. 2005
Landslide numerical risk factor Roy and Saha 2019
Long short-term memory Li et al. 2021
Naïve Bayes Ali et al. 2021
Machine learning Lai and Tsai 2019
Monte Carlo Feizizadeh and Blaschke 2013
Ordered weighted average
Feizizadeh and Blaschke 2012; Feizizadeh and Blaschke
2013
Random forest
Ali et al. 2021; Ghorbanzade et al. 2019; Lai and Tsai
2019
Shannon entropy Roodposhti et al. 2016; Zhao et al. 2017
Support vector machine
Ghorbanzadeh et al. 2019; Li et al. 202; Roy et al. 2019;
Vakhshoori and Zare 2016
Support vector machine regression Kavzoglu et al. 2012
Variable weight combination Li et al. 2021
Weight of evidence Nohani et al. 2019; Roy et al. 2019
Weighted linear combination
Feizizadeh and Blaschke 2012; Feizizadeh and Blaschke
2013

40

Chapter 3 Methodology
This project aims to generate a model in which landslide susceptibility maps are created through
the use of weighted criteria. Two different methods for weighting were used to analyze the data:
weighted overlay and fuzzy overlay. The base data used for these different methods are identical
but are differentiated due to respective geoprocesses, and weighting was determined using
previous studies as a guide.
3.1. Research Design
The purpose of this project is to not only generate a model in which landslide
susceptibility locations may be predicted, but to also determine what criteria weigh more heavily
in determining landslide initiation. Data of all types have been used in a variety of studies that
used a number of methods, though this project did not use the complete list of criteria detailed in
Chapter 2.
For this project, Esri’s ArcGIS Pro is the primary software used in modeling landslide
susceptibility in the south-central Front Range. The data for this project were collected and
formatted before undergoing susceptibility analysis. Two methods – weighted overlay and fuzzy
overlay – were used, resulting in maps that can then be compared to determine the most accurate
predictions. These two methods were chosen to allow for repeatability and to test how different
methods affect results. A simplified workflow is shown in Figure 5.

41

Figure 5 Project workflow

42

42
The DEM was chosen to be used as the basis for both the scale and the snap raster
because of its high resolution and coverage area. Any generated rasters would be created using
the 10-meter resolution of the DEM, as well as the bounds. Other snap rasters were considered,
but the DEM was ultimate chosen due to the fact that it is the base upon which almost all of the
other analyses build upon.
A map projection was selected for the project, chosen based on what projection best fit
the AOI: in this case, NAD 1983 UTM Zone 13N. NAD 1983 Colorado State Plane Central FIPS
502 was considered but not used due to the fact that the three Colorado State Plane projections
focus on latitudinal bands of the state, as opposed to longitudinal bands (USA Contiguous Albers
Equal Area Conic USGS). As landslides vary in size and scale, data with the highest resolution
was used as the snap raster. For this project, the data with the highest resolutions available were
elevation and slope at ⅓ arc-second (10-meter), and elevation was selected as the snap raster.
Subsequent rasters generated from various shapefiles retained the same resolution as the snap
raster.
3.2. Choice of Study Area
The study area was chosen based on several factors, the first of which was the average
frequency of landslide occurrences. Colorado has a high incidence rate of landslides that
originate from the western half of the state that lies within the bounds of the Rocky Mountains.
The fact that the state has a long history of recorded landslides further made it a feasible
candidate study location, and the USGS also has multiple marked locations marked as “high
risk” for landslide activity. The fact that the Colorado Geological Survey also had landslide data
on their website also helped in determining what area to focus on for this study.

43

43
Once the initial data was downloaded, an AOI was selected based on data coverage, and
this location was determined using the USGS’s landslide hazard inventory and county
boundaries. Areas in the landslide inventory with a high density of high confidence in landslide
activity were focused on first. The region with the highest density of likely landslide locations
encompasses the Front Range in north-central Colorado. The Pairwise Buffer tool was used to
generate a buffer of five miles around county boundaries, and the bounds of the AOI were drawn
using points determined by the buffers of three counties: Larimer, Boulder, and Jefferson
Counties (Figure 6). These counties were chosen due to the fact that a majority of the high-risk
landslide hazard locations from the landslide inventory fell within these counties. Initially, these
three counties were to be the AOI for this project, but the cluster of high confidence that
straddled Clear Creek and Summit Counties to the west-southwest of the three counties could not
be ignored, and so the AOI was expanded to follow latitudinal and longitudinal lines. Once the
AOI was established, the rest of the data was clipped to the AOI for ease of use.

44

44

Figure 5 Buffers on Colorado county boundaries with the three chosen counties (Larimer,
Boulder, and Jefferson) lightly shaded in grey
3.3. Criteria Selection and Data Preparation
Eight different criteria were used for this study: elevation, slope, precipitation, drainage
proximity, drainage density, lithology, lineament proximity, and road proximity. These criteria
were chosen based on a combination of usage in the literature, data availability, and AOI
coverage (Table 4). Each of the criteria covered in 2.3 was searched to see if the associated data
could be included in this project and, if found, retained if the data coverage extended over the
whole of the AOI. Five different data sets were used to generate the eight criteria that were
ultimately used in this project. The data come from different sources and in different formats,

45

45
some of which required more preparation than others. The following sections detail these criteria
and data preparation processes.
Table 3 Variables employed and raw data utilized for this project
Criteria Data Format Source Description
Elevation DEM geoTIFF USGS
DEM with coverage across the
contiguous US
Slope DEM geoTIFF USGS
DEM with coverage across the
contiguous US
Precipitation Precipitation shapefile USDA
Averaged annual rainfall by state
from 1981 to 2010
Drainage
Proximity
Rivers shapefile NOAA
Rivers and streams within the
contiguous US
Drainage
Density
Rivers shapefile NOAA
Rivers and streams within the
contiguous US
Lithology Geology shapefile USGS
Geologic units and faults with
attribute data by state
Lineaments Geology shapefile USGS
Geologic units and faults with
attribute data by state
Road Proximity Roads shapefile
Colorado,
USCB
Major roads and highways within
the state of Colorado
3.3.1. Elevation
Elevation ranking was taken from a number of sources, as almost every study included
elevation as one of their criteria. Many of the studies observed a correlation between landslide
frequency and higher elevation (Ayalew and Yamagishi 2004; Chen and Li 2020; Du et al. 2017;
Feizizadeh and Blaschke 2013; Feizizadeh et al. 2014; Mallick et al 2018; Patil et al. 2019;
Roodposhti et al. 2016; Shou and Chen 2021; Vakhshoori and Zare 2016), though the number of
studies that detailed their ranking categorization were few and far in between.
Elevation was derived from a series of DEMs downloaded from the USGS. The USGS
has multiple resolutions available to the public that cover the entirety of the contiguous US and
the state of Alaska. The ⅓ arc-second (10-meter) resolution was chosen for this project because
of all the resources searched, it was the highest-resolution DEM available that covered the entire

46

46
extent of the AOI. The DEM used in this project consisted of a mosaic of smaller geoTIFFs 1° ×
1° in size that were combined using the Merge tool, and the color bar for the resultant DEM was
normalized into a single uniform color bar for better visualization. Elevation that was then
clipped to fit the AOI was derived from the merged DEM (Figure 7). Within the AOI, the
elevation ranged from 1,438 meters to 4,356 meters (4,717.8 feet to 14,291.3 feet).

Figure 6 Elevation in meters

47

47
3.3.2. Slope
Slope was derived from the merged DEM (Figure 8). Within the AOI, the slope ranged
from 0° to 85.1° For slope ranking, Patil et al. (2019) and Çellek (2020) indicated that slope
gradients of between roughly 20° to 40° experienced landslides the most frequently, with
landslide frequencies tapering off as slopes both decreased below 20° and increased greater than
40°.

Figure 7 Slope in degrees derived from elevation

48

48
3.3.3. Precipitation
Precipitation was chosen as a criterion due to the frequency of its use in other studies, as
well as the fact that the data was readily available. The data consists of averaged annual rainfall
in inches that date from 1981 to 2010. The data was sourced from the USDA and was
downloaded as a shapefile that covered the entire state. The clipped AOI data has a range of 10
inches to 52 inches (Figure 9). Precipitation was a somewhat common criterion used, and studies
found a correlation between higher landslide frequencies and increased amounts of rainfall (Ali
et al. 2020; Chen et al. 2015; Feizizadeh and Blaschke 2013; Feizizadeh et al. 2014; Nandi and
Shakoor 2009; Roy et al. 2019; Zhao et al. 2017).

49

49

Figure 8 Precipitation in inches
3.3.4. Drainage Proximity
A shapefile of major rivers in the contiguous US was downloaded from the NOAA
(Figure 12). This shapefile covers the contiguous United States and includes portions of rivers
and streams that originate in Canada. Drainage systems data was easily accessible and drainage
proximity factored into a number of studies (Ali et al. 2020; Erener et al. 2016; Ghorbanzadeh et
al. 2019; Nandi and Shakoor 2009; Nohani et al. 2019; Roccati et al. 2021; Vahidnia et al. 2010),
which revealed a correlation between distance from rivers and streams and landslide frequency.

50

50
The shapefile was clipped to the AOI and the Euclidean Distance tool was used to generate
distance buffers, which is shown in Figure 10.

Figure 9 Drainage proximity in meters with clipped drainage systems superimposed
3.3.5. Drainage Density
Drainage density used the same shapefile as drainage proximity and was generated with
the aid of the Line Density tool (Figure 12). Drainage density was, unfortunately, not as well
documented in the literature, and Saha et al. (2019) was the only one to detail how they
categorized the river density criterion.

51

51

Figure 10 Drainage density in meters/meters squared with clipped drainage systems
superimposed

3.3.6. Lithology
Geology was downloaded from the USGS. The shapefile covered the whole of the state
of Colorado before being clipped to the AOI (Figure 12). The different lithological compositions
were then visualized according to the associated key. Lithology was used as a criterion in many
studies (Ali et al. 2020; Chen and LI 2020; Du et al. 2017; Kavzoglu and Colkeson 2013;
Lombardo and Mai 2018; Pham et al. 2020; Roy and Saha 2019; Roy et al. 2019; Saha et al.
2005; Zhao et al. 2017). Despite the frequency of its use, however, Ott (2020) was the only

52

52
author to categorize lithologies by erodibility, and the rankings used in this project are based on
Ott (2020)’s work.

Figure 11 Lithologic classifications

3.3.7. Lineament Proximity
The lineament shapefile covered the entirety of Colorado and was color-coded according
to standard geological map key colors (Figure 13). Lineament proximity was used in several
studies, and many studies determined a correlation between lineament proximity and landslide

53

53
frequency (Chen and Li 2020; Erener et al. 2016; Ghorbanzadeh et al. 2019; Mallick et al. 2018;
Pham et al. 2020; Vakhshoori and Zare 2016).

Figure 12 Lineament proximity in meters with lineaments superimposed

3.3.8. Road Proximity
Road data was provided by both the US Census Bureau and the state of Colorado in the
form of shapefiles. A shapefile consisting of primary and secondary roads within Colorado was
downloaded from the Census Bureau while a shapefile containing major streets came from the
Colorado government. These two shapefiles were merged in order to form a more complete
reference for transportation passages throughout the state using the Merge tool before being

54

54
clipped to the AOI bounds (Figure 14). Road proximity was included in a number of studies
because the construction of roads – particularly in areas of dynamic topographical change – have
a tendency to destabilize slope gradients by way of creating extremely steep slopes in order to
create enough space level enough to build a road. This abrupt change in slope greatly increases
the chances of slope failure, which in turn may develop into a landslide (Ayalew and Yamagishi
2004; Chen and Chen 2021; Nohani et al. 2019; Pawluszek and Borkowski 2016; Shou and Chen
2021).

Figure 13 Road proximity in meters with clipped roads layer superimposed

55

3.3.9. Unselected Criteria
As mentioned in Chapter 2, landslide susceptibility mapping can and does utilize a
variety of different criteria. Some of those mentioned had been intended for use in this project,
but for various reasons ended up being discarded. The five criteria that were planned but rejected
were aspect, LULC, soil composition, water saturation, porosity, and population/census tracts.
Aspect was derived from elevation but was dropped due to insufficient data on what cardinal
direction landslides tended to occur on within the AOI and therefore an appropriate weighting
scheme would not be possible for the AOI. While data for LULC that covers the entire extent of
the AOI exists, it was not discovered until the analysis was already completed. Soil composition
had missing attributes in the attribute table that rendered the data inadequate for the purposes of
this project. Water saturation data was spread out throughout the AOI, but the data points were
sparce enough and spread out far enough that interpolation and extrapolation were not feasible.
The data for porosity downloaded in a format that required additional processing outside the
scope of this project to be used.

56

3.4. Weighted Overlay
Weighted overlay is but one of many MCDA methods that take into account criterion
ranks and value functions (Malczewski and Rinner 2015). Weighted linear combination,
weighted linear average, weighted summation, and simple additive weighting are other names
this method is known by. This process is straight forward in that it is based on assumptions of
additivity and linearity, in which the former indicates criteria are independently preferential of
each other. The latter assumes that the preferential weight of a criterion is constant on every level
it is considered in.
3.4.1. Reclassification of Criteria
For the weighted overlay, the data needed to be reclassified into a uniform scaling. The
Reclassification tool was used here to change the scales from their original values to a 1 to 5
ranking with 5 being of high importance and 1 being of low importance. The breakup of ranking
categories is described in further detail by criteria.
3.4.1.1. Elevation
Elevation was reclassified according to height above sea level. In this case, the higher the
elevation, the greater the ranking, as there is a rough correlation between landslide activity and
elevation (Clague and Roberts 2012). Table 5 shows the cutoff values for each rank, and Figure
15 is a visual representation of the reclassified elevation.
Table 4 Elevation reclassification ranks and cutoffs
Elevation Rank Cutoff
(meters) 5 <3500

4 3000
3 2500
2 2000
1 1500

57

Figure 14 Reclassified elevation

3.4.1.2. Slope
The reclassification of slope utilized the cutoffs of several studies (Chen and Chen 2021;
Chen and Li 2020; Feizizadeh and Blaschke 2013; Vakhshoori and Zare 2016) for the rank
cutoffs used in this project. The ranking of these cutoffs were based on Patil et al. (2020)’s
statistics for their study area. Table 6 breaks down the slope cutoffs for each rank and Figure 16
is a visual representation of reclassified slopes.

58

Table 5 Slope reclassification ranks and cutoffs
Slope Rank Cutoff
(degrees) 5 20-30

5 30-40
3 10-20
3 40-50
2 0-10
2 50-60
1 60-70
1 <70

Figure 15 Reclassified slope

59

3.4.1.3. Precipitation
The reclassification of precipitation was a simplistic scaling using the minimum and
maximum averaged rainfall in the shapefile. In order to reclassify precipitation, the shapefile had
to be rasterized. The Polygon to Raster tool was used for this process. The cutoffs were equally
distributed within that range with a higher rank given to increased average rainfall (Ali et al.
2020; Nandi and Shakoor 2009). The cutoffs are summarized in Table 7 and visualized in Figure
17.
Table 6 Precipitation reclassification ranks and cutoffs
Precipitation Rank Cutoff
(inches) 5 >50
4 50
3 37.5
2 25
1 12.5

60

Figure 16 Reclassified precipitation

3.4.1.4. Drainage Proximity
Reclassification of drainage proximity required the use of the Euclidean Distance tool.
The cutoffs were chosen based on work from Chen and Chen (2021), Erener et al. (2016), Patil et
al. (2019), and Zhao et al. (2014). Table 8 shows the ranking cutoffs and Figure 18 visualizes
these rankings.

61

Table 7 Drainage proximity reclassification ranks and cutoffs
Drainage Proximity Rank Cutoff
(meters) 5 3750
4 7500
3 11250
2 15000
1 >15000

Figure 17 Reclassified drainage proximity

62

3.4.1.5. Drainage Density
Drainage density was not as commonly used in the literature, though Roodposhti et al.
(2016) and Saha et al. (2005) use this criterion in their own studies. Table 9 delineates the cutoffs
for the different ranks, while Figure 19 visualizes these cutoffs.
Table 8 Drainage density reclassification ranks and cutoffs
Drainage Density Rank Cutoff
(meters/square meters) 5 >4

4 4

3 3
2 2
1 1

63

Figure 18 Reclassified drainage density

3.4.1.6. Lithology
Lithology required a different approach to reclassification because the symbology used
for this shapefile is not numerical. Ott (2020) summarized the erodibility of different lithologic
compositions, and the rankings were generated based on that author’s work. The shapefile used
for lithology also had to be converted to a raster in order to be reclassified, and therefore the
Polygon to Raster tool was used. The Fuzzy Membership tool was then used to assign rankings to
the different lithologies. The breakdown of ranks is shown in Table 10 and visualization is
featured in Figure 20.

64

Table 9 Lithology reclassification ranks and categories
Lithology Rank Category

5 Unconsolidated, undifferentiated
4 Sedimentary, carbonate
3 Igneous, volcanic
3 Sedimentary, undifferentiated
2 Igneous and Sedimentary, undifferentiated
2 Metamorphic and Sedimentary, undifferentiated
1 Igneous, intrusive
1 Metamorphic, gneiss
1 Metamorphic, undifferentiated
1 Sedimentary, clastic
N/A Water

65

Figure 19 Reclassified lithology

3.4.1.7. Lineament Proximity
The reclassification of lineaments was relatively simplistic, as it only required the use of
the Euclidean Distance tool to generate proximity buffers. The buffer cutoffs were guided by
studies such as Ali et al. (2020), Du et al. (2017), Erener et al. (2016), Nohani et al. (2019), Roy
et al. (2019), and Vojtekova and Vojtek (2020). Lineaments were used due to both the frequency
of its use in other studies, as well as ease of access to the data. The cutoffs are shown in Table 11
and displayed in Figure 21.

66

Table 10 Lineament reclassification ranks and cutoffs
Lineament Proximity Rank Cutoff
(meters) 5 6250

4 12500
3 18750
2 25000
1 >25000

Figure 20 Reclassified lineament proximity

3.4.1.8. Road Proximity
Road proximity was a frequently used criterion in the literature (Ali et al. 2020; Ayalew
and Yamagishi 2004; Feizizadeh and Blaschke 2013; Patil et al. 2019; Vakhshoori and Zare

67

2006). Reclassification was based distance from a road, and the cutoffs are shown in Table 12
and visualized in Figure 22.
Table 11 Road proximity reclassification ranks and cutoffs
Road Proximity Rank Cutoff
(meters) 5 5000

4 10000
3 15000
2 20000
1 >20000

Figure 21 Reclassified road proximity

68

3.4.2. Weighting of Criteria
The Weighted Overlay tool was used to generate a landslide susceptibility map using the
calculated ranks. The rankings in Tables 5-12 were used in the weighted overlay using the
reclassified rasters and summarizes the criteria properties that went into the weighted overlay.
Shit et al. (2016) utilized a weighted overlay for their study, and the equation for the weighted
overlay is as follows:

𝑆 =
∑ 𝑊 𝑖 𝑆 𝑖𝑗
∑ 𝑊 𝑖 ( 19 )
where S is the spatial unit value in the output map, Sij is the ith spatial class weight of jth factor
map, and Wi is the weight ith factor map.
The criteria were ranked based on how previous studies weighted their chosen criteria,
which is summarized in Table 13. Of the listed criteria, four studies provided the actual
percentages each criterion was given for their work. Not all of the criteria used in this project
were used within the four studies with given percentages. Feizizadeh and Blaschke (2014), for
example included every criterion used in this study except river density.

69

Table 12 Criteria used for reclassification
Study Slope Aspect
Precipi-
tation
Faults
Road
Proximity
River
Proximity
River
Density
Ali et al. 2020
-/6.43/12.13/
17.33/
23.27+
-/69.78/
143.39/
215.59/
287.79
-/800/900/
1000/1250
0/2498/
5333/8304/
11680
0/100/219/
362/557
0/103/211/
327/479
-
Ayalew and
Yamagishi
2004
-/4/16/31/
46+
W/NW/NE/
SW/N/E/S/
SE/F
- 50/100/150
50/100/
150
50/100/150 -
Bragagnolo et
al. 2020
-/7.07/13.2/
19.3/25.5+
0/60/180/
270/359
- - - - -
Chen and
Chen 2021
-/10/20/30/
40/50/+
W/NW/NE/
SW/N/E/S/
SE/F
- -
0/100/200/
300/400/+
0/200/400/
600/800/+
-
Chen and Li
2020
-/10/20/30/
40/50/60/
70/+
F/N/NE/E/S
E/S/SW/W/
NW
-/1221.86/
1502.36/
1954.28/
2639.95+
-/1000/2000/
3000/4000+
0/200/400/
600/800+
-/200/400/
600/800+
-
Chen et al.
2015
-/2/4/6/8/+ - - - - 200/400/600 -
Du et al. 2017 0/15/30/45+
F/N/NE/E/S
E/S/SW/W/
NW
-/1000/1500/
2000/2500/
3000+
-/500/1000/
1500/2000+
-/200/400/
600/800/
1000+
-/200/400/
600/800/
1000+
-
Ercanoglu
and
Gokceoglu
2015
- - - - - - -
Erener et al.
2016
0/5/10/15/
20/25/30/35/
40+
F/N/NE/E/S
E/S/SW/W/
NW
-
0/500/1500/
2500/3500/
5000/6500/
8000/9000+
0/100/200/
300/400/
500/700/
900+
0/100/200/
300/400/
500/750/
1000+
-
Feizizadeh
and Blaschke
2013
0/10.1/20.1/
30.1/40.1+
F/N/E/S/W
-/251/301/
350/401+
0/1001/
2001/3001/
4000+
0/26/51/
76/100+
0/51/101/
151/200+
-
Feizizadeh et
al. 2014
Continuous
F/N/NE/E/S
E/S/SW/W/
NW
Continuous Continuous Continuous Continuous -
Ghorbanzade
h et al. 2019
Continuous
F/N/NE/E/S
E/S/SW/W/
NW
- - - - -
Huggel et al.
2012
- - - - - - -
Kavzoglu et
al. 2013
0/5/10/15/
20/25/30/
35+
F/N/NE/E/S
E/S/SW/W/
NW
- 25 - 8
Korup 2012 - - - - - - -
Lehmann et
al. 2019
- - - - - - -
Lombardo
and Mai 2018
- - - - - - -
Mallick et al.
2018
Continuous
F/N/NE/E/S
E/S/SW/W/
NW
Continuous Continuous - Continuous -

70

Study Slope Aspect
Precipi-
tation
Faults
Road
Proximity
River
Proximity
River
Density
Nandi and
Shakoor 2009
0/7.1/14.1/
21.1/35.1/
42.1/49.1/
56.1/63.1+
-
-/92.7/93.98/
95.26/96.53/
97.80/99.07/
100.34/101.
61/102.67+
- -
0/401/801/
1201/1601/
2001/2401/
2801/3201/
3601
-
Nohani et al.
2019
0/5/15/30/
45+
F/N/E/S/W -
0/100/200/
300/400+
0/100/200/
300/400+
0/100/200/
300/400+
-
Patil et al.
2019
30/20/10/40/
-10/50/60/
70+
-
80.14/89.26/
97.24/107.9
6/120.95+
5000/10000/
15000/
20000/
25000/
30000/
35000
5000/10000/
15000/
20000/
25000
2000/4000/
6000/8000/
10000
-
Pawluszek
and
Borkowski
2016
Continuous
F/N/NE/E/S
E/S/SW/W/
NW
- -
-/50/100/
150/200+
-/50/100/
200/500+
-
Pham et al.
2020
0/14.54/
29.09/43.63/
58.18+
F/N/NE/E/S
E/S/SW/W/
NW
-
0/101/201/
301/401/
500+
0/101/201/
301/401/
500+
0/101/201/
301/401/
500+
-
`Pourghasemi
et al. 2012
0/6/16/31/
51/70+
N/NE/E/SE/
S/SW/W/N
W
-
0/100/200/
300/400+
0/100/200/
300/400/
500+
-/100/200/
300/400+
-
Pourghasemi
et al. 2020
Continuous
F/N/NE/E/S
E/S/SW/W/
NW
- Continuous Continuous Continuous -
Regmi et al.
2014
- - - - - - -
Rengers et al.
2016
- - - - - - -
Roccati et al.
2021
0/11/21/36/
51/76/
100+ (%)
F/N/NE/E/S
E/S/SW/W/
NW
- - <5/>5 <10 -
Roodposhti et
al. 2016
Continuous
F/N/NE/E/S
E/S/SW/W/
NW
Continuous Continuous Continuous Continuous Continuous
Roy and Saha
2019
0/9.32/
18.44/27.34/
36.66+
F/N/NE/E/S
E/S/SW/W/
NW
1877.38/
1991.97/
2090.45/
2167/
2239.06+
0/1.54/2085/
4.2/5.75+
0/1.74/3.94/
6.72/10.22+
0/0.42/1.1/
1.66/2.26+
-
Roy et al.
2019
0/9.32/
18.44/27.34/
36+
F/N/NE/E/S
E/S/SW/W/
NW
1877/1991/
2090/2167/
2239+
0/1.34/2.61/
3.92/5.51+
0/1.74/3.94/
6.79/10.28+
0/0.42/1.1/
1.66/2.24+
-
Saha et al.
2005
-15/16/26/
36/45+
F/N/NE/E/S
E/S/SW/W/
NW
-
-
504/505/100
9/1513/
2017/2521/
3025+
- -
-310/311/
620+
Saito et al.
2009
- - - - - - -
Schulz et al.
2009
- - - - - - -
Shou and
Chen 2021
-25/26/27/
28/29/30+
-
Unstable/
0/31/101/
151/201/
Stable
- - - -
Vahidnia et
al. 2010
Continuous Continuous - Continuous - Continuous -

71

Study Slope Aspect
Precipi-
tation
Faults
Road
Proximity
River
Proximity
River
Density
Vakhshoori
and Zare
2016
0/5/15/25/
35/50+
F/N/NE/E/S
E/S/SW/W/
NW
600/700/
800/900/
1000
0/200/400/
600/1000+
0/250/600/
750+
0/100/250/
500+
-
Vojtekova
and Vojtek
2020
0/2.1/5.1/
15.1/35+
F/N/NE/E/S
E/S/SW/W/
NW
-
-200/201/
401/601/
801+
-
-100/101/
201/301/
401+
-
Wayllace et
al. 2019
- - - - - -
Zhao et al.
2017
0/11/21/31/
41/50+
F/N/NE/E/S
E/S/SW/W/
NW
650/700/
750/800/
850/900/
950/1000/
1050+
- -
0/201/401/
600/1200+
-
Zhou et al.
2021
- - - - - - -
Zhu et al.
2014
- - - - - - -

The percentages were summed and normalized, as shown in Table 14, with missing
values normalized to 0.
Table 13 Calculation of normalizing percentages

Criteria
Feizizadeh and
Blaschke (2014)
Feizizadeh
et al. (2014)
Kavzoglu
et al. (2012)
Mallick
et al. (2018)
Elevation 0.02 - 0.0265 -
Slope 0.141 0.177 0.29 0.261
Precipitation 0.172 0.062 - 0.178
River Proximity 0.112 0.13 - -
River Density - 0.101 0.0355 -
Lithology 0.21 0.15 0.3074 0.056
Lineament
Proximity
0.124 0.092 - 0.128
Road Proximity 0.036 0.131 0.0181 0.103
Sum of Percentages 0.815 0.843 0.6775 0.726

ArcGIS Pro, however, requires that inputted percentages are entered as integers that sum
up to 100%, and the sum of the percentages shown in the “Percentage” column second from the
right in Table 15, which were rounded to two significant figures, did not equal 100% but rather
99%. To determine which criterion would receive the final 1%, the difference between the

72

percentages and the actual non-rounded results were calculated, and the criterion with the
greatest negative difference received the additional 1%. These differences are shown in the
“Difference” column to the far right in Table 15, and the criterion with the greatest negative
difference was river density.
Table 14 Normalization of percentages with missing rounding percent
Criteria, Normalized
Feizizadeh and
Blaschke
(2014)
Feizizadeh
et al.
(2014)
Kavzoglu
et al.
(2012)
Mallick
et al.
(2018)
Averaged
Percentage
Percentage Difference
Elevation 0.024540 0.000000 0.039114 0.000000 0.015914 0.02 0.004086
Slope 0.173006 0.209964 0.428044 0.359504 0.292630 0.29 -0.002630
Precipitation 0.211043 0.073547 0.000000 0.245179 0.132442 0.13 -0.002442
River Proximity 0.137423 0.154211 0.000000 0.000000 0.072909 0.07 -0.002909
River Density 0.000000 0.119810 0.052399 0.000000 0.043052 0.04 -0.003052
Lithology 0.257669 0.177936 0.453727 0.077135 0.241617 0.24 -0.001617
Lineament
Proximity
0.152147 0.109134 0.000000 0.176309 0.109397 0.11 0.000603
Road Proximity 0.044172 0.155397 0.026716 0.141873 0.092040 0.09 -0.002040
Sum of Percentages 1.000000 1.000000 1.000000 1.000000 1.000000 0.99 -0.010000

Once the additional 1% was added to river density, the total summed percentage resulted
in 100%. Table 16 shows the finalized percentages given for each criterion.
Table 15 Missing percent added to variable with maximum negative difference
Criteria, Finalized Percentage
Elevation 0.02
Slope 0.29
Precipitation 0.13
River Proximity 0.07
River Density 0.05
Lithology 0.24
Lineament Proximity 0.11
Road Proximity 0.09
Sum of Percentages 1.00

73

The finalized percentage results are shown in Table 17 along with the reclassified rankings for
each criterion.
Table 16 Finalized reclassification ranks and percentages of criteria for weighted overlay
Criteria Rank Category Percentage
Elevation 5 <3500
0.02
(meters) 4 3000
3 2500
2 2000
1 1500
Slope 5 20-30
0.29
(degrees) 5 30-40
3 10-20
3 40-50
2 0-10
2 50-60
1 60-70
1 <70
Precipitation 5 >50
0.13
(inches) 4 50
3 37.5
2 25
1 12.5
River Proximity 5 3750
0.07
(meters) 4 7500
3 11250
2 15000
1 >150000
River Density 5 >4
0.05
(length/unit area) 4 4
3 3
2 2
1 1

74

Criteria Rank Category Percentage
Lithology 5 Unconsolidated, undifferentiated
0.24
4 Sedimentary, carbonate
3 Igneous, volcanic
3 Sedimentary, undifferentiated
2 Igneous and Sedimentary, undifferentiated
2 Metamorphic and Sedimentary, undifferentiated
1 Igneous, intrusive
1 Metamorphic, gneiss
1 Metamorphic, undifferentiated
1 Sedimentary, clastic
N/A Water
Lineament 5 6250
0.11
Proximity 4 12500
(meters) 3 18750
2 25000
1 >25000
Road Proximity 5 5000
0.09
(meters) 4 10000
3 15000
2 20000
1 >20000

Reclassifying slope required knowledge of how frequently landslides occur at certain slope
gradients. To that end, Patil et al. (2019) calculated what percentage of landslides occur within
specific ranges, and the ranking scheme used correlates to frequency. Table 18 explains how
ranks were calculated, and Table 1 shows the slope gradient ranges with frequency percentages
and the corresponding ranks.
Table 17 Ranking for slope derived from frequency
Rank
Frequency
(percent)
1 <1%
2 1%-10%
3 10%-20%
4 20%-30%
5 >30%

75

Table 18 Slope categorization by frequency
Slope
(degrees)
Frequency
(percent)
Rank
0°-10° 7.000% 2
10°-20° 15.000% 3
20°-30° 30.000% 5
30°-40° 31.000% 5
40°-50° 14.000% 3
50°-60° 3.000% 2
60°-70° 0.200% 1
<70° 0.001% 1

3.5. Fuzzy Overlay
Fuzzy overlay is different from weighted overlay in that fuzzy memberships are used in
place of rankings. Different fuzzy memberships are used according to what aspect of a criterion
is emphasized, and these are then put into a fuzzy function to yield a result. Depending on the
type of function used, results again vary depending on what aspect of the results is to be
highlighted.
3.5.1. Fuzzy Membership Layers
In addition to the criteria used in this project being reclassified and therefore used in a
weighted overlay, the same criteria were also used in a fuzzy overlay. The criteria were
processed with the Fuzzy Membership tool in order to prioritize certain aspects of each one.
Different membership types were used for different criteria, based on what features ranked more
important than others. The Fuzzy Membership tool transformed the data into a 0 to 1 scale based
on what aspect of the criterion is considered more or less important, with 1 being most important
and 0 being least important. The fuzzy membership type for each criterion is listed in Table 20.

76

Table 19 Fuzzy membership for each criterion
Category Criteria Fuzzy Membership
Topography Elevation fuzzy large
Slope fuzzy Gaussian
Hydrology Precipitation fuzzy large
Drainage Proximity fuzzy small
Drainage Density fuzzy large
Subsurface Lithology fuzzy large using reclassified data
Lineament Proximity fuzzy small
Surface Road Proximity fuzzy small

The equation for the fuzzy Gaussian operator was written out by Kritikos and Davies
(2014) as:

𝜇 ( 𝑥 ) = 𝑒 −𝑓 1( 𝑥 −𝑓 2)
2

( 20 )
where μ(x) is the membership value of category x, which is the observed variable value or crisp
value, f1 is the standard deviation or spread, and f2 is the midpoint. The spread has a range from
0 to 1. All things being equal, the larger the f1 value, the narrower the spread, while the smaller
the f1 value, the wider the spread, as seen in Figure 23.

Figure 22 Visual of how fuzzy Gaussian transforms original data into a normal distribution with
different spread values (Kritikos and Davies 2014)

77

3.5.1.1. Fuzzy large elevation
Fuzzy large was chosen as the fuzzy membership for elevation. Fuzzy large emphasizes
larger values more heavily than smaller values, which would be equivalent to the reclassification
scheme used in the weighted overlay (Figure 24).

Figure 23 Fuzzy large elevation
3.5.1.2. Fuzzy Gaussian slope
Using the Fuzzy Membership tool on slope took some trial and error to ensure that the
range of slope gradients were properly weighted. As mentioned before in 3.4.2, Patil et al. (2019)
lists the frequency in which landslides occurred for given slope gradient ranges. Using their

78

landslide frequency statistics, the midpoint for the fuzzy Gaussian operator was 30, given that
Patil et al. (2019) highest landslide frequency percentages ranged from 20° – 40°. A spread of
0.1 – the default – was used (Figure 25).

Figure 24 Fuzzy Gaussian slope
3.5.1.3. Fuzzy large precipitation
The precipitation raster generated for the Weighted Overlay tool was also required for the
Fuzzy Membership tool. The fuzzy large operator was used to place emphasis on areas with
higher average rainfall (Figure 26).

79

Figure 25 Fuzzy large precipitation

3.5.1.4. Fuzzy small drainage proximity
Drainage proximity utilized the fuzzy small operator (Figure 27). In this way, the Fuzzy
Membership tool output would resemble that used for the weighted overlay in that the closer an
area is to a drainage channel, the higher the landslide susceptibility.

80

Figure 26 Fuzzy small drainage proximity

3.5.1.5. Fuzzy large drainage density
The output from using the Line Density tool to calculate drainage density resulted in a
raster that had large data gaps where there were no rivers or streams. Because the Fuzzy
Membership tool cannot run on a raster with data gaps, the Reclassification tool was used to
convert areas with no data into areas with a data value of 0. The resultant raster was reclassified
and then used to generate a fuzzy membership output using the fuzzy large operator (Figure 28).

81

Figure 27 Fuzzy large drainage density

3.5.1.6. Fuzzy large lithology
The lithology raster used in the weighted overlay was also required for the Fuzzy
Membership tool. Lithology also required preparation, though this was in the form of first
determining which types of lithology weathered and eroded easiest. This data came from Ott
(2020), and the different lithology types were reclassified using the same ranking method
mentioned in 3.5.1 (Figure 29).

82

Figure 28 Fuzzy large lithology
3.5.1.7. Fuzzy small lineaments
The fuzzy small operator was used for lineament proximity (Figure 30). This follows the
same rationale as the ranking used in the weighted overlay with a closer proximity indicating a
higher risk of landslide activity (Chen and Li 2020; Nohani et al. 2019; Patil et al. 2019; Saha et
al. 2005; Vakhshoori and Zare 2016).

83

Figure 29 Fuzzy small lineament proximity

3.5.1.8. Fuzzy small road proximity
Road proximity also utilized the fuzzy small operator (Figure 31). The choice of fuzzy
small for the fuzzy operator followed the same logic as the ranking in the weighted overlay of
closer proximity to roads means increased landslide susceptibility (Erener et al. 2016; Patil et al.
2019; Vakhshoori and Zare 2016).

84

Figure 30 Fuzzy small road proximity

3.5.2. Selection of Fuzzy Overlay Method
The Fuzzy Overlay tool offers five different fuzzy overlay operators: And, Or, Sum,
Product, and Gamma. The fuzzy membership-converted rasters (listed in Table 20) were used,
and the fuzzy Gamma operator was selected after examining each preliminary result using the
five operators on the default settings in ArcGIS Pro. The fuzzy Gamma function works by
multiplying the fuzzy Algebraic Sum with the fuzzy Algebraic Product, of which both are raised
to the power of γ. Vakhshoori and Zare (2016) explain how fuzzy Gamma is derived:

85

𝜇 𝑠 ( 𝑥 ) = 1 − ∏ 𝜇 𝑖 𝑛 𝑖 =1
( 𝑥 ) ( 21 )
𝜇 𝑝 ( 𝑥 ) = ∏ 𝜇 𝑖 𝑛 𝑖 =1
( 𝑥 ) ( 22 )

𝜇 𝛾 ( 𝑥 ) = [𝜇 𝑠 ( 𝑥 ) ]
𝛾 × [𝜇 𝑝 ( 𝑥 ) ]
1−𝛾 ( 23 )
where μs(x) is fuzzy Algebraic Sum, μp(x) fuzzy Algebraic Product, μ γ fuzzy Gamma, γ a
parameter in the range of 0 to 1, n the number of criteria being used, and μi(x) the map with a
fuzzy membership function. The closer the output is to 1 the more susceptible an area is to
having a landslide occur.
Changing γ values changes the results. A γ equal to 1 produces results that are identical to
the results of fuzzy Algebraic Sum, while a γ equal to 0 yields results equal to the results of fuzzy
Algebraic Product. Various values for γ were tested to determine what would be most
appropriate for the final fuzzy Gamma γ value.
Fuzzy Gamma was chosen as the representative fuzzy overlay output for several reasons.
Fuzzy And were not chosen because the data was not supposed to be completely exclusive by
requiring all of the criteria to be present. Fuzzy Or, on the other hand, was too inclusive and
skewed the results by overemphasizing areas with low landslide feasibility. Fuzzy Sum, similar
to Fuzzy Or, overemphasized areas with multiple lower-weighted criteria. Fuzzy Product was
similar to Fuzzy And but to a much more exclusive degree.
With fuzzy Gamma, the choice of a γ value of 0.9 was selected because it represented a
reasonable balance between the endmember fuzzy Sum and fuzzy Product. A γ value of 0.5,
despite being an even balance between fuzzy Sum and fuzzy Product, forced the output to lean
more heavily towards the fuzzy Product side. The 0.9 γ value output adequately highlighted

86

areas of higher susceptibility while neither overemphasizing nor underemphasizing areas of
lower landslide susceptibility.

87

Chapter 4 Results
Weighted and fuzzy overlays were used to generate the final results for this project. The
weighted overlay used criteria reclassified to an equal scale, and ranks for each criterion were
calculated based on the utilization frequency in other studies to determine the criterion’s
importance in landslide susceptibility. The fuzzy overlay ultimately used the fuzzy Gamma
operator on fuzzy membership criteria rasters and a chosen γ value to determine landslide
susceptibility.
Weighting plays a key role in both the weighted overlay and the fuzzy overlay, though
more so in the former than the latter. How ranks are determined is dependent on the availability
of data, the quality of the data, and how relevant a data is to the project in question. Due to the
fact that landslides tend to occur in areas with changes in elevation, most authors consider slope
to be of high importance with regards to assessing landslide susceptibility, and this decision is
reflected in how the slope criterion was weighted for this project for the weighted overlay. It is
therefore unsurprising the amount of influence it had on the weighted overlay result. As for the
fuzzy overlay result, all of the criteria were given equal weight given that the Fuzzy Overlay tool
only allows the user to add the necessary rasters and choose the desired fuzzy operator.
The results for the weighted overlay and fuzzy overlay were both expected and
unexpected in that the regions of higher predicted landslide activity for the most part overlap
each other. The unexpected aspects of the results were the differences in how areas with lower
landslide potential was weighted as well as how bias influences results, particularly in the
weighted overlay result.

88

4.1. Weighted Overlay Result
The weighted overlay result (Figure 32) reveals how the rankings interacted to highlight
areas of higher landslide susceptibility. The scale for the overlay result was set from 1 to 5 to
match the reclassification scheme, with 5 indicating high landslide susceptibility and 1 indicating
low susceptibility. The color bar ranges from 2 to 5, suggesting that with the weighting of the
criteria used in this project, the entirety of the area of interest is susceptible to some degree of
landslide activity.

89

Figure 31 Weighted overlay result
A comparison of the weighted overlay result with the USGS landslide inventory (Figure
33) reveals that the results from the weighted overlay do not entirely correlate with the landslide
inventory. Much of the high confidence locations in the inventory fall within areas calculated to

90

have medium landslide susceptibility according to the weighted overlay results. The weighted
overlay result correlated much better with locations flagged by the USGS as having possible or
probable landslides in the area.

91

Figure 33 Comparison of (A) the weighted overlay result with (B) the USGS’s landslide inventory locations superimposed

92

Both slope (29%) and lithology (24%), as the two most heavily weighted criteria, greatly
affected the output. The mountainous areas to the west have the highest landslide susceptibility
outputs, where the slope gradients are highest, with susceptibility decreased to the east until the
area relatively flattens out where central Denver is located. Likewise, the fluvial deposits east of
where central Denver is situated had higher susceptibility scores due to the fact that those rivers
and streams are actively depositing unconsolidated sediment eroded from the Front Range.
Precipitation (13%) certainly caused an increase in landslide susceptibility, where higher
amounts of rainwater are more likely to turn dry soil into slippery mud. Lineament proximity
(11%), with the majority of known mapped lineaments located within the Front Range proper,
primarily boost landslide susceptibility in areas that statistically have higher elevation and
steeper slope gradients when compared to the relatively lower elevation and flatter area that
central Denver covers. Road and river proximity (9% and 7%, respectively) both cover much of
the area of interest, while river density (5%) highlights confluence areas and areas with a higher
density of rivers and streams. Elevation (2%) plays a role in landslide susceptibility in that
generally speaking, the higher the elevation, the more likely a landslide may occur.
Of note are the higher-ranked areas that more susceptible areas scattered around the Front
Range peaks. With regards to slope, higher landslide susceptibility appears to be closely
associated with slopes that range from 20° to 50° (Figure 34). This agrees with Patil et al.
(2020)’s study, which indicated that slope gradients between 20° to 40° tend to have the highest
landslide frequencies.

93

Figure 34 Slope and weighted overlay comparison: (A) slope; (B) slope with weighted overlay result semitransparent and
superimposed; (C) weighted overlay result

94

The weighted overlay result from the slope criterion’s weight was expected; the result
from the lithology criterion’s weight was more of a surprise given how much weight the
unconsolidated sediment of the fluvial flood plains in the Denver metroplex added to the result
with one river in particular – the St. Vrain River (Figure 35). The expectation was that the flood
plains would not rank so highly, given that the area lies within the urbanized corridor that
consists of Denver, Boulder, and Fort Collins. Unconsolidated sediment on the slopes of the
Front Range – deposited in the incised valleys due to the drainage systems mentioned earlier –
also increased the ranking of landslide frequency. Cross inspection of the weighted overlay
result to elevation, precipitation, drainage systems, lineaments, and roads yielded no easily
visible correlations.

95

Figure 35 Lithology and weighted overlay comparison: (A) lithology; (B) lithology with weighted overlay result semitransparent
and superimposed; (C) weighted overlay result

96

4.2. Fuzzy Overlay Result
The fuzzy overlay result (Figure 36) yielded a very different output. The γ value chosen
that best displays the results of the fuzzy overlay is 0.9. The scale used for fuzzy overlay covers a
range from 0 to 1, with 0 indicating very low landslide susceptibility and 1 indicating very high
landslide susceptibility. Unlike the weighted overlay result, the fuzzy overlay result displays a
large majority of the AOI as not particularly susceptible to landslide activity. The fact that the
much larger swaths of land have a susceptibility rating closer to 0 is indicative of the fact that the
fuzzy overlay’s weighting result is much stricter in terms of which areas are considered more
susceptible to landslides than others.

97

Figure 32 Fuzzy overlay result using the fuzzy Gamma operator with a γ value of 0.9
The comparison between the fuzzy overlay result and the USGS’s landslide inventory
displays greater commonality with regards to predicted landslide hazard locations (Figure 37).
Areas marked as high confidence overlap with regions the fuzzy overlay result considered to
have higher landslide susceptibility. To a lesser degree, areas noted as probable or possible with

98

regards to landslide occurrence overlap with lower landslide susceptibility in the fuzzy overlay
result. This is indicative of the fact that the weighting generated through the various fuzzy
membership operators seems to coincide with the predictions of external sources.

99

Figure 37 Comparison of (A) the fuzzy overlay result with (B) the USGS’s landslide inventory locations superimposed

100

Two criteria in particular seem to have a higher correlation to increased landslide
susceptibility than the rest: slope and drainage systems. Landslide susceptibility tends to be
higher in areas with more ideal slope gradients (20°-40°), and a side-by-side comparison is
displayed in Figure 38.

101

Figure 38 Slope and fuzzy overlay comparison: (A) slope; (B) slope with fuzzy overlay result superimposed; (C) fuzzy overlay
result

102

That localized areas of particularly high landslide susceptibility also tend to coincide with
incised valleys from said drainage systems is likely due to increased erosion that comes with
closer proximity to drainage systems (Figure 39). The other criteria do not seem to have as much
of an impact on the result than the aforementioned two.

103

Figure 39 Drainage systems and fuzzy overlay comparison: (A) drainage systems; (B) drainage systems with fuzzy overlay result
semitransparent and superimposed; (C) fuzzy overlay result

104

Chapter 5 Discussion and Conclusion
This project looked at two out of many different MCDA weighting schemes for landslide
susceptibility mapping: weighted overlay and fuzzy overlay. These two methods were selected
because there exists little in the way of direct comparisons between the two in literature with
regards to landslide susceptibility. The ability for the software used to be able to handle such
analyses had to also be taken into consideration. The various criteria used for this project was
selected based on data availability and resolution, as well as what authors deemed important
enough to include in their own studies.
Bias and limitations play a significant role in how they affect the results as they may
emphasize certain criteria more than others. The fact that both rankings and fuzzy memberships
may be chosen differently depending on the user means that the results, even if using the same
data and same methodology, yields different outputs. This in turn affects how people’s
understanding of the data is perceived, which also influences how and where people may live
and what kinds of laws might be put in place to protect them.
5.1. Bias and Limitations
The results of the weighted overlay and the fuzzy overlay had both similarities and
differences that were visually distinct, though these distinctions were partially a result of the
inclusion and weighting of specific criterion. The results revealed that the selection of the
method of weighting as well as how variables are weighted play important roles in result
expectations. This, in turn, hints at a cyclical influence between how weighting biases affect
results, and how results can generate biases that affect the decision-making process with regards
to landslide susceptibility mapping.

105

As noted in Chapter 2, there are a wide variety of variables that were used in landslide
susceptibility studies. Because of the number of criteria collectively used, it is not possible for
any given study to include all of them, and it is also likely that this project did not cover the full
range of criteria used in landslide susceptibility studies. It is impractical for any study to include
every possible variable used in landslide susceptibility mapping – depending on the size of the
AOI, the amount of data required could be astronomically high.
Data such as elevation is readily accessible across the globe, and in varying resolutions.
The suitability of a given resolution is dependent on the size of the AOI and the scale of the
subject(s) of analysis. DEMs may be generated from satellite imagery, topographic maps, and
LiDAR – though the former two are more likely to be “bare-earth” models while the latter
probably includes surficial features such as buildings and trees (USGS, n.d.). From this,
derivative variables such as slope and aspect may be generated in resolutions equal to the
original DEM. Within Esri ArcGIS Pro’s Spatial Analyst toolbox are fourteen tools specifically
made to generate some derivative layer of elevation in the Surface toolset, not including a tool to
manually add supplementary information to a surface layer. On the other hand, other variable
data such as desiccation height and undesiccation height, water condition, landslide-rainfall
index are specific to individual studies in which the study author(s) likely generated the data
themselves. Such data on a wider scale would not be available as it does not exist outside of the
AOIs of the studies in question.
In addition to variable resolutions, sourcing data is also a necessary consideration. With
regards to the data, it can be either readily available to the public or proprietary, in which a user
would need to purchase access to said data. Data from certain national organizations, such as the
USGS, USDA, and NOAA, as well as national, state, and local governments, and universities

106

usually offer some data for public consumption. The data that is available may or may not be
cover the study area in question, and it may also not be of viable quality depending on the project
scope. The format of the data also had to be taken into consideration, as data incompatible with
the software being utilized is not useful to the study. In the event that a data format could be
converted without a loss in quality, such measures are likely to be taken in order to include that
criterion in a study.
The eight criteria used in this project were chosen from a combination of factors. The
landslide inventory from the USGS was used to search for locations within the US that had high
landslide incidences. Once the location was chosen, searching for data that would be suitable for
the AOI began, and the criteria listed in Chapter 2 were searched for. Of all the criteria listed in
Chapter 2, the key data to any landslide susceptibility study was elevation – without elevation,
derivatives such as slope would not exist. Apart from one study out of the thirty-four examined
in this project, elevation was included as a criterion for their analysis. The second key criterion
was aspect, of which thirty-one studies included as part of their dataset, followed by slope at
twenty-nine studies.
Aspect was not included in this project because of a lack of information regarding which
cardinal directions landslides were more likely to occur on. LULC data was found within the
AOI, but the extent did not cover the whole of the area in question and ultimately had to be
discarded. Soil composition, water saturation, and porosity were downloaded in unusable formats
that could not be converted and were similarly deemed unnecessary. Census tract data were
readily available but a method of distilling census data into spatially representative population
density data was beyond the scope of this project, and therefore census tract data was abandoned.

107

Choosing rankings for the weighted overlay, and fuzzy memberships and γ values for the
fuzzy overlay ingrain a subjective bias in the results. This is due to the fact that, depending on
what criteria the user or decision maker deems more important than others, some variables may
be weighted more heavily in the weighted overlay. The fact that no means of objectively
weighting variables exists only emphasizes the fact that bias is an implicit part of the results of
any analysis that uses a weighting system. The same bias comes into play when fuzzy
memberships are chosen for the individual criterion, as well as which fuzzy overlay operator is
used. Subjectivity is therefore introduced by the selection of fuzzy membership by tweaking its
parameters, and the choice of fuzzy overlay operator is also dependent on what sort of outcome
the endmember user is expecting.
With regards to this project, there exist limitations that needed to be acknowledged from
the onset. Software choice is somewhat dictated by the analysis methodology to be used and vice
versa – if the software cannot support the desired methodology, then either a different
methodology must be chosen or created, or a change in software is required so that one capable
of handling the desired methodology is selected. There is also the possibility that the number of
variables must also be constrained to a quantity that the chosen analysis software can handle.
Having too many variables would inevitably cause the processing time to slow dramatically. The
same can be said of data size – datasets that are too large processes much slower than smaller
datasets, and this could be an issue if projects are time-sensitive.
Data quality or resolution can drastically affect the results of an analysis. Working a
small study area with data resolutions too coarse for the AOI may render any results unsuitable
for further analysis or future work. Quality of the data in use is important, as the higher the
resolution, the finer the detail and more comprehensive the results are. The downside of having

108

finer quality data is, as mentioned previously, the size of the data in question. File sizes too large
may be a detriment to hardware storage and memory capacities – perhaps even cloud-based
storage capacities – as well as processing time and speed.
Data compatibility is an issue that needs to be addressed whenever a project requires
data. Different sources of data provide different formats that may or may not work with certain
software. It is important to determine then what data is necessary and what is superfluous – upon
which data deemed essential might require a conversion from its original format into another that
is software compatible. Should a transformation into a compatible format fail, then either a
separately sourced data set needs to be found, or the criterion that uses the data in question is
discarded from the project.
5.2. Societal Impacts
Landslide susceptibility mapping is an important facet of disaster planning and
management. As populations spread outward and encroach on the foothills of mountains, people
migrate ever closer to areas prone to landslide activity. This movement towards landslide-prone
areas means that more lives, as well as property and infrastructure, are at risk. With the rise in
global population and climatic changes due to global warming, predicting potential landslide
locations becomes increasingly important in regions already prone to landslides.
While the global average temperature is slowly climbing, local average temperatures in
parts of the world have dropped. These localized changes are exaggerated at the poles and
regions of high altitude, and it is for this reason that landslide susceptibility mapping –
particularly in mountainous regions with nearby urban centers – is such an important of urban
planning. The increasing global population has led to wider, more extensive urban sprawl. As
more people move to larger cities and, in turn, choose to live in the suburbs, cities that are

109

already in close proximity to mountain ranges are spreading upwards into the foothills of said
mountains. This puts not only people, but infrastructure and property as well, in regions of higher
landslide risk. Roads that traverse through and along mountains require level surfaces, and so
slopes are excavated in order to accommodate construction. This excavation, however, comes at
the cost of slope stability, thereby increasing the risk of a landslide occurring.
The spreading of population centers and the creation of roads also brings along other
forms of infrastructure, such as buildings both residential and commercial, electricity, water, and
communications networks, and potentially other methods of rapid transit. Most of these
constructions require stable, if not level, foundations for structural longevity and integrity, and
destabilizing the soil under and around these new constructions must be taken into consideration.
Landslide susceptibility mapping studies undertaken to better understand the slope stability of
the construction of these new construction projects may also play a feature role in legislation to
ensure the safety of both the infrastructure being built, as well as the people working on the
construction projects. This may be further extended to legislation regarding the protection of
both residents and commercial businesses, which ultimately means greater safety precautions are
put into place for the protection of those with a landslide hazard zone.
The following figures respectively showcase details for both the weighted and fuzzy
overlays in relation to Front Range and the Denver metroplex. The overlay layers are
semitransparent to allow visibility of not only Denver and its satellite cities, but also a few of the
major roads and highways that run through the cities and the hillshade from the basemap for a
sense of relative relief.
The weighted overlay result (Figure 40) shows higher susceptibility within the Front
Range where drainage systems exist. The weighted overlay result suggests that for the most part,

110

residents of Denver and its surroundings are most susceptible to landslides near where the Clear
Creek drainage system exits the Front Range – the Clear Creek drainage system is the fluvial
system that crosses the center of Figure 40. This means that when there is precipitation upstream
in the Front Range, residents in western Denver are at higher risk, and while cities like Golden
and Boulder are established along the edge of the Front Range, lawmakers may mitigate damage
and injury by passing laws that require property and infrastructure to be built a certain distance
away from the base of the mountains. While the weighted overlay results implies that there is an
increase chance of landslide activity to the north of Denver due to the presence of unconsolidated
sediment, the fact that there is minimal slope change drops the chances of a landslide
significantly.

111

Figure 33 Detail of weighted overalay

112

The fuzzy overlay result (Figure 41) is similar to the weighted overlay with regards to its
characterization of the Clear Creek drainage system, but there is increased emphasis on smaller
drainage systems to the north that the weighted overlay does not demonstrate to the same extent.
The result from the fuzzy overlay indicates that a larger percentage of the population in the
Denver metroplex are more vulnerable to landslide activity. Unlike the weighted overlay, the
fuzzy overlay result does not consider unconsolidated sediment within the Denver metroplex to
be a landslide hazard for reasons stated before. In this way, the fuzzy overlay’s result is more
accurate than the weighted overlay.

113

Figure 34 Detail of fuzzy overlay

114

5.3. Future Application and Work
Further work could be done with this project, as the results may be considered
preliminary if expanded upon. The results of this project could be the first step in creating a more
robust project for landslide susceptibility mapping in the Front Range. In refining this project,
more variables might be added to increase the accuracy and precision of potential landslide
locales. While every possible criterion should not be included, select data that was not included
here, including soil composition, water saturation, LULC, population/census tracts, and aspect.
Each of these variables were considered and ultimately discarded for reasons listed in Chapter 3,
but their inclusion in this project would have enriched the results immensely.
The results from this project could also be used as a means of comparing methodologies.
Provided the variables used are weighted as evenly as possible across methods, an analysis of
which method works better or worse for landslide susceptibility analysis might be useful in
narrowing down choices for similar projects in the future. As mentioned in Chapter 2, a variety
of methods have been used for landslide susceptibility mapping in the past, and the results of this
project might be compared to those other methodologies if they use the same data and AOI. It is
also possible for the results of this project to be merged into a larger, state-wide or even nation-
wide landslide susceptibility map, one that could provide greater detail if the resolution of this
project’s output is of better quality than what is otherwise available.
If given the opportunity or time to further or improve upon this project, the data that had
been discarded would have been assimilated into the project. If available, information regarding
wind direction would have also been included, as the direction of prevailing winds have a
significant impact on climate – and more specifically, precipitation, weathering, and erosional
patterns, as well as amount of solar radiation an area receives. It is possible that other weighting

115

schemes might have been included to create a more robust comparison to better flesh out both
the effectiveness and accuracy of the weighting methods as well as incite deeper discussion on
the relationship between variables and bias. These would be the primary reasons to further
pursue research through this project, which would, in and of itself, be a hefty undertaking.

116

References
Ahmed, B. 2015. “Landslide Susceptibility Mapping Using Multi-Criteria Evaluation
Techniques in Chittagong Metropolitan Area, Bangladesh.” Landslides 12: 1077–1095.
https://doi.org/10.1007/s10346-014-0521-x.
American Geosciences Institute. n.d. “How Much Do Landslides Cost the U.S. in Terms of
Monetary Losses?” Accessed July 15, 2021.
https://www.americangeosciences.org/critical-issues/faq/how-much-do-landslides-cost-
terms-monetary-losses.
Amitrano, D., R. Guida, D. Dell’Aglio, G. Di Martino, D. Di Martire, A. Iodice, M. Costantini,
F. Malvarosa, and F. Minati. 2019. “Long-Term Satellite Monitoring of the Slumgullion
Landslide Using Space-Borne Synthetic Aperture Radar Sub-Pixel Offset Tracking.”
Remote Sensing 11: 369. https://doi.org/10.3390/rs11030369.
Aoudia, A., G. Costa, and P. Suhadolc. 1970. "GIS for Earthquake Hazards Mitigation in the
Friuli Area (NE Italy)." WIT Transactions on Information and Communication
Technologies 21.
Baidya, P., D. Chutia, S. Sudhakar, C. Goswami, J. Goswami, V. Saikhom, P.S. Singh, and K.K.
Sarma. 2014. “Effectiveness of Fuzzy Overlay Function for Multi-Criteria Spatial
Modeling—A Case Study on Preparation of Land Resources Map for Mawsynram Block
of East Khasi Hills District of Meghalaya, India.” Journal of Geographic Information
System 6: 605–612. https://doi.org/10.4236/jgis.2014.66050.
Blońska, E., J. Lasota, W. Piazczyk, M. Wiecheć, A. Klamerus-Iwa. 2018. “The Effect of
Landslide on Soil Organic Carbon Stock and Biochemical Properties of Soil.” Journal of
Soils and Sediments 18: 2727–2737. https://doi.org/10.1007/s11368-017-1775-4.
Bogaard, T., Y. Gugliem, V. Marc, C. Emblanch, C. Bertrand, and J. Mudry. 2007.
“Hydrogeochemistry in Landslide Research: A Review.” Bulletin de la Société
Géologique de France 178, no. 2: 113–126. https://doi.org/10.2113/gssgfbull.178.2.113.
Buckley, J.J. 1985. “Fuzzy Hierarchical Analysis.” Fuzzy Sets and Systems 17, no. 3: 233–247.
https://doi.org/10.1016/0165-0114(85)90090-9.
Carver, S.J. 1991. “Integrating Multi-Criteria Evaluation with Geographical Information
Systems.” International Journal of Geographical Information System 5, no. 3: 321–339.
https://doi.org/10.1080/02693799108927858.
Çellek, S. 2013. “Landslide Susceptibility Analysis of Sinop-Gerze Region.” PhD diss.
Karadeniz Technical University, Turkey.
—. 2020. “Effect of the Slope Angle and Its Classification on Landslide.” Natural Hazards and
Earth System Sciences: Preprint. https://doi.org/10.5194/nhess-2020-87.

117

—. 2021 "The Effect of Aspect on Landslide and Its Relationship with Other Parameters" in
Landslides, edited by Y. Zhang, Q. Cheng. London: IntechOpen.
https://doi.org/10.5772/intechopen.99389.
Cerri, R.I., V. Rosolen, F.A.G.V. Reis, A.J. Pereira Filho, F. Vemado, L. do Carmo Giordano,
and B. Maques Gabelini. 2020. “The Assessment of Soil Chemical, Physical, and
Structural Properties as Landslide Predisposing Factors in the Serra do Mar Mountain
Range (Caraguatatuba, Brazil).” Bulletin of Engineering Geology and the Environment
79: 3307–3320. https://doi.org/10/1007/s10064-020-01781-1.
Çetinkaya, C., E. Özceylan, M. Erbaş, and M. Kabak. 2016. “GIS-Based Fuzzy MCDA
Approach for Siting Refugee Camp: A Case Study for Southeastern Turkey.”
International Journal of Disaster Risk Reduction 18: 218–231.
https://doi.org/10.1016/j.ijdrr.2016.07.004.
Che, V.B., K. Fontijn, G.G.J. Ernst, M. Kervyn, M. Elburg, E. Van Ranst, and C.E. Suh. 2011.
“Evaluating the Degree of Weathering in Landslide-Prone Soils in the Humid Tropics:
The Case of Limbe, SW Cameroon.” Geoderma 170: 378–389.
https://doi.org/10.1016/geoderma.2011.10.013.
Chen, Y., R. Liu, D. Barrett, L. Gao, M. Zhou, L. Renzullo, and I. Emelyanova. 2015. “A Spatial
Assessment Framework for Evaluating Flood Risk Under Extreme Climates.” Science of
the Total Environment 538: 512–523. https://doi.org/10.1016/j.scitotenv.2015.08.094.
Clague J.J. and N.J. Roberts. 2012. “Landslide Hazard and Risk.” in Landslides: Types,
Mechanisms and Modeling. Cambridge: Cambridge University Press.
https://doi.org/10.1017/CBO9780511740367.
Clague, J.J. and D. Stead. 2012. Landslides: Types, Mechanisms and Modeling. Cambridge:
Cambridge University Press. https://doi.org/10.1017/CBO9780511740367.002.
Coe, J.A., J.W. Godt, R.L. Baum, R.C. Buckham, and J.A. Michael. 2004. “Landslide
Susceptibility from Topography in Guatemala.” in Landslides: Evaluation and
Stabilization. Edited by W.A. Lacerda, M. Ehrlich, S.A.B. Fontura and A.S.F. Sayão. 69–
78. London: Taylor & Francis Group. https://doi.org/10.1201/b16816-8.
Colorado Geological Survey. n.d. “Landslides.” Accessed August 13, 2021.
https://coloradogeologicalsurvey.org/hazards/landslides/.
Cruden, D.M., and D.J. Varnes. 1996. “Landslide Types and Processes. Landslides: Investigation
and Mitigation.” Special Report 247: 36–75, Washington: Transportation Research
Board.
Dağ, S., 2007. “Landslide Susceptibility Analysis of Çayeli (Rize) and Its Surrounding by
Statistical Methods.” PhD diss. Karadeniz Technical University, Turkey.

118

Daly, C., R.P. Nielson, and D.L. Phillips. 1993. “A Statistical-Topographic Model for Mapping
Climatological Precipitation over Mountainous Terrain.” Journal of Applied Meteorology
33: 140–158.
De Montis, A., P. De Toro, B. Droste-Franke, I. Omann, and S. Stagl. 2004. "Assessing the
Quality of Different MCDA Methods." in Alternatives for Environmental Valuation:
115–149. London: Routledge. https://doi.org/10.4324/9780203412879.
Dempsey, C. 2020. “What is the Difference Between Elevation and Altitude?” Last modified
March 23, 2023. https://www.geographyrealm.com/what-is-the-difference-between-
elevation-and-altitude/.
Denver Regional Council of Governments. n.d. “Data” Accessed: July 24, 2022.
https://data.drcog.org/data?page=1&q=&sort=title.
Di Luzio, M. 2010. “Seamless Daily Precipitation for the Conterminous United States.”
Accessed June 5, 2022.
https://datagateway.nrcs.usda.gov/GDGHome_DirectDownLoad.aspx.
Di Maio, C., J. De Rosa, R. Vassallo, R. Coviello, and G. Macchia. 2020. “Hydraulic
Conductivity and Pore Water Pressures in a Clayey Landslide: Experimental Data.”
Geosciences 10, no. 102: 13. https://doi.org/10.3390/geosciences10030102.
Donnarumma, A., P. Revellino, G. Grelle, and F.M. Guadango. 2013. “Slope Angle as Indicator
Parameter of Landslide Susceptibility in a Geologically Complex Area.” in Landslide
Science and Practice, edited by C. Margottini, P. Canuti, and K. Sassa 425–433. Berlin:
Springer-Verlag Berlin Heidelberg. https://doi.org/10.1007/978-3-642-31325-7_56.
Du, G.L., Y.S. Zhang, J. Iqbal, Z.H. Yang, and X. Yao. 2017. “Landslide Susceptibility Mapping
Using an Integrated Model of Information Value Method and Logistic Regression in the
Bailongjiang Watershed, Gansu Province, China.” Journal of Mountain Science 14, no. 2:
249–268. https://doi.org/10.1007/s11629-016-4126-9.
Eberhardt, E. 2012. “Landslide Monitoring: The Role of Investigative Monitoring to Improve
Understanding and Early Warning of Failure.” in Landslides: Types, Mechanisms and
Modelling, edited by J.J. Clague and D. Stead. 222–234. Cambridge: Cambridge
University Press. https://doi.org/10.1017/CBO9780511740367.020.
Erbaş, M., M. Kabak, E. Özceylan, and C. Çetinkaya. 2018. “Optical Siting of Electric Vehicle
Charging Stations: A GIS-Based Fuzzy Multi-Criteria Decision Analysis.” Energy 163:
1017–1031. https://doi.org/10.1016/j.energy.2018.08.140.
Ercanoglu, M. and C. Gokceoglu. 2002. “Assessment of Landslide Susceptibility for a
Landslide-Prone Area (North of Yenice, NW Turkey) by Fuzzy Approach.”
Environmental Geology 41: 720–730. https://doi.org/10.1007/s00254-001-0454-2.
Erener, A., A. Mutlu, and H.S. Düzgün. 2016. “A Comparative Study for Landslide
Susceptibility Mapping Using GIS-Based Multi-Criteria Decision Analysis (MCDA),

119

Logistic Regression (LR) and Association Rule Mining (ARM).” Engineering Geology
203: 45–55. https://doi.org/10.1016/j.enggeo.2015.09.007.
Feizizadeh, B. and T. Blaschke. 2013. “GIS-Multicriteria Decision Analysis for Landslide
Susceptibility Mapping: Comparing Three Methods for the Urmia Lake Basin, Iran.”
Natural Hazards 65: 210 –2128. https://doi.org/10.1007/s11069-012-0463-3.
—. 2014. “An Uncertainty and Sensitivity Analysis Approach for GIS-Based Multicriteria
Landslide Susceptibility Mapping.” International Journal of Geographical Information
Science 28, no. 3: 610–638. https://doi.org/10.1080/13658816.2013.869821.
Florinsky, I.V., 2016. “Chapter 14 – Lineaments and Faults.” in Digital Terrain Analysis in Soil
Science and Geology (Second Edition), edited by I.V. Florinsky. 353–376. London:
Elsevier. https://doi.org/10.1016/B978-0-12-804632-6.09991-0.
Francioni, M., F. Calamita, J. Coggan, A. De Nardis, M. Eyre, E. Miccadei, T. Piacentini, D.
Stead, and N. Sciarra. 2019. “A Multi-Disciplinary Approach to the Study of Large Rock
Avalanches Combining Remote Sensing, GIS and Field Studies: The Case of the Scanno
Landslide, Italy.” Remote Sensing 11: 1570. https://doi.org/10.3390/rs11131570.
Froude, M.J. and D.N. Petley. 2018. “Global Fatal Landslide Occurrence from 2004 to 2016.”
Natural Hazards and Earth System Sciences 18: 2161–2181.
https://doi.org/10.5194/nhess-18-2161-2018.
Ghorbanzadeh, O., T. Blaschke, K. Gholamnia, S.R. Meena, D. Tiede, and J. Aryal. 2019.
“Evaluation of Different Machine Learning Methods and Deep-Learning Convolutional
Neural Networks for Landslide Detection.” Remote Sensing 11: 196.
https://doi.org/10.3390/rs11020196.
The GIS Encyclopedia. n.d. “Aspect (geography).” Last modified October 23, 2013, 02:33.
http://www.wiki.gis.com/wiki/index.php/Aspect_(geography).
The GIS Encyclopedia. n.d. “Elevation.” Last modified November 27, 2011, 01:23.
http://www.wiki.gis.com/wiki/index.php/Elevation.
Glade, T., M. Anderson, and M.J. Crozier. 2005. Landslide Hazard and Risk. Chichester,
England: John Wiley & Sons, Ltd. https://doi.org/10.1002/9780470012659.
Glade, T. and M.J. Crozier. 2005 “Landslide Hazard and Risk: Issues, Concepts and Approach.”
in Landslide Hazard and Risk, edited by T. Glade, M. Anderson, and M.J. Crozier. 1–40.
West Sussex: John Wiley & Sons, Ltd. https://doi.org/10.1002/9780470012659.ch1.
Gomberg, J., P. Bodin, W. Savage, and M.E. Jackson. 1995. “Landslide Faults and Tectonic
Faults, Analogs?: The Slumgullion Earthflow, Colorado.” Geology 23, no. 1: 41–44.
Gomberg, J., W. Schulz, P. Bodin, and J. Kean. 2011. “Seismic and Geodetic Signatures of Fault
Slip at the Slumgullion Landslide Natural Laboratory.” Journal of Geophysical Research
116: B09404. https://doi.org/10.1029/2011JB008304.

120

Hasekioğullari, G. D. 2011. “Assessment of Parameter Effects in Eroducing Landslide
Susceptibility Maps.” MS thes. Hacettepe University, Turkey.
Hasanloo, M., P. Pahlavani, and B. Bigdeli. 2019. “Flood Risk Zonation Using a Multi-Criteria
Spatial Group Fuzzy-AHP Decision Making and Fuzzy Overlay Analysis.” Remote
Sensing and Spatial Information Sciences 42: 18. https://doi/org/10.5194/isprs-archives-
XLII-4-W18-455-2019.
Hassan, I., M.A. Javed, M. Asif, M. Luqman, S.R. Ahmad, A. Ahmad, S. Akhtar, and B.
Hussain. 2020. “Weighted Overlay Based Land Suitability Analysis of Agriculture Land
in Azad Jammu and Kashmir Using GIS and AHP.” Pakistani Journal of Agricultural
Research 57, no. 6: 1509–1519. https://doi.org/10.21162/PAKJAS/20.9507.
Highland, L.M. 2008. The Landslide Handbook —A Guide to Understanding Landslides. Reston,
VA: U.S. Geological Survey. https://pubs.usgs.gov/circ/1325/.
Hu, X. and R. Bürgmann. 2020. “Rheology of a Debris Slide From the Joint Analysis of
UAVSAR and LiDAR Data.” Geophysical Research Letters 47: e2020GL087452.
https://doi.org/10.1029/2020GL087452.
Huang, I.B., J. Keisler, and I. Linkov. 2011. “Multi-Criteria Decision Analysis in Environmental
Sciences: Ten Years of Applications and Trends.” Science of the Total Environment 409:
3578–3594. https://doi.org/10.1016/j.scitotenv.2011.06.022.
Huggel, C., N. Khabarov, O. Korup, and M. Obersteiner. 2012. “Physical Impacts of Climate
Change on Landslide Occurrence and Related Adaptation.” in GIS Landslide, edited by
H. Yamagishi and N.P. Bhandary, 121–133. Tokyo: Springer Japan.
https://doi.org/10.1017/CBO9780511740367.003.
Islam, Z. and G. Metternicht. 2005. “The Performance of Fuzzy Operators on Fuzzy
Classification of Urban Land Covers.” Photogammetric Engineering & Remote Sensing
71, no. 1: 59–68. https://doi.org/10.14358/PERS.71.1.59.
Iwahashi, J., S. Watanabe, and T. Furuya. 2003. “Mean Slope-Angle Frequency Distribution and
Size Frequency Distribution of Landslide in Higashikubiki Area, Japan.” Geomorphology
50: 394–364. https://doi.org/10.1016.S0169-555X(02)00222-2.
Jankowski, P. 1995. “Integrating Geographical Information Systems and Multiple Criteria
Decision-Making Methods.” International Journal of Geographical Information Systems
9, no. 3: 251–273. https://doi.org/10.1080/02693799508902036.
Johnson, B., C. Campbell, and H.J. Melosh. 2016. “The Reduction of Friction in Long Runout
Landslides as an Emergent Phenomenon.” Journal of Geophysical Research Earth
Science 121: 881–889. https://doi.org/10.1002/2015JF003751.
Kayastha, P. 2015. “LandslideSusceptibility Mapping and Factor Effect Analysis Using
Frequency Ratio in a Catchment Scale: A Case Study from Garuwa Sub-Basin, East

121

Nepal.” Arabian Journal of Geosciences 8: 8601–8613. https://doi.org/10.1007/s12517-
015-1831-6.
Kavzoglu, T., E.K. Sahin, and I. Colkesen. 2013. “Landslide Susceptibility Mapping Using GIS-
Based Multi-Criteria Decision Analysis, Support Vector Machines, and Logistic
Regression.” Landslides 11, no. 3: 425–439. https://doi.org/10.1007/s10346-013-0391-7.
Kim, J.Y. 2019. "Introduction to Fuzzy Set Theory and The Hyperplane Separation Theorem of
Fuzzy Convex Sets." Masters thesis, University of Washington.
https://sites.math.washington.edu/~morrow/336_20/papers19/Jon.pdf.
Knepper, Jr., D.H. 2002. “Planning for the Conservation and Development of Infrastructure
Resources in Urban Areas—Colorado Front Range Urban Corridor.” In U.S. Geological
Survey Circular 1219. Denver: U.S. Geological Survey.
https://pubs.usgs.gov/circ/c1219/.
Korup, O. 2012. “Landslides in the Earth System.” in GIS Landslide, edited by H. Yamagishi
and N.P. Bhandary, 10–23. Tokyo: Springer Japan.
https://doi.org/10.1017/CBO9780511740367.003.
Kritikos, T., and T. Davies. 2014. “Assessment of Rainfall-Generated Shallow Landslide/Debris-
Flow Susceptibility and Runout Using a GIS-Based Approach: Application to Western
Southern Alps of New Zealand.” Landslides 12, no. 6: 1–25.
https://doi.org/10.1007/s10346-014-0533-6.
Lai, J.S., and F. Tsai. 2019. “Improving GIS-Based Landslide Susceptibility Assessments with
Multi-temporal Remote Sensing and Machine Learning.” Sensors 19: 25.
https://doi.org/10.3390/s19173717.
Lee, S. 2005. "Application of Logistic Regression Model and Its Validation for Landslide
Susceptibility Mapping Using GIS and Remote Sensing Data." International Journal of
Remote Sensing 26, no. 7: 1477–1491. https://doi.org/10.1080/01431160412331331012.
Lehmann, P., J. von Ruette, and D. Or. 2019. “Deforestation Effects on Rainfall-Induced
Shallow Landslides: Remote Sensing and Physically-Based Modelling.” Water Resources
Research 55: 9962–9976. https://doi.org/10.1029/2019WR025233.
Li, J., W. Wang, and Z. Han. 2021. “A Variable Weight Combination Model for Prediction on
Landslide Displacement Using AR Model, LSTM Model, and SVM Model: A Case
Study of the Xinming Landslide in China.” Environmental Earth Sciences 80: 356.
https://doi.org/10.1007/s12665-021-09696-2.
Li, Y., X. Wang, and H. Mao. 2020. “Influence of Human Activity on Landslide Susceptibility
Development in the Three Gorges Area.” Natural Hazards 104: 2115–2151.
https://doi.org/10.1007/s11069-020-0426406.

122

Lin, C.Y., C.Y. Lin, and C. Chompuchan. 2017. “Risk-Based Models for Potential Large-Scale
Landslide Monitoring and Management in Taiwan.” Geomatics, Natural Hazards and
Risk 8, no. 2:1505–1523. https://doi.org/10.1080/19475705.2017.1345797.
Lombardo, L., and P.M. Mai. 2018. “Presenting Logistic Regression-Based Landslide
Susceptibility Results.” Engineering Geology 244: 14–24.
https://doi.org/10.1016/j.enggeo.2018.07.019.
Ma, Y., J. Zhao, Y. Sui, S. Liao, and Z. Zhang. 2020. “Application of Knowledge-Driven
Methods for Mineral Prospectivity Mapping of Polymetallic Sulfide Deposits in the
Southwest Indian Ridge between 46° and 52°E.” Minerals 10: 970.
https://doi.org/10.3390/min10110970.
Madison, A., E. Fielding, Y. Sheng, and K. Cavanaugh. 2019. “High-Resolution Spaceborne,
Airborne and In Situ Landslide Kinematic Measurements of the Slumgullion Landslide in
Southwest Colorado.” Remote Sensing 11: 265. https://doi.org/10.3390/rs11030265.
Maes, J., M. Kervyn, A. de Hontheim, O. Dewitte, L. Jacobs, K. Mertens, M. Vanmaercke, L.
Vranken, J. Poesen. 2017. “Landslide Risk Reduction Measures: A Review of Practices
and Challenges for the Tropics.” Progress in Physical Geography 4, no. 2: 191–221.
http://doi.org/10.1177/0309133316689344.
Mallick, J., R.K. Singh, M.A. AlAwadh, S. Islam, R.A. Khan, and M.N. Qureshi. 2018. “GIS-
Based Landslide Susceptibility Evaluation Using Fuzzy-AHP Multi-Criteria Decision-
Making Techniques in the Abha Watershed, Saudi Arabia.” Environmental Earth
Sciences 77: 276. https://doi.org/10.1007/s12665-018-7451-1.
Malczewski, J. 2004. “GIS-Based Land-Use Suitability Analysis: A Critical Overview.”
Progress in Planning 62: 3–65. https://doi.org/10.1016/S0305-9006(03)00079-5.
Malczewski, J. and C. Rinner. 2015. Multicriteria Decision Analysis in Geographic Information
Science. New York: Springer. https://doi.org/10.1007/978-3-540-74757-4.
Martel, S.J. 2004. “Mechanics of Landslide Initiation as a Shear Fracture Phenomenon.” Marine
Geology 203: 319–339. https://doi.org/10.1016/S0025-3227(03)00313-X.
McKean, J. and J. Roering. 2004. “Objective Landslide Detection and Surface Morphology
Mapping Using High-Resolution Airborne Laser Altimetry.” Geomorphology 57: 331–
351. https://doi.org/10.1016/S0169-555X(03)00164-8.
Moreiras, S.M. and A. Coronato. 2010. “Landslide Processes in Argentina.” in Natural Hazards
and Human-Exacerbated Disasters in Latin America, edited by Edgardo M. Latrubesse.
301–332. Amsterdam: Elsevier. https://doi.org/10.1016/S0928-2025(08)10015-3.
National Oceanic and Atmospheric Administration. 2007. “Full Set of Rivers.” Last modified
May 16, 2007. https://www.weather.gov/gis/Rivers.

123

National Renewable Energy Laboratory (NREL). n.d. “Wind Prospector.” Accessed July 27,
2022. https://maps.nrel.gov/wind-
prospector/?aL=w_eWZI%255Bv%255D%3Dt%26P0Tu_1%255Bv%255D%3Dt%26P0
Tu_1%255Bd%255D%3D1&bL=ujuxuD&cE=0&lR=0&mC=38.28131307922966%2C-
106.23779296875&zL=6
Nelson, P.H., and S.L. Santus. 2011. “Gas, Water, and Oil Production from Wattenberg Field in
the Denver Basin, Colorado.” In U.S. Geological Survey Open-File Report 2011-1175.
Reston, Virginia: U.S. Geological Survey. https://pubs.usgs.gov/of/2011/1175/pdf/OF11-
1175.pdf.
Ott, R.F. 2020. “How Lithology Impacts Global Topography, Vegetation, and Animal
Biodiversity: A Global-Scale Analysis of Mountainous Regions.” Geophysical Research
Letters 47: e2020GL088649. https://doi.org/10.1029/2020GL088649.
Padilla, C., Y. Onda, T. Iida, S. Takahashi, and T. Uchida. 2014. “Characterization of the
Groundwater Response to Rainfall on a Hillslope with Fractured Bedrock by Creep
Deformation and Its Implication for the Generation of Deep-Seated Landslides on Mt.
Wanitsuka, Kyushu Island.” Geomorphology 204: 444–458.
https://doi.org/10.1016.j.geomorph2013.08.024.
Palcic, I. and B. Lalic. 2009. “Analytical Hierarchy Process as a Tool for Selecting and
Evaluating Projects.” International Journal of Simulation Modelling 1: 16–26.
https://doi.org/10.2507/IJSIMM08(1)2.112.
Park, J.H. 2015. “Analysis on the Characteristics of the Landslide – With a Special Reference on
Geo-Topographical Characteristics –.” Journal of Korean Society of Forest Science 104,
no. 4: 588–597. https://doi.org/10.14578/jkfs.2015.104.4.588.
Pawluszek, K. and A. Borkowski. 2016. “Impact of DEM-Derived Factors and Analytical
Hierarchy Process on Landslide Susceptibility Mapping in the Region of Rożnów Lake,
Poland.” Natural Hazards 86: 919–952. https://doi.org/10.1007/s11069-016-2725-y.
Petley, D. 2012. “Remote Sensing Techniques and Landslides.” in Landslides: Types,
Mechanisms and Modelling, edited by J.J. Clague and D. Stead, 159–171. Cambridge:
Cambridge University Press. https://doi.org/10.1017/CBO9780511740367.015.
Pourghasemi, H.R., B. Pradhan, and C. Gokceoglu. 2012. "Application of Fuzzy Logic and
Analytical Hierarchy Process (AHP) to Landslide Susceptibility Mapping at Haraz
Watershed, Iran." Natural hazards 63, no. 2: 965–996. https://doi.org/10.1007/s11069-
012-0217-2.
Preuth, T., T. Glade, and A. Demoulin. 2009. “Stability Analysis of a Human-Influenced
Landslide in Eastern Belgium.” Geomorphology 120: 38–47.
http://doi.org/j.geomorph.2009.09.013.
Qiu, C., and Y. Mitani. 2017. “Development of a GIS-Based 3D Slope Stability Analysis System
for Rainfall-Induced Landslide Hazard Assessment.” in GIS Landslide, edited by H.

124

Yamagishi and N.P. Bhandary, 71–93. Tokyo: Springer Japan.
https://doi.org/10.1007/978-4-431-54391-6.
Roslee, R., A.C. Mickey, N. Simon, and M.N. Norhisham. 2017. “Landslide Susceptibility
Analysis (LSA) Using Weighted Overlay Method (WOM) Along the Genting Sempah to
Bentong Highway, Pahang.” Malaysian Journal of Geosciences 1 no. 2: 13–19.
https://doi.org/10.26480/mjg.02.2017.13.19.
Regmi, N.R., J.R. Giardino, and J.D. Vitek. 2013. “Characteristics of Landslides in Western
Colorado, USA.” Landslides 11: 589–603. https://doi.org/10.1007/s10346-013-0412-6.
Reid, M.E., R.G. LaHusen, R.L. Baum, J.W. Kean, W.H. Schulz, and L.M. Highland. 2012.
Real-Time Monitoring of Landslides. Denver: U.S. Geological Survey.
https://doi.org/10.3133/FS20123008.
Rengers, F.K., L.A. McGuire, J.A. Coe, J.W. Kean, R.L. Baum, D.M. Staley, and J.W. Godt.
2016. “The Influence of Vegetation on Debris-Flow Initiation During Extreme Rainfall in
the Northern Colorado Front Range.” Geology 44, no. 10: 823–826.
https://doi.org/10.1130/G38096.1.
Roodposhti, M.S., J. Aryal, H. Shahabi, and T. Safarrad. 2016. “Fuzzy Shannon Entropy: A
Hybrid GIS-Based Landslide Susceptibility Mapping Method.” Entropy 18: 343.
https://doi.org/10.3390/e18100343.
Rotaru, A., D. Oajdea, and P. Răileanu. 2007. “Analysis of the Landslide Movements.”
International Journal of Geology 3, no. 1: 70–79.
https://naun.org/multimedia/NAUN/geology/ijgeo-10.pdf.
Saaty, R.W. 1987. “The Analytic Hierarchy Process—what it is and how it is used.” Mathl
Modelling 9, no. 3–5: 161–176.
Saha, A.K., R.P. Gupta, I. Sarkar, M.K. Arora, and E. Csaplovics. 2005. “An Approach for GIS-
Based Statistical Landslide Susceptibility Zonation—with a Case Study in the
Himalayas.” Landslides 2: 61–69. https://doi.org/10.1007/s10346-004-0039-8.
Schulz, W.H., J.P. McKenna, J.D. Kibler, and G. Biavati. 2009. “Relations between Hydrology
and Velocity of a Continuously Moving Landslide—Evidence of Pore-Pressure Feedback
Regulating Landslide Motion?” Landslides 6: 181–190. https://doi.org/10.1007/s10346-
009-0157-4.
Shit, P.K., G.S. Bhunia, and R. Maiti. 2016. “Potential Landslide Susceptibility Mapping Using
Weighted Overlay Model (WOM).” Modeling Earth Systems and Environment 2: 21.
https://doi.org/10.1007/s40808-016-0078-x.
Shou, K.J. and J. Chen. 2021. “On the Rainfall Induced Deep-Seated and Shallow Landslide
Hazard in Taiwan.” Engineering Geology 288: 106156.
https://doi.org/10.1016/j.enggeo.2021.106156

125

Shou, K.J. and J.F. Lin. 2016. “Multi-Scale Landslide Susceptibility Analysis Along a Mountain
Highway in Central Taiwan.” Engineering Geology 212: 120–135.
https://doi.org/10.1016/j.enggeo.2016.08.009
Stanley, T., and D.B. Kirschbaum. 2017. “A Heuristic Approach to Global Landslide
Susceptibility Mapping.” Natural Hazards 87, no. 1: 145–164.
https://doi.org/10.1007/s11069-017-2757-y.
State of Colorado. 2011. “Major River Basin Boundaries.” Last modified February 9, 2011.
https://cdss.colorado.gov/gis-data/gis-data-by-category.
—. 2015. “Counties in Colorado.” Last modified January 17, 2023.
https://data.colorado.gov/Transportation/Counties-in-Colorado/67vn-ijga.
—. 2015. “Major Roads in Colorado.” Last modified January 17, 2023.
https://data.colorado.gov/Transportation/Major-Roads-in-Colorado/e7ye-tasg.
Suchet, P.A., J.L. Probst, and W. Ludwig. 2003. “Worldwide Distribution of Continental Rock
Lithology: Implications for the Atmospheric/Soil CO2 Uptake by Continental Weathering
and Alkalinity River Transport to the Oceans.” Global Biogeochemical Cycles 17, no. 2:
1083. https://doi.org/10.1029/2002GB001891.
Tan, Q., M. Bai, P. Zhou, J. Hu, and X. Qin. 2020. “Geological Hazard Risk Assessment of Line
Landslide Based on Remotely Sensed Data and GIS.” Measurement 169: 10.
https://doi.org/10.1016/j.measurement.2020.108370.
Triantaphyllou, E. and K. Baig. 2005. “The Impact of Aggregating Benefit and Cost Criteria in
Four MCDA Methods.” IEEE Transactions on Engineering Management 52, no. 2: 213–
226. https://doi.org/10.1109/TEM.2005.845221.
Tweto, O. 1979. “MI-16 Geologic Map of Colorado.” Accessed May 10, 2022.
https://mrdata.usgs.gov/geology/state/state.php?state=CO.
United States Census Bureau, Department of Commerce. 2019. “TIGER/LineShapefile, 2019,
Primary and Secondary Roads State-based Shapefile.” Updated: January 15, 2021.
https://catalog.data.gov/dataset/tiger-line-shapefile-2019-state-colorado-primary-and-
secondary-roads-state-based-shapefile.
United States Geological Survey. 2018. “U.S. Landslide Inventory.” Last modified April 5,
2022.
https://uscssi.maps.arcgis.com/home/search.html?q=owner%3A%22esjones_USGS%22
&t=content&restrict=false.
—. 2022. “Elevation Products (3DEP): 1/3 arc-second DEM.” Last modified July 07, 2022.
https://apps.nationalmap.gov/downloader/.
—. n.d. “Lithclass polygon colors.” Accessed November 14, 2022.
https://mrdata.usgs.gov/catalog/lithclass-color.php.

126

—. n.d. “How Many Deaths Result from Landslides Each Year?” Accessed July 15, 2021.
https://www.usgs.gov/faqs/how-many-deaths-result-landslides-each-year?qt-
news_science_products=0#qt-news_science_products.
—. n.d. “What is a digital elevation model (DEM)?” Accessed February 8, 2023.
https://www.usgs.gov/faqs/what-digital-elevation-model-
dem#:~:text=DEMs%20are%20created%20from%20a,IfSAR%20(Alaska%20only)%20d
ata.
University of Colorado. 2015. “Colorado Watershed Boundary Dataset (WBD) Hydrologic Unit
8.” Last modified December 16, 2015.
https://ark.colorado.edu/ark:/47540/5c8ff914a84a6c000a68f3a8.
—. 2017. “BLM - Colorado State Boundary.” Last modified January 9, 2017.
https://geo.colorado.edu/catalog/47540-pf100p61s2w3.
Wang, G. and K. Sassa. 2009. “Seismic Loading Impacts on Excess Pore-Water Pressure
Maintain Landslide Triggered Flowslides.” Earth Surface Processes and Landforms 34:
232–241. https://doi.org/10.1002/esp.1708.
Wayllace, A., B. Thunder, N. Lu, A. Khan, and J.W. Godt. 2019. “Hydrological Behavior of an
Infiltration-Induced Landslide in Colorado, USA.” Geofluids 2019: 14.
https://doi.org/10.1155/2019/1659303.
Wieczorek, G.F. and P.P. Leahy. 2008. “Landslide Hazard Mitigation in North America.”
Environmental & Engineering Geoscience 14, no. 2: 133–144.
https://doi.org/10.2113/gseegeosci.14.2.133.
Wiki.GIS.com: The GIS Encyclopedia. Updated 27 November 2011, 02:23 UTC. Encyclopedia
on-line. Available from http://wiki.gis.com/wiki/index.php/Elevation. Internet. Retrieved
10 September 2022.
Winter, M.., D. Peeling, D. Palmer, and J. Peeling. 2019. “Economic Impacts of Landslides and
Floods on a Road Network.” AUC Geographica 54, no. 2: 201–207.
https://doi.org/10.14712/23361980.2019.18.
Yamagishi, H. 2017. “Identification and Mapping of Landslides.” in GIS Landslide, Edited by H.
Yamagishi and N.P. Bhandary, 3–11. Tokyo: Springer Japan.
https://doi.org/10.1007/978-4-431-54391-6_1.
Zakaria, Z., I. Sophian, Z.S. Sabila, and L.H. Jihadi. 2018. “Slope Safety Factor and Its
Relationship with Angle of Slope Gradient to Support Landslide Mitigation at Jatinangor
Education Area, Sumedang, West Java, Indonesia.” IOP Conference Series: Earth and
Environmental Science 145: 012052. https://doi.org/10.1088/1755-1315/145/1/012052.
Zhao, H., L. Yao, G. Mei, T. Liu, and Y. Ning. 2017. “A Fuzzy Comprehensive Evaluation
Method Based on AHP and Entropy for a Landslide Susceptibility Map.” Entropy 19:
396. https://doi.org/10.3390/e19080396.

127

Zhou, W., M.D. Minnick, J. Chen, J. Garrett, and E. Acikalin. 2021. “GIS-Based Landslide
Susceptibility Analyses: Case Studies at Different Scales.” Natural Hazards Review 22,
no. 3: 16. https://doi.org/10.1061/(ASCE)NH.1527-6996.0000485.
Zhu, A.X., R. Wang, J. Qiao, C.Z. Qin, Y. Chen, J. Liu, F. Du, Y. Lin, and T. Zhu. 2014. “An
Expert Knowledge-Based Approach to Landslide Susceptibility Mapping Using GIS and
Fuzzy Logic.” Geomorphology 214: 128–138.
https://doi.org/10.1016/j.geomorph2014.02.003.

Abstract (if available)

Abstract Landslide susceptibility mapping incorporates variables such as slope, precipitation, and lithology, among others, alongside a wide range of different methodologies in order to generate maps that may aid in landslide prediction. Criteria in the literature is expansive and varied, and the weighting methods used equally so. Weighted overlay and fuzzy overlay were chosen and compared using a select number of criteria as a means of testing which method would yield a better, more accurate result. Between the two, fuzzy overlay appears to be the more accurate of the two methods after evaluating the outputs, and this is due to the ways in which the two methods classify criteria. Of the eight criteria used, slope has been the most influential criterion for both methods with lithology coming in as a surprisingly strong factor for the weighted overlay and drainage systems as a strong influence for the fuzzy overlay. This influence is reflected in the locations of areas of higher landslide susceptibility and reveal that weighting and bias have definite effects on the outputs. There then exists a circular influence between the outputs shaping decisions that may affect large numbers of people and decisionmakers’ opinions affecting criteria emphasis. Of the two methods used, fuzzy overlay produced less biased results than weighted overlay, as the emphasis used in weighted overlay are highly subjective and influenced by the user.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Site selection of medium density housing in Tacoma, Washington: where to put “missing middle” housing

PDF

Providing a new low-cost primary care facility for under-served communities: a site suitability analysis for Service Planning Area 6 in Los Angeles County, California

PDF

Predicting the presence of historic and prehistoric campsites in Virginia’s Chesapeake Bay counties

PDF

A critical assessment of the green sea turtle central west Pacific distinct population segment utilizing maxent modeling on nesting site locations

PDF

Spatial narrative of the invasive lionfish in the Western Atlantic and Caribbean oceans: a GIS story map

PDF

Site suitability analysis for implementing tidal energy technology in southern California

PDF

A spatial and temporal exploration of how satellite communication devices impact mountain search and rescue missions in California’s Sierra Nevada mountain range

PDF

Rails to trails web mapping application for the great Redwood trails: mapping northern California’s repurposed trails

PDF

Trojan Food Finder: a web-based GIS campus food sharing application

PDF

Developing improved geologic maps and associated geologic spatial databases using GIS: Candy Mountain and Badger Mountain, WA

PDF

Selection of bridge location over the Merrimack River in southern New Hampshire: a comparison of site suitability assessments

PDF

Geographic local routing for social connection: a novel application for the integration of routing technology and multi-user environments

PDF

Assessing the use of normalized difference chlorophyll index to estimate chlorophyll-A concentrations using Landsat 5 TM and Landsat 8 OLI imagery in the Salton Sea, California

PDF

An examination of close-range photogrammetry and traditional cave survey methods for terrestrial and underwater caves for 3-dimensional mapping

PDF

Spatiotemporal analysis of the SLOSH and ADCIRC storm surge models: a case study of hurricane Ida

PDF

Climate conservation priorities: using MCDA to identify future refuge for the Joshua Tree Forest in California

PDF

A spatial analysis of violent crime cold spots: testing the capable guardian component of the routine activity theory

PDF

Using GIS to explore the tradeoffs in hydrographic survey planning: an investigation of sampling, interpolation, and local geomorphology

PDF

Urban green space accessibility and environmental justice: a GIS-based analysis in the city of Phoenix, Arizona

PDF

Scenario-based site suitability analysis and framework for biodiversity conservation: agricultural zone, Galapagos Archipelago, Ecuador

Asset Metadata

Creator Lee, Patricia L. (author)

Core Title A comparison of weighted and fuzzy overlays in mapping landslide susceptibility, south-central front range, Colorado

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Degree Conferral Date 2023-05

Publication Date 05/11/2023

Defense Date 03/14/2023

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

Tag fuzzy overlay,GIS,landslide susceptibility mapping,mutli-criteria decision analysis,OAI-PMH Harvest,weighted overlay

Format theses (aat)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Sedano, Elisabeth J. (committee chair), Loyola, Laura C. (committee member), Swift, Jennifer N. (committee member)

Creator Email patriciallee85@gmail.com,plee3804@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC113120787

Unique identifier UC113120787

Identifier etd-LeePatrici-11823.pdf (filename)

Legacy Identifier etd-LeePatrici-11823

Document Type Thesis

Format theses (aat)

Rights Lee, Patricia L.

Internet Media Type application/pdf

Type texts

Source 20230511-usctheses-batch-1042 (batch), 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 author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright.

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

Repository Email cisadmin@lib.usc.edu