University of Southern California Dissertations and Theses

An analysis of the North Rainier Elk Herd area, Washington: change detection and habitat modeling with remote sensing and GIS

(USC Thesis Other)

An analysis of the North Rainier Elk Herd area, Washington: change detection and habitat modeling with remote sensing and GIS

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

AN ANALYSIS OF THE NORTH RAINIER ELK HERD AREA, WASHINGTON:
CHANGE DETECTION AND HABITAT MODELING WITH REMOTE SENSING
AND GIS

By

Joshua J. Benton

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

May 2013

Copyright 2013 Joshua J. Benton
i

Dedication
Dedicated to Cassidy Benton.

ii

Acknowledgements
First and foremost I would like to thank my wife Tabby for her support during this process.
Without her essentially doing everything while I focused on school, this would not have been
possible. I would also like to thank my thesis committee, Dr. Travis Longcore, Dr. Flora Paganelli
and Dr. John Wilson, for their constant support through this whole process.
Special thanks must also be given to the authors/creators of the Westside Elk Nutrition and
Habitat Use Models who created a very exciting method of analyzing elk habitat and to Michelle
Tirhi who helped set me on the path to this project in more ways than one.

iii

Table of Contents
Dedication ................................................................................................................................... i
Acknowledgements..................................................................................................................... ii
List of Tables .............................................................................................................................. iv
List of Figures .............................................................................................................................. v
Abstract ...................................................................................................................................... 1
Chapter 1: Background................................................................................................................ 2
Chapter 2: Study Area ................................................................................................................. 6
Chapter 3: Methods .................................................................................................................... 8
Chapter 4: Results ..................................................................................................................... 33
Chapter 5: Discussion and Conclusions ...................................................................................... 46
References Cited ....................................................................................................................... 50
Appendix .................................................................................................................................. 52

iv

List of Tables
Table 1: Description of fields added to the changed areas polygon attribute table to be used in
the vegetation update toolbox. ................................................................................................. 17
Table 2: Equations used to predict biomass (kg/ha) of three forage classes based on stand and
forest overstory conditions in three forest zones and three study areas
a
in western Oregon and
Washington (Boyd et al. 2011). ................................................................................................. 20
Table 3: Final equations to predict dietary digestible energy (DDE) for elk based on abundance
(kg/ha) of two forage classes in different habitats and 3 study areas
a
in western Oregon and
Washington (from Table 2, Boyd et al. 2011). ............................................................................ 21
Table 4: Original DDE classification table proposed by Cook et al. (2004)................................... 33
Table 5: DDE divided into six classes (Boyd et al. 2011). ............................................................ 34

v

List of Figures
Figure 1: Study area map showing the Carbon River watershed and the major streams: Puyallup
River, Carbon River, Voight’s Creek, and South Prairie Creek. ...................................................... 7
Figure 2: Overview of the analysis toolboxes within the WENHUM, showing importance of
vegetation layer to subsequent analyses (Boyd et al., 2011). ..................................................... 10
Figure 3: The "Update Base Veg Toolbox" that was used to update the vegetation layer. .......... 11
Figure 4: NAIP imagery from 2006 that was used to create the original vegetation layer. .......... 12
Figure 5: NAIP imagery from 2011 that was used to update the vegetation layer in the
WENHUM. ................................................................................................................................ 13
Figure 6: Detail of a forested area circa 2006, prior to clear cutting. The polygon represents the
extent of the change that was found and is a portion of the change areas polygon layer........... 14
Figure 7: Detail of a changed area circa 2011, after clear cutting. The polygon represents the
extent of the change that was found and is a portion of the change areas polygon layer........... 15
Figure 8: The areas of change found within the study area. ....................................................... 16
Figure 9: Schematic showing the nutrition data model in the Nooksack region, as provided in the
“Elk Nutrition Toolbox”. A zoomed-in, segmented view is in the appendix (Figures A1 – A4) ..... 19
Figure 10: Map showing modeling regions for the WENHUM. Displayed here are the Nooksack,
Willapa Hills and Springfield regions. These regions are used for assigning the correct nutrition
model equations (Table 2) to the study area. ............................................................................ 22
Figure 11: Map showing the differences between the Nooksack and the Willapa Hills models.
The differences that are evident along the boundary line do not appear to follow any natural
cause. ....................................................................................................................................... 23
Figure 12: This is the model provided in the “Elk Covariates Toolbox”. A segmented and zoomed-
in view is in the appendix (Figures A5 – A11). ............................................................................ 26
Figure 13: Map showing the roads as well as the area designated as containing potential
closures. ................................................................................................................................... 29
Figure 14: Map showing the Potential Natural Vegetation zones used in the application of the
nutrition equation in the “Elk Covariates Toolbox”. ................................................................... 30
vi

Figure 15: Topography of the study area as characterized by a 30 m x 30 m resolution Digital
Elevation Model provided in the WENHUM............................................................................... 31
Figure 16: Schematic showing the predictive habitat use model in the “Elk Use Toolbox” of the
WENHUM. ................................................................................................................................ 32
Figure 17: Raw DDE values for 2006 with change areas outlined in black. Values ...................... 35
Figure 18: Raw DDE values for 2011 with change areas outlined in black. It is visible here that the
color within the outlines is representative of category 6. Most of the DDE values in the change
areas have values of 6. .............................................................................................................. 36
Figure 19: Change in percentage for each class of DDE for the years 2006 and 2011. ................ 37
Figure 20: Distance to the edge of cover and forage for 2006, using the original vegetation layer.
................................................................................................................................................. 38
Figure 21: Distance to the edge of cover and forage for 2011, using updated vegetation layer. . 39
Figure 22: Distance to open roads with all candidate roads listed as open. ................................ 40
Figure 23: Distance to open roads with all candidate roads listed as all closed. ......................... 41
Figure 24: Mean slope for study area, calculated for a 350 m radius circle around each pixel from
a slope grid derived from the DEM. ........................................................................................... 42
Figure 25: Comparison of the predicted use results for 2006 and 2011: (a) with all roads open in
2006; (b) with all roads closed 2006; (c) with all roads open 2011; and (d) all roads closed 2011.
................................................................................................................................................. 44
Figure 26: Changes in percentages of each elk use categories for 2006, both with potential road
closures modeled as “all open” (AO) and “all closed” (AC). ........................................................ 45
Figure 27: Changes in percentages for each of the elk use categories for 2011, both with
potential road closures modeled as “all open” (AO) and “all closed” (AC). ................................. 45
Figure 28: Map showing roads and 2011 predicted use with all candidate roads modeled as
closed to public use. In the northwestern portion of the map, a high density of roads and a high
predicted use occur together, these two conditions are in reality mutually exclusive. ............... 47

vii

Figure A 1: A close-up view of a portion of the nutrition model for the Nooksack Region (1 of 4).
................................................................................................................................................. 53
Figure A 2: A close-up view of a portion of the nutrition model for the Nooksack Region (2 of 4).
................................................................................................................................................. 54
Figure A 3: A close-up view of a portion of the nutrition model for the Nooksack Region (3 of 4).
................................................................................................................................................. 55
Figure A 4: A close-up view of a portion of the nutrition model for the Nooksack Region (4 of 4).
................................................................................................................................................. 56
Figure A 5: A close up view of the Elk covariates model (1 of 7). ................................................ 58
Figure A 6: A close up view of Elk covariates model (2 of 7). ...................................................... 59
Figure A 7: A close up view of the Elk covariates model (3 of 7). ................................................ 60
Figure A 8: A close up view of the Elk covariates model (4 of 7). ................................................ 61
Figure A 9: A close up view of the Elk covariates model (5 of 7). ................................................ 62
Figure A 10: A close up view of the Elk covariates model (6 of 7). .............................................. 62
Figure A 11: A close up view of the Elk covariates model (7 of 7). .............................................. 63

1

Abstract The North Rainier Elk Herd (NREH) is one of ten designated herds in Washington State,
all managed by the Washington Department of Fish and Wildlife (WDFW). To aid in the
management of the herd, the WDFW has decided to implement a spatial ecosystem analysis.
This thesis partially undertakes this analysis through the use of a suite of software tools, the
Westside Elk Nutrition and Habitat Use Models (WENHUM). This model analyzes four covariates
that have a strong correlation to elk habitat selection: dietary digestible energy (DDE); distance
to roads open to the public; mean slope; and distance to cover-forage edge and returns areas of
likely elk habitation or use. This thesis includes an update of the base vegetation layer from
2006 data to 2011, a series of clear cuts were identified as areas of change and fed into the
WENHUM models. The addition of these clear cuts created improvements in the higher quality
DDE levels and when the updated data is compared to the original, predictions of elk use are
higher. The presence of open or closed roads was simulated by creating an area of possible
closures, selecting candidate roads within that area and then modeling them as either “all open”
or “all closed”. The simulation of the road closures produced increases in the higher levels of
predicted use.

2

Chapter 1: Background
The North Rainier Elk Herd (NREH), as one of the ten designated herds in Washington State, is
managed by the Washington Department of Fish and Wildlife (WDFW). Although primarily
managed by the WDFW, other interested stakeholders include the Muckleshoot Indian Tribe
(MIT) and government and private landowners within the herd area boundary, including the
Washington Departments of Transportation (WSDOT) and Natural Resources (WDNR) and the
Hancock Timber Resource Group. The WDFW describes the NREH as an important state
resource that provides recreational, cultural and aesthetic values to the public. The NREH is also
valued highly by the Native American people of the area for ceremonial and subsistence uses
(Spencer 2002).
Elk (Cervus elaphus) are herbivores and consequently closely associated with the plants they eat.
This basic concept predates the early GIS modeling of elk habitat that in turn is the foundation
for the work in this thesis. In some of the early attempts to demonstrate the relationship
between elk and plants, individual animals showed a preference for areas that are higher in
quality when the habitat is heterogeneous (Stephens and Krebs 1986; Collins et al. 1978; Weins
1976; Martinka 1969). To further establish a relationship between elk and the food they eat
regardless of the environment, Kufeld (1973) looked at the value of specific plant species as
studied in previous work. He aggregated several studies surrounding diet and preferences of
Rocky Mountain elk (Cervus elaphus subspecies???). This work was conducted with the
knowledge that previous studies had mostly been centered on specific locations throughout the
Rocky Mountains of the western U.S. and Canada. Kufeld (1973) was able to categorize the plant
species and rank them by the intensity at which elk sought them. Some plants may be ranked
3

higher or lower because of limited availability or because of differences in palatability (Kufeld
1973).
Not only do elk prefer areas associated with higher quality forage, productivity of herds (i.e.
health of females during parturition and calf survival) is strongly related to time of the year and
quality/quantity of available forage. The times leading up to and during pregnancy greatly
influence birth rates and cow and calf survival. As early as 1958, Swank (1958) found a strong
correlation between the quality and quantity of available forage and the productivity of deer
herds, a species with a similar life history. Poor forage also has been found to adversely affect
Rocky Mountain cow elk in Wyoming through stress and weight loss limiting the survival of
calves (Thorne et al. 1976). Feeding trials on 30 captive female elk confirmed that a reduction in
quality of dietary digestible energy during the summer months affected reproductive functions,
with even just moderate reductions in DDE resulting in delayed estrus (Cook et al. 2001).
Furthermore, lactating female elk can be used as a quality indicator to link nutrition and habitat
selection because they have been found to have energy requirements 2–3 times the level found
during times when not lactating (Robbins, as cited in Beck et al. 2006).
Another basic component of the relationship between elk and their environment is the idea that
food is not the only factor in how elk choose their location. In 1979, a study found a link
between the sizing and spacing of forest stands and openings, habitat quality and road densities
with elk usage of the areas (Thomas et al., as cited in Wisdom et al. 1986), which in turn led to
the widespread use of cover-forage ratios and road-densities as indicators of elk habitat quality.
The weakness of this method was that it required several assumptions: that areas of forage and
cover were of adequate size; forage quality was not limiting; and that thermal cover needs were
4

met (Witmer and DeCalesta 1985). It was apparent then that more specific indices of the sizes
and spacing of forage and cover, habitat quality, and the effects of human disturbances were
needed (Witmer and DeCalesta 1985). Shortly thereafter, a model was created that evaluated
four criteria: sizing and spacing of forage and cover; density of roads open to motor vehicles;
cover quality; and forage quality (Wisdom et al. 1986) that was an important influence in the
creation of the WENHUM.
With this much knowledge of the relationships between elk and their environment it was only a
matter of time before geographic information systems (GIS) were applied to the study of elk. In
1997, existing digital land-cover data were used to evaluate elk habitat in Illinois to facilitate the
reintroduction of the eastern subspecies of elk (Cervus elaphus canadensis, Van Deelen et al.,
1997). Bian and West (1997) used GIS modeling to study elk, created a logistic regression model
to measure the observed calving grounds with other habitat variables such as dietary and other
needs. GIS was used to implement the model and predict other elk calving grounds throughout
the grasslands in the study.
This thesis will use the knowledge of the relationship between elk and their environment
through the use of the Westside Elk Nutrition and Habitat Use Models (WENHUM) to analyze
the habitat within the study area. The WENHUM is a suite of software tools that model elk
habitat and nutrition that were developed by a consortium of scientists from state, federal and
private entities. It uses a base vegetation layer to calculate four covariates, variables that may
have an influence on the outcome, that form the basis of the habitat use predictions. First will
be an update of the base vegetation layer that serves as an important factor for the subsequent
toolboxes contained within the WENHUM. Next the four covariates will be calculated.
5

During the process there will be a detailed account of the use of the WENHUM including errors
contained within some of the models. The result of this thesis will be the predicted use by elk
for the study area as well as other outputs created along the way including DDE data, an
assessment of the available nutrition.

6

Chapter 2: Study Area
The study area is the Carbon River watershed (as established by the National Hydrography
Dataset; NHD, WA Department of Ecology, 2012) with the addition of a 2.5 km buffer (Figure 1).
A buffer is added to account for the effects of roads and vegetation that are adjacent but on the
outside of the study area. The distance of 2.5 km was chosen for this analysis as an estimation of
the effects of the surrounding environment. The site is in the western portion of Washington
State and located almost entirely within Pierce County, covers 924 km
2
and overlaps the
northwest corner of Mount Rainier National Park (199 km
2
). The major waterway is the Carbon
River, which is in turn fed by Voight’s Creek and South Prairie Creek. The landscape is dominated
by its highest peak, Mt. Rainier, a solitary peak at a height of 4,394 m. The lowest point in the
study area is 27 meters and runs along the Carbon River as it in turn feeds into the Puyallup
River which runs into Puget Sound. The two largest population centers, Orting and Buckley have
a combined population of 11,333 (U.S. Census Bureau 2012; City of Buckley 2012).
The climate of the area is cool with relatively dry summers and mild, wet and cloudy winters.
Sunny days per month average 4–8 in winter, 8–15 in spring and fall and 15–20 in the summer
(Western Regional Climate Center 2013). The dominant trees found within the study area
include western hemlock (Tsuga heterophylla), Pacific Fir (Abies amabilis) and Mountain
Hemlock (Tsuga mertensiana).

7

Figure 1: Study area map showing the Carbon River watershed and the major streams: Puyallup River,
Carbon River, Voight’s Creek, and South Prairie Creek.
8

Chapter 3: Methods
The WENHUM is composed of four toolboxes: “Update Base Veg Toolbox”; “Elk Covariate
Toolbox”; “Elk Nutrition Toolbox”; and “Elk Use Toolbox” (Figure 2). The toolboxes and models
analyze elk nutrition and habitat use specifically in western Washington and Oregon. Each of the
toolboxes contains one or more models which are used to identify the four covariates that have
consistently provided the most support for observed patterns of elk movements which in turn
are used in the identification of the predicted habitat use(Boyd et al. 2011):
 elk dietary digestible energy (DDE; higher DDE, higher predicted elk use);
 distance to roads (farther from roads, higher use);
 percent slope (flatter slopes, higher use); and
 distance to cover/forage (closer to edge, higher use).
A modified or updated vegetation layer is a product of the first of the four toolboxes, “Base Veg
Update Toolbox”. The base vegetation layer is an important part of the remaining tools, it is the
basis for some the assumptions about the habitat. Second, the “Elk Nutrition Toolbox”
calculates DDE for the study area using the base vegetation layer. Third, the “Elk Covariate
Toolbox”, establishes the remaining three covariates needed to run the fourth tool, the “Elk Use
Toolbox”. The “Elk Use Toolbox” creates a map that shows the predicted probability of elk using
an area.
The use of the WENHUM is the main focus of this thesis and a major component of being able to
use the WENHUM is the use of an up-to-date vegetation layer. Vegetation data were originally
compiled from National Agriculture Imagery Program (NAIP) images taken circa 2006. NAIP
9

imagery is a product of the Farm Service Agency (FSA) in an effort to collect the imagery for the
entire conterminous United States at a 1-meter resolution. The imagery used for this thesis is
captured at a spatial resolution of 1 meter (NAIP products prior to 2008 vary) from an aircraft
platform in three bands: Red; Green; and Blue (natural color; USDA 2009). It is a raster with 1 m
by 1 m pixels.
The first step in the analysis was to determine the study area. I originally planned to analyze the
entire NREH area comprising 7,144 km
2
(9,131 km
2
with a 4 km buffer). I soon realized that the
size of the area was excessive; the magnitude of the image files associated with that area were
unmanageable and processing times on available equipment were prohibitive. A decision was
then made to work with a portion of the NREH area. The wider area can easily be analyzed in
separate sections using the data and clarification to the WENHUM methods developed in this
thesis.

10

Figure 2: Overview of the analysis toolboxes within the WENHUM, showing importance of
vegetation layer to subsequent analyses (Boyd et al., 2011).

The area chosen was the Carbon River Watershed, as established by the National Hydrography
Dataset (NHD; WA Department of Ecology, 2012). This area was chosen because it is within the
original study area and it is known to contain elk (personal observation). Any areas that could be
designated as uninhabitable by elk prior to any analysis (e.g. city, residential, and other
urbanized areas) were left in the study area to judge the ability of the WENHUM to recognize
them as unsuitable.
11

The first step in the use of the WENHUM was to update the available vegetation layer using the
“Update Base Veg Toolbox” (Figure 3). The layer provided with the WENHUM, referred to as
“gnn_2006” (gnn = gradient nearest neighbor), is defined by Boyd et al. (2011) as “… a modeling
method that incorporates multivariate statistics and imputation to produce a variety of
vegetation maps, based on ground data and mapped (explanatory) data. For elk nutrition and
habitat use modeling, we used key fields from the March 2010 release of the GNN species-size
model, developed for Northwest Forest Plan Effectiveness Monitoring.”
The data consist of many fields including: HW100 – the proportion of stems in the dominant
canopy layer that are considered hard wood tree species multiplied by 100; STNDHGT – stand
height in meters; CANCOV – canopy cover, percent of all live trees; and Elk_Hab_Ma – a field
that defines whether vegetation provides elk habitat or not (using modeled values). The layer is
based on data established in 2006 and was updated to 2011.
Initially I evaluated the use of a supervised classification to derive the vegetation layer update
but due to technical difficulties arising from the constraints and limitations of available
hardware, software and geospatial data, a “heads-up” approach was used. A heads-up approach
is a common approach defined by the use of an image, typically a satellite image or
orthophotograph displayed as a basemap, and then features such as buildings or parcels are
scanned visually and drawn on top (Esri 2013).
The heads-up digitizing process was undertaken in ArcMap with the same NAIP imagery used in
the creation of the vegetation layer, the most current available from 2011. Polygons were hand
drawn around areas where significant change had occurred. The first step was to create a new
feature class polygon layer, referred to as the vegetation update polygon layer, which was used
12

as a parameter in the “Update Base Veg Toolbox”. An edit session was started using the “Create
Features” toolset to digitize polygons. The study area was subdivided into NHD sub-watersheds
and each unit was carefully scanned by eye, using the swipe function to switch between the
2006 (Figure 4) and 2011 (Figure 5) images quickly, searching for significant change. When an
area of change was found (Figures 6 and 7) a polygon was drawn around it and saved as the
change areas polygon layer (Figure 8).
11

Figure 3: The "Update Base Veg Toolbox" that was used to update the vegetation layer.

12

Figure 4: NAIP imagery from 2006 that was used to create the original vegetation layer.
13

Figure 5: NAIP imagery from 2011 that was used to update the vegetation layer in the
WENHUM.
14

Figure 6: Detail of a forested area circa 2006, prior to clear cutting. The polygon represents the
extent of the change that was found and is a portion of the change areas polygon layer.
15

Figure 7: Detail of a changed area circa 2011, after clear cutting. The polygon represents the
extent of the change that was found and is a portion of the change areas polygon layer.
16

Figure 8: The areas of change found within the study area.
17

When all 72 areas of change were found within the study area, the attribute table for each was
edited using the “Update Base Veg Toolbox” by updating the following fields in Table 1.

Table 1: Description of fields added to the changed areas polygon attribute table to be used in
the vegetation update toolbox.
Field Name Description
Value A long integer data type, this is what the toolbox will
look for when creating a grid from the polygon layer.
HW Proportion of hard woods, 0 - 10
CANCOV Percent canopy cover, 0 - 100
STANDHGT Height of stand (meters), ≥1

The value field within the attribute table for the changed areas polygon layer is an arbitrary
number indicating each unique combination of values for the fields related to hardwoods
proportions, canopy cover, and stand height. This is the field that was used in the join process in
the update.
Next, the attribute table from the existing vegetation grid was modified to include the values
from the changed areas. The entries here matched the “Value” field in the update polygon layer
and because all areas of change were considered identical, there was only a single entry. Then
all of the appropriate data is entered as parameters in the model provided in the “Update Base
Veg Toolbox”.
18

Before the “Update Base Veg Toolbox” could be run it had to be edited in an edit session within
ArcMap 10.1 Model Builder (Esri 2012a) because it was not producing the desired results. The
user guidelines stated that the newly updated vegetation layer would be output as a layer titled
“new_veg” when it actually had retained a title designated from the step “Copy Raster” titled
“temp_view_CopyRaster”. Once it was determined that this was the layer that was desired, the
title applied at this step was edited to “new_veg”.
The next model is the Elk Nutrition Model (Figure 9); this model creates one of the four
covariates, mean DDE as well as “raw” DDE estimates. The raw DDE output of the nutrition
model can be used alone to assess the nutritional values available to elk. The model uses a
series of equations (Tables 2 and 3) to calculate the biomass of available forage in two
vegetation series in three regions in western Washington and western Oregon: Nooksack;
Willapa Hills; and Springfield (Figure 10). The two vegetation series are TSME-ABAM (mountain
hemlock [Tsuga mertensiana] / Pacific silver fir [Abies amabilis]) and THSE (western hemlock
[Tsuga heterophylla]).
19

Figure 9: Schematic showing the nutrition data model in the Nooksack region, as provided in the “Elk Nutrition Toolbox”. A zoomed-in,
segmented view is in the appendix (Figures A1 – A4)

20

Table 2: Equations used to predict biomass (kg/ha) of three forage classes based on stand and
forest overstory conditions in three forest zones and three study areas
a
in western Oregon and
Washington (Boyd et al. 2011).
TSME & ABAM
b
habitats, all seasons, all study areas
AB
c
= 657.6 – 11.28(CC) + 0.0458(CC
2
) + 553.06(HW)
NB = 527.8 – 6.09(CC) + 590.49(HW)
SB = 1/(0.00833 + 0.00062(CC))

TSHE habitats, all seasons, by study area
AB
Nk
= 707.3 – 13.93(CC) + 0.0731(CC
2
) + 383.17(HW)
AB
WH
= 707.3 – 6.28(CC) - 0.0154(CC
2
) + 383.17(HW)
AB
Sp
= 490.5 – 11.70(CC) + 0.0731(CC
2
) + 383.17(HW)
NB
Nk
= 671.8 – 16.91(CC) + 0.1092(CC
2
) + 268.13(HW)
NB
WH
= 477.4 – 3.90(CC) – 0.0151(CC
2
) + 268.13(HW)
NB
Sp
= 308.5 – 7.59(CC) + 0.0473(CC
2
) + 268.13(HW)
SB
Nk
= 80.1 – 0.66(CC) + 99.83(HW)
SB
WH
= 212.6 – 2.20(CC) + 99.83(HW)
SB
Sp
= 166.2 – 1.68(CC) + 99.83(HW)
a
The 3 study areas are Nooksack (Nk, northern Cascades near Mount Baker), Willapa Hills (WH,
coastal foothills west of Centralia, WA), and Springfield (Sp, central Cascades, west of
Springfield, OR).
b
Habitat codes are: TSME = Tsuga mertensiana forest series, ABAM = Abies amabilis forest
series, TSHE = Tsuga heterophylla forest series.
c
Forage class codes (equation variable names) are: NB = biomass (kg/ha) of neutral plant species
(those plants that elk neither significantly avoided nor selected), SB = biomass (kg/ha) of
selected plant species (those plant species that elk significantly selected), and AB = biomass
(kg/ha) of accepted species (SB and NB combined). Predictor variable codes are: CC = canopy
cover (%) of all live trees; HW = proportion of stems in dominant canopy layer that are
hardwood tree species (e.g., red and other alders, big leaf maple, and paper birch).

The study area is divided nearly down the middle by the boundary line between the Nooksack
and Willapa Hills regions (Figure 11). I first attempted to split the study area into two halves for
the elk nutrition model runs to obtain individual results for both the Nooksack and Willapa Hills
21

sets of equations. From those results it became apparent that there were inconsistencies that
were not biologically realistic. For example, this approach resulted in abrupt changes in
calculated DDE along the boundary. Areas that were adjacent had different DDE values when in
reality, or when viewed on the ground, they should not have. On the ground there would be
differences of course but they would not follow this artificial boundary line (Figure 12). The
differences were attributed to the equations that were used to predict biomass within the two
regions.
Table 3: Final equations to predict dietary digestible energy (DDE) for elk based on abundance
(kg/ha) of two forage classes in different habitats and 3 study areas
a
in western Oregon and
Washington (from Table 2, Boyd et al. 2011).
TSME & ABAM
b
habitats, all seasons, all study areas
DDE = 2.44 + 0.000889(NB)
c
+ 0.00308(SB) - 0.00000546(SBNB)

TSHE habitats, all seasons, by study area

DDE
Nk
= 2.362 + 0.00108(NB) + 0.000504(SB) – 0.00000361(SBNB)
DDE
WH
= 2.278 + 0.00062(NB) + 0.00120(SB) – 0.00000172(SBNB)
DDE
Sp
= 2.300 + 0.00108(NB) + 0.00129(SB) – 0.00000418(SBNB)

a
The 3 study areas are Nooksack (Nk, northern Cascades near Mount Baker), Willapa Hills (WH,
coastal foothills west of Centralia, WA), and Springfield (Sp, central Cascades west of Springfield,
OR).
b
Habitat codes are: TSME = Tsuga mertensiana forest series, ABAM = Abies amabilis forest
series, TSHE = Tsuga heterophylla forest series.
c
Forage class codes (equation variable names) are: NB = biomass (kg/ha) of neutral plant species
(those plants that elk neither significantly avoided nor selected), SB = biomass (kg/ha) of
selected plant species (those plant species that elk significantly selected), and SBNB = the
interaction of SB x NB (i.e., the product of SB and NB).
22

Figure 10: Map showing modeling regions for the WENHUM. Displayed here are the Nooksack,
Willapa Hills and Springfield regions. These regions are used for assigning the correct nutrition
model equations (Table 2) to the study area.

23

Figure 11: Map showing the differences between the Nooksack and the Willapa Hills models.
The differences that are evident along the boundary line do not appear to follow any natural
cause.

24

I then decided that because the portion of the study area that falls under the Nooksack region
was larger, that I would use the Nooksack Region model. At a minimum, this approach ensures
that the habitat suitability estimates will be consistent within the study area, and could be
repeated with the other region’s model if needed in the future.
While running the nutrition model for the Nooksack region, I found further errors in the
WENHUM model. Some of the outputs were not being saved to the designated output folder. I
then opened model in an edit session in ArcMap (Esri 2012a) and scanned the outputs from
several of the tools to verify labels and permanency. Permanency refers to the option for an
output that allows it to be designated as “intermediate,” which means that it is created and
used for the next step but is deleted at the end of the model run. Several of the outputs were
erroneously marked as intermediate, particularly the raw DDE output along with others that
were not a part of this analysis.
The three remaining covariates: distance to roads open to the public; mean slope; and distance
to cover/forage edges were generated with the “Elk Covariates Toolbox” (Figure 12). This model
required five different inputs:
1. Vegetation Grid
2. Study Area Boundary
3. Roads Data (Figure 13)
4. PNV (potential natural vegetation) Zones (Figure 14)
5. DEM Grid (Figure 15)
26

Figure 12: This is the model provided in the “Elk Covariates Toolbox”. A segmented and zoomed-in view is in the appendix (Figures A5 –
A11).
27

The road data were downloaded from the Washington Office of Financial Management’s
website in the form of TIGER/line data from the U.S. Census Bureau (WOFM 2012). To allow its
use in the model, a field titled “Open” was added. This field would be the designation of
whether or not the roads were open to public use. Because data were not available on the
status of roads as open or closed, I conducted a sensitivity analysis to show the range of
variation in model results from having potentially closed roads modeled as both all-open or all-
closed.
To determine which roads would be modeled as potentially opened and closed, an area of likely
closures was determined. To accomplish this, land ownership and land use data were evaluated
to determine which of these areas were most likely to have closures. I decided that land owned
by the U.S. Forest Service (USFS), the Washington Department of Natural Resources (WDNR)
and other areas designated as “timberland” by the WDNR were the most likely to be closed
(Figure 13). Once a polygon indicating the area was created, the roads within that area were
narrowed down further to roads with no names and USFS roads labeled as “National Forest
Development Roads” as the closure candidates. The models were then run to show all roads
either opened or closed to illustrate the effect.
To calculate the distance to roads open to public use, the model separates all of the roads into
either open or closed. It then calculates the distance of each pixel to the nearest open road in
meters.
The method of calculating the distance to cover forage relies on the designation of pixels into
cover, forage or nodata cells. The isolated nodata cells (< 2x2) are then eliminated by assigning
them to either cover or forage based on the surrounding cells. Cover areas with canopy cover
28

≥40% and stand height > 2 m, are defined as areas occupying at least a 3x3 cell area and then
the smaller areas are redefined as forage. Forage is defined as areas that are 3x3 or greater with
smaller areas defined as “not classified.” To calculate the distances to cover/forage edges, the
model determines the boundary lines between the cover and forage areas and calculates the
distance in meters of pixels to the boundary lines.
The vegetation grids are the original and the updated vegetation layers. The PNV is a layer
provided with the WENHUM (Figure 14) and is used in the calculation of distance to cover
forage edge. The mean slope is calculated for a 350-meter radius circle around each pixel from a
slope grid derived from the DEM (Figure 15).
29

Figure 13: Map showing the roads as well as the area designated as containing potential
closures.
30

Figure 14: Map showing the Potential Natural Vegetation zones used in the application of the
nutrition equation in the “Elk Covariates Toolbox”.
31

Figure 15: Topography of the study area as characterized by a 30 m x 30 m resolution Digital
Elevation Model provided in the WENHUM.
32

The final toolbox used was the “Elk Use Toolbox” (Figure 16) that took the four covariates and
used them within a single equation to create a single grid that predicts the likelihood that elk
will use the habitat. The equation is:
(

) (

) (

) (

)
where: Y = predicted elk use; X
1
= mean DDE; X
2
= distance to open public roads (meters); X
3
=
distance to cover/forage edge (meters); X
4
= mean slope.

Figure 16: Schematic showing the predictive habitat use model in the “Elk Use Toolbox” of the
WENHUM.

33

Chapter 4: Results
The “Update Base Veg Toolbox” produces a new vegetation grid that can either represent real
changes over time as an update of the base vegetation layer or simulated changes to model
proposed management actions. This analysis updated the original to reflect the changes to the
forest structure during the period 2006 - 2011. The actual vegetation grid is not used on its own
as an analysis tool but as an input into the other models.
The primary output from the Elk Nutrition model is the information pertaining to dietary
digestible energy (DDE), which is output in multiple forms. The two that we are concerned with
are DDE (measured in kcal/g) and mean-DDE. The values in the DDE output were originally
summarized by Cook et al. (2004) into four categories (Table 4).

Table 4: Original DDE classification table proposed by Cook et al. (2004).

Description DDE (kcal/g of food)
Excellent > 2.90
Good 2.75 – 2.90
Marginal 2.40 – 2.75
Poor < 2.40

One problem with the summary provided by Cook et al. (2004) is that nearly all of the DDE levels
within the WENHUM modeling areas are below the excellent and good categories. Boyd et al.
(2011) later created another classification scheme that uses six classes defined by the
percentage of pixels that fall within each class (Table 5).
34

Table 5: DDE divided into six classes (Boyd et al. 2011).
Class Description DDE
1 Poor < 2.40
2 Low-Marginal ≥ 2.40 to < 2.575
3 High-Marginal ≥ 2.575 to < 2.75
4 Low-Good ≥ 2.75 to < 2.825
5 High-Good ≥ 2.825 to < 2.90
6 Excellent ≥ 2.90

When viewing the results from the “Elk Nutrition Toolbox” (Figures 17 and 18) it is easy to see
that there is essentially no change from 2006 to 2011 except in the designated change areas. In
the change areas, where the change was clear cutting, the elk habitat has improved the DDE to
the sixth class. There is a 2.25% increase in the total area of DDE class 6 (Figure 19) representing
a significant increase in the amount of the most desirable DDE category.
Also evident is that the two lowest DDE classes, “poor” and “low-marginal”, have decreased and
the remaining middle classes, “high-marginal”, “low-good”, and “high-good”, have remained
nearly unchanged (Figure 19).
35

Figure 17: Raw DDE values for 2006 with change areas outlined in black. Values
36

Figure 18: Raw DDE values for 2011 with change areas outlined in black. It is visible here that the
color within the outlines is representative of category 6. Most of the DDE values in the change
areas have values of 6.
37

Figure 19: Change in percentage for each class of DDE for the years 2006 and 2011.

The “Elk Covariate Toolbox” produced the following outputs:
 distance to cover/forage edge
o 2006 (Figure 20)
o 2011 (Figure 21);
 distance to the nearest road open to public use
o all candidate roads modeled as “open” (Figure 22)
o all candidate roads modeled as “closed” (Figure 23);
 and mean slope (Figure 24).

0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
1 2 3 4 5 6
Percenatge of Total Study Area
DDE Category
2006 2011
38

Figure 20: Distance to the edge of cover and forage for 2006, using the original vegetation layer.
39

Figure 21: Distance to the edge of cover and forage for 2011, using updated vegetation layer.
40

Figure 22: Distance to open roads with all candidate roads listed as open.
41

Figure 23: Distance to open roads with all candidate roads listed as all closed.
42

Figure 24: Mean slope for study area, calculated for a 350 m radius circle around each pixel from
a slope grid derived from the DEM.
43

The output from the “Elk Use Toolbox” for example, Boyd et al. (2011) caution that the results
for any given pixel are not standardized and that the number values can vary substantially from
analysis to analysis, they are an index where the higher the number the higher the predicted
use. For this reason, all results have been classified identically to support comparisons. The
classification used is Geometrical, described by Esri as “a scheme that creates class breaks based
on class intervals that have a geometrical series. The algorithm creates geometric intervals by
minimizing the sum of squares of the number of elements in each class. This ensures that each
class range has approximately the same number of values with each class and that the change
between intervals is fairly consistent” (Esri 2012b). By using this method we can ensure that the
study area is divided into equal regions for each use level, as recommended by Boyd et al.
(2011).
When viewing the comparative maps (Figure 25) it is easy to see the effect of the open and
closed roads. The areas of higher use increase in the region where roads have been modeled to
be closed (Figure 25-A and 25-D). In 2006 and 2011 all of the predicted use categories increased
with the exception of the lowest category and category 6 in 2006 which remained static at
0.006% (Figures 26 and 27). This data illustrates the importance of the road data, having
accurate data of road closures/openings can be have an effect on the use probability; the
amount of the effect is dependent on the density and number of closures.

44

Figure 25: Comparison of the predicted use results for 2006 and 2011: (a) with all roads open in
2006; (b) with all roads closed 2006; (c) with all roads open 2011; and (d) all roads closed 2011.
45

Figure 26: Changes in percentages of each elk use categories for 2006, both with potential road
closures modeled as “all open” (AO) and “all closed” (AC).

Figure 27: Changes in percentages for each of the elk use categories for 2011, both with
potential road closures modeled as “all open” (AO) and “all closed” (AC).
0%
10%
20%
30%
40%
50%
60%
70%
80%
1 2 3 4 5 6
Percentage of Total Study Area
Predicted Elk Use Category
2006 (AO) % 2006 (AC) %
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
50%
1 2 3 4 5 6
Percentage of Total Study Area
Predicted Elk Use Category
2011 (AO) % 2011 (AC) %
46

Chapter 5: Discussion and Conclusions
Analyses show that at least in the short term that clear cuts can provide improved habitat for
elk, as shown by the increase in total habitat between 2006 and 2011.
The DDE showed an increase in the more desirable categories, particularly in category six
(“excellent”) between 2006 and 2011. This resulted from clear cuts providing an increase in
available forage. Benefits from the clear cuts will likely decrease as the size of individual clear
cuts increases because this will increase the distance to cover/forage edges. In addition to this,
as time passes since the time of the clear cut, DDE and associated predicted use will decrease as
the stand height increases and the other variables change as well.
At the onset of the study it was decided to leave in urbanized areas that were clearly not elk
habitat and to judge the ability of WENHUM to classify them. It is evident in that while the
models use proximity to open roads as an important covariate, exceptional habitat must exist in
those areas since they labeled as “high” in terms of predicted habitat use (Figure 28). Given that
the WENHUM predicts habitat based on proximity to open roads, and all roads in that area are
classed as open, the amount of high predicted use is surprising. The model does not take into
account human population densities or urbanization and this analysis shows that (Figure 28). In
the map it shows how in the northwest portion of the study area there is a high concentration of
roads open to the public but also a high predicted elk use, which does not actually occur.
Therefore, areas that can be deemed to be uninhabitable by elk prior to any analysis should be
removed at the onset to minimize these kinds of problems.
47

Figure 28: Map showing roads and 2011 predicted use with all candidate roads modeled as
closed to public use. In the northwestern portion of the map, a high density of roads and a high
predicted use occur together, these two conditions are in reality mutually exclusive.
48

I encountered challenges in the use of the WENHUM (new table). In some instances, models
were not producing the results stated in the manual, results that were required as input for the
next step. The models had to be edited so that all the appropriate results were not designated
as “intermediate” or were named in a way so that the desired result was known (Table 6).
Table 6: Description of the editing required for toolboxes in the WENHUM.
Model Problem Solution
“Update Base Veg Toolbox” Newly created base
vegetation layer Incorrectly
named.
Edited title for the output
raster dataset in the “Copy
Raster” tool within the model.
“Elk Nutrition Toolbox” Outputs not being saved to
designated folders.
Removed “intermediate”
designation from several
required outputs .

Based on the results of this analysis I think that it is safe to say that the WENHUM can be used to
evaluate the much larger NREH area. Although it would be wise to break it down into sections to
be combined at different points to produce whole maps. Caution should be taken with the study
area and its overlap of the modeling regions.
In response to the initial attempt of conducting a classification and change analysis, it is my
opinion that it is possible but would require an immense amount of computing/processing time
that may make it prohibitive in most cases. There was also an anticipated issue with creating the
change areas polygons following the change detection. The results were likely to be too
granulated and not formable into clean, easy to use polygons. It should also be noted that
judging by the results obtained here using the “heads-up digitization” of the change area
polygons, it may not be necessary.
49

The WENHUM created by Boyd et al. (2011) are a fascinating way to view and to model elk
habitat. There is an even greater amount of detail that can be used to study an area especially
when taking into account forestry practices other than clear cutting. In situations where there
has been more thinning or selective cutting a thorough and more detailed analysis of an area
would be required. Caution should be taken as the size of the study area increases, finite detail
on forest stands and road closures will be increasingly more difficult.

50

References Cited
Beck J L, Peek J M, and Strand E K 2006 Estimates of elk summer range nutritional carrying
capacity constrained by probabilities of habitat selection. Journal of Wildlife
Management 70: 283–94
Bian L, and West E 1997 GIS Modeling of elk calving habitat in a prairie environment with
statistics. Photogrammetric Engineering and Remote Sensing 63: 161–67
Boyd J, Coe P, Cook J, Cook R, Johnson B, Naylor B, Nielson R, Rowland M, Wisdom M 2011 User
guidelines for application, summary, and interpretation of west side elk nutrition and
habitat use models - draft version 1.0. , March, pp.1–65
City of Buckley 2012. About Buckley. Www document:
http://www.cityofbuckley.com/documents/about_buckley.html
Collins B, Urness J, and Austin, D, 1978 Elk diets and activities on different lodgepole pine
habitat segments. Journal of Wildlife Management 42: 799–810
Cook J, Johnson B, Cook R, Riggs R, Delcurto T, Bryant L, Irwin L 2004 Effects of summer-autumn
nutrition and parturition date on reproduction and survival of elk. Wildlife Monographs
155: 1–61
Cook R, Murray D, Cook J, Zager P, Monfort S, 2001 Nutritional influences on breeding dynamics
in elk. Canadian Journal of Zoology 79: 845–853
Deelen Van T, Mckinney L, Joselyn M, John E, Buhnerkempe J 1997 Can we restore elk to
Southern Illinois: the use of existing digital land-cover data to evaluate potential habitat.
Wildlife Society Bulletin 25: 886–894
Esri, 2012a. ArcGIS 10.1
Esri, 2012b. ArcGIS Help 10.1 - Classifying numerical fields for graduated symbology. Www
document:
http://resources.arcgis.com/en/help/main/10.1/index.html#//00s50000001r000000
Esri, 2013. GIS Dictionary – Heads-up digitizing. Www document:
http://support.esri.com/en/knowledgebase/GISDictionary/term/heads-up%20digitizing
Kufeld C 1973 Foods eaten by the rocky mountain elk. Journal of Range Management 26: 106–
113
Martinka J 1969 Population ecology of summer resident elk in Jackson Hole, Wyoming. Journal
of Wildlife Management 33: 465–481
51

Spencer R 2002 Washington State elk herd plan: north rainier elk herd, Olympia, WA,
Department of Fish and Wildlife
Stephens W and Krebs R 1986 Foraging theory. Princeton: Princeton University Press
Swank G 1958 The mule deer in Arizona chaparral: and an analysis of other important deer
herds. Phoenix, AR, Game and Fish Department
Thorne T, Dean E and Hepworth G 1976 Nutrition during gestation in relation to successful
reproduction in elk. Journal of Wildlife Management 40: 330–335
U.S. Census Bureau 2012 State and county quickfacts - Orting, WA. Www document:
http://quickfacts.census.gov/qfd/states/53/5352005.html
USDA 2009 National agriculture imagery program (naip) information sheet. Www document:
http://www.fsa.usda.gov/FSA/apfoapp?area=home&subject=prog&topic=nai
WA Department of Ecology 2012 GIS Data. Www document:
http://www.ecy.wa.gov/services/gis/data/data.htm.
Weins J.A. 1976 Population responses to patchy environments. Annual Review of Ecology and
Systematics 7: 81–120.
Western Regional Climate Center 2013 Climate of Washington. Www document:
http://www.wrcc.dri.edu/narratives/WASHINGTON.htm
Wisdom M, Bright L, Carey C, Hines W, Pederson R, Smithey D, Thomas J, Witmer G 1986 A
model to evaluate elk habitat in western Oregon, Portland, OR, USDA Forest Service.
Witmer W and DeCalesta S 1985 Effect of forest roads on habitat use by Roosevelt elk.
Northwest Science, 59: 122–125.
WOFM 2012 Get GIS Data. Www document: http://www.ofm.wa.gov/pop/geographic/tiger.asp.

52

Appendix

Figure A 1: An overview of the subsections that follow (Figures A2 – A5) for the Nooksack
version of the “Elk Nutrition Toolbox” model.
53

Figure A 2: A close-up view of a portion of the nutrition model for the Nooksack Region (1 of 4).
54

Figure A 3: A close-up view of a portion of the nutrition model for the Nooksack Region (2 of 4).
55

Figure A 4: A close-up view of a portion of the nutrition model for the Nooksack Region (3 of 4).
56

Figure A 5: A close-up view of a portion of the nutrition model for the Nooksack Region (4 of 4).
57

Figure A 6: An overview of the subsections that follow (Figures A7 – A13) for the “Elk Covariates
Toolbox”.
58

Figure A 7: A close up view of the Elk covariates model (1 of 7), the portion that calculates the
distance to cover/forage edges.
59

Figure A 8: A close up view of Elk covariates model (2 of 7), the continuation of the portion that
calculates the distance to cover/forage edges.
60

Figure A 9: A close up view of the Elk covariates model (3 of 7), the continuation of the portion
that calculates the distance to cover/forage edges.
61

Figure A 10: A close up view of the Elk covariates model (4 of 7), the continuation of the portion
that calculates the distance to cover/forage edges.
62

Figure A 11: A close up view of the Elk covariates model (5 of 7), the continuation of the portion
that calculates the distance to cover/forage edges.

Figure A 12: A close up view of the Elk covariates model (6 of 7), the portion that calculates the
distance to roads open to the public.

63

Figure A 13: A close up view of the Elk covariates model (7 of 7), portion of that calculates the
mean slope.

Abstract (if available)

Abstract The North Rainier Elk Herd (NREH) is one of ten designated herds in Washington State, all managed by the Washington Department of Fish and Wildlife (WDFW). To aid in the management of the herd, the WDFW has decided to implement a spatial ecosystem analysis. This thesis partially undertakes this analysis through the use of a suite of software tools, the Westside Elk Nutrition and Habitat Use Models (WENHUM). This model analyzes four covariates that have a strong correlation to elk habitat selection: dietary digestible energy (DDE)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Utilizing GIS and remote sensing to determine sheep grazing patterns for best practices in land management protocols

PDF

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

PDF

Habitat suitability modeling of Mexican spotted owl (Strix occidentalis lucida) in Gila National Forest, New Mexico

PDF

3D visualization models as a tool for reconstructing the historical landscape of the Ballona Creek watershed

PDF

Spatial distribution of the endangered Pacific pocket mouse (Perognathus ssp. pacificus) within coastal sage scrub habitat at Dana Point Headlands Conservation Area

PDF

Evaluating glacier movement fluctuations using remote sensing: a case study of the Baird, Patterson, LeConte, and Shakes Glaciers in central southeastern Alaska

PDF

Predictive habitat distribution modeling of sperm whale (Physeter macrocephalus) within the central Gulf of Alaska utilizing passive acoustic monitoring

PDF

Remote analysis of avalanche terrain features: identifying routes, avoiding hazards

PDF

Using Landsat and a Bayesian hard classifier to study forest change in the Salmon Creek Watershed area from 1972–2013

PDF

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

PDF

Modeling open space acquisition in Boulder, Colorado

PDF

Integration of topographic and bathymetric digital elevation model using ArcGIS interpolation methods: a case study of the Klamath River Estuary

PDF

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

PDF

Is the likelihood of waterfowl presence greater on conserved lands?

PDF

An evaluation of soil sampling methods in support of precision agriculture in Northeastern North Carolina

PDF

Evaluating transit and driving disaggregated commutes through GTFS in ArcGIS

PDF

Using Maxent modeling to predict habitat of mountain pine beetle in response to climate change

PDF

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

PDF

Alaska Hike Search: designing a mapping application for Alaskan trails and user contributed hazards

PDF

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

Asset Metadata

Creator Benton, Joshua J. (author)

Core Title An analysis of the North Rainier Elk Herd area, Washington: change detection and habitat modeling with remote sensing and GIS

School College of Letters, Arts and Sciences

Degree Master of Science

Degree Program Geographic Information Science and Technology

Publication Date 02/08/2013

Defense Date 01/17/2013

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

Tag ArcGIS,Elk,GIS,habitat,Model,OAI-PMH Harvest,remote sensing

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Longcore, Travis R. (committee chair), Paganelli, Flora (committee member), Wilson, John P. (committee member)

Creator Email jjbenton@usc.edu,joshuabenton3@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-219366

Unique identifier UC11294080

Identifier usctheses-c3-219366 (legacy record id)

Legacy Identifier etd-BentonJosh-1432.pdf

Dmrecord 219366

Document Type Thesis

Rights Benton, Joshua J.

Type texts

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

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

Repository Name University of Southern California Digital Library

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