Monitoring Floodplain Restoration Using UAV Lidar and 2D Hydraulic Modeling on the Greenwater River, Washington

Item

Title

Eng Monitoring Floodplain Restoration Using UAV Lidar and 2D Hydraulic Modeling on the Greenwater River, Washington

Date

2018

Creator

Eng Zierdt, Brian

Subject

Eng Environmental Studies

extracted text

MONITORING FLOODPLAIN RESTORATION USING UAV LIDAR AND
2D HYDRAULIC MODELING ON THE GREENWATER RIVER, WASHINGTON

by
Brian Zierdt

A Thesis
Submitted in partial fulfillment
of the requirements for the degree
Master of Environmental Studies
The Evergreen State College
September, 2018

© 2018 by Brian Zierdt. All rights reserved.

This Thesis for the Master of Environmental Studies Degree
by
Brian Zierdt

has been approved for
The Evergreen State College
by

________________________
Edward A. Whitesell, Ph. D.
Member of the Faculty

________________________
Date

ABSTRACT

Monitoring Floodplain Restoration Using UAV Lidar and 2D Hydraulic Modeling on the
Greenwater River, Washington
Brian Zierdt
Anthropogenic changes to the landscape have reduced both fish habitat and the
natural flood protection of streams and rivers. Shifting trends in river discharge also
present an increased risk to salmon survival, and highlight the importance of floodplain
restoration projects to boost resiliency to climate change. Lidar-derived topographic data
input into a hydraulic model can be utilized to quantify the benefits of floodplain
restoration. UAV lidar technology can provide more detailed topographic outputs than
conventional lidar flown with manned aircraft. This study used both conventional and
UAV lidar within a 2D hydraulic model, run using HEC-RAS 5.0.3, to analyze how well
the Greenwater River Floodplain Restoration Project achieved proposed floodplain
reconnection and velocity reduction goals. Second, it explores the potential benefits of
using high-resolution UAV lidar. Results show that the Greenwater River restoration had
a positive impact on project metrics, with an 8.2% gain in floodplain inundation area and
an 8.5% reduction of velocities in the main channel at the 100-year flood. Dense tree
canopy in the project area reduced the potential 1 cm detail of the UAV lidar output,
resulting in a 1-foot DEM. A comparison of model results run on the native post-project
terrain and a downsampled 3-foot terrain, the resolution of the pre-project data, resulted
in very little change in spatial patterns, with a 1.3% reduction of inundation area and a
0.5% reduction in velocities across the floodplain at the 100-year flood. Benefits of highresolution UAV lidar for the production of detailed roughness values and the assessment
of fine-scale habitat features is discussed, although the latter would likely require the
capture of blue-green bathymetric lidar and not only near-infrared lidar captured for this
study. UAV lidar is ultimately shown to be a cost-effective method of obtaining a highly
detailed topographic model for smaller projects of a few 100 acres or less.

Table of Contents

1.0 INTRODUCTION ....................................................................................................... 1
2.0 LITERATURE REVIEW ............................................................................................ 8
2.1 Introduction .............................................................................................................. 8
2.2 Monitoring the Effectiveness of Stream Restoration ............................................... 9
2.3 Climate Change and Salmon.................................................................................. 12
2.4 Hydraulic Modeling Using Lidar ........................................................................... 15
2.5 Summary ................................................................................................................ 20
3.0 METHODS ................................................................................................................ 21
3.1 Introduction ............................................................................................................ 21
3.2 Study Area ............................................................................................................. 22
3.3 Data ........................................................................................................................ 23
3.3.1 Lidar ................................................................................................................ 24
3.3.2 Discharge and Basin Statistics ........................................................................ 28
3.3.3 Aerial Imagery ................................................................................................ 31
3.3.4 Land Cover and Manning’s n Roughness Values ........................................... 33
3.4 Modeling ................................................................................................................ 36
3.5 Results Analysis ..................................................................................................... 40
4.0 RESULTS .................................................................................................................. 41
4.1 Introduction ............................................................................................................ 41
4.2 Inundation Area ..................................................................................................... 43
4.2.1 Inundation Area - Mean Annual Minimum .................................................... 45
4.2.2 Inundation Area – Bankfull ........................................................................... 46
4.2.3 Inundation Area – 5-Year .............................................................................. 47
4.2.4 Inundation Area – 10-Year ............................................................................ 48
4.2.5 Inundation Area – 25-Year ............................................................................ 49
4.2.6 Inundation Area – 50-Year ............................................................................ 50
4.2.7 Inundation Area – 100-Year .......................................................................... 51
4.3 Flow Velocity......................................................................................................... 52
4.3.1 Flow Velocities – 10-Year .............................................................................. 52
4.3.2 Flow Velocities – 100-Year ............................................................................ 54
4.4 DTM Resolution Effect on Hydraulic Modeling ................................................... 56
iv

5.0 DISCUSION .............................................................................................................. 59
5.1 Introduction ............................................................................................................ 59
5.2 Pre- to Post-Project Comparisons .......................................................................... 60
5.2.1 Inundation Area .............................................................................................. 61
5.2.2 Flow Velocities ............................................................................................... 62
5.3 Ecological Importance ........................................................................................... 63
5.3.1 Salmon Populations ........................................................................................ 64
5.3.2 Changes in Peak Flood Occurrence ................................................................ 66
5.3.3 Ecological Interaction ..................................................................................... 68
5.4 DTM Resolution Effect.......................................................................................... 69
5.5 Cost Benefit ........................................................................................................... 71
5.6 Future Research ..................................................................................................... 72
6.0 CONCLUSION .......................................................................................................... 74
REFERENCES ................................................................................................................. 76
APPENDICES .................................................................................................................. 81

v

List of Figures
Figure 1.1. Location of Greenwater River Floodplain Restoration .....................................5
Figure 2.1. Oblique view of the three-dimensional lidar point cloud of all laser returns,
2007 Greenwater restoration area ....................................................................18
Figure 3.1. Photos of 2017 lidar drone flight ...........................................................................25
Figure 3.2. Pre- and post-project lidar terrain used for hydraulic modeling ............................27
Figure 3.3. Greenwater basin drainage areas ...........................................................................29
Figure 3.4. 2017 Aerial imagery ..............................................................................................32
Figure 3.5. National Land Cover Dataset within the Greenwater project area ........................33
Figure 3.6. Sample of data used to delineate Manning’s n roughness values ..........................35
Figure 3.7. Depiction of combined structured and unstructured computational mesh.
Breakline along FR 70 road surface is in red. .......................................................38
Figure 3.8. Example of cell and cell face detailed hydraulic tables (Brunner, 2016) ..............39
Figure 4.1. Engineered log jam locations shown on the post-project bankfull inundation......42
Figure 4.2. Inundation areas for all flood events .....................................................................44
Figure 4.3. Mean annual minimum inundation areas pre- and post-project ............................45
Figure 4.4. 1.6-year bankfull inundation areas pre- and post-project ......................................46
Figure 4.5. 5-year inundation areas pre- and post-project .......................................................47
Figure 4.6. 10-year inundation areas pre- and post-project .....................................................48
Figure 4.7. 25-year inundation areas pre- and post-project .....................................................49
Figure 4.8. 50-year inundation areas pre- and post-project .....................................................50
Figure 4.9. 100-year inundation areas pre- and post-project ...................................................51
Figure 4.10. 10-year velocities pre- and post-project ..............................................................53
Figure 4.11. 100-year velocities pre- and post-project ............................................................55

vi

Figure 4.12. 100-year inundation areas on post project 1-foot and 3-foot gridded terrain ..... 56
Figure 4.13. Mean annual minimum inundation areas on post-project 1-foot and 3-foot
gridded terrain....................................................................................................... 58
Figure 5.1. StreamNet fish distributions by species ................................................................ 65
Figure 5.2. Annual peak streamflow 1912–2017 (top), Streamflow 2008–2017 (middle),
Annual peak flow events > 10-year flood per decade (bottom) ........................... 67
Figure 5.3. Relationship between peak flow events and salmon life histories in the
Greenwater River .................................................................................................. 69

vii

List of Tables
Table 3.1. Drainage areas and peak flood levels......................................................................30
Table 4.1. 10-year velocities pre- to post-project ....................................................................52
Table 4.2. 100-year velocities pre- to post-project ..................................................................54
Table 4.3. Inundation areas on post-project 1-foot and 3-foot gridded terrain ........................57
Table 4.4. Mean velocities on post-project 1-foot and 3-foot gridded terrain .........................58
Table 5.1. Lidar cost comparison .............................................................................................72

viii

Acronyms
cfs: cubic feet per second
DNR: Department of Natural Resources
DEM: digital elevation model
DSM: digital surface models
DTM: digital terrain model
ESU: evolutionarily significant unit
ESA: Endangered Species Act
ELJ: engineered log jam
FIPS: Federal Information Processing Standard
FR 70: Forest Road 70
GFDL: Geophysical Fluid Dynamics Laboratory
HEC-RAS: Hydrologic Engineering Center’s River Analysis System
HECI: Herrera Environmental Consultants, Inc.
LWD: large woody debris
NAIP: National Agriculture Imagery Program
NIR: near-infrared
NAD83: North American Datum (1983)
NAD88: North American Vertical Datum (1988)
NLCD: National Land Cover Database
NOAA: National Oceanic and Atmospheric Administration
RCO: Recreation and Conservation Office
SPSSEG: South Puget Sound Salmon Enhancement Group
TIN: Triangulated Irregular Network
UAV: unmanned aerial vehicle
USGS: U.S. Geological Survey
WDFW: Washington Department of Fish and Wildlife
WSDOT: Washington State Department of Transportation
WRIA: Watershed Resource Inventory Areas

ix

x

Acknowledgements

I would most like to thank my wife Shawna, who has supported me through the
many years of transitioning my career back to that which I had always imagined for
myself, working towards a goal of restoring and conserving our natural world. I would
like to thank my three children, Nora, Jonas, and Ula, who gave up a bit of their dad to
this transition, and who, along with all of my family and friends are as excited as I am for
the completion of this thesis and my MES. I would also like to thank the team at the
South Puget Sound Salmon Enhancement Group for all of their knowledge and support
through my master’s work, as well as the opportunity to continue this work
professionally. Finally, I would like to thank my thesis reader, Ted Whitesell, for all of
his guidance and patience with me throughout the writing process.

xi

1.0 INTRODUCTION
Rivers and streams in the Pacific Northwest have been home to Pacific salmon
species for over 6 million years (Waples, Pess, & Beechie, 2008). These salmon play an
important role as a keystone species within the aquatic ecosystem that they are part of
and the terrestrial riparian ecosystem that they move through. In the past century
overharvesting and anthropogenic changes to the landscape, resulting in separation from
and degradation of habitat, have led to the elimination of Pacific salmon across 40% of
their historic range and reduced returns to 6-7% of their historic numbers (Gresh,
Lichatowich, & Schoonmaker, 2000). Dams, culverts, and levees block rivers and change
flow patterns, logging and other forms of deforestation have removed thermal protection
and food sources, and agricultural and urban stormwater runoff is polluting waters that
are detrimental to the survival of native salmon populations. In response, 6 salmonid
species, which include 18 evolutionarily significant units (ESU), have been listed under
the Endangered Species Act (ESA) in Washington State since 1991 (RCO, 2009).
Recovering salmon populations through restoration to increase the health and natural
function of our aquatic systems is a unifying goal across many government, tribal, and
nonprofit entities. To accomplish this over a billion dollars is spent annually on river
restoration projects in the US and it is a large focus for environmental management and
policy decisions (Bernhardt et al., 2005).
With so much spent on salmon restoration, it is important to monitor the efficacy
of these projects to ensure that we adapt our management practices to be most effective.
These efforts not only help support fish and river systems but increase the ecosystem
services we get from these precious resources. From clean water and flood resilience to
1

fishing and recreation, it is imperative that we continue revitalizing and restoring our
natural waterways. In managing this work, it is also important to take into account the
trends of a shifting climate.
Climate models show a distinct shift to warming temperatures caused by
anthropogenic influences increasing greenhouse gases and changing the global carbon
cycle, and recent global levels have already surpassed all other climate anomalies over
the past 1500 years (Mann et al., 2009). When linking models of future climate, land
cover, hydrology, and salmon populations, a large negative impact is seen to occur in
freshwater salmon habitat and river basins that are fed by the current snowline, and
salmon populations in these basins become especially vulnerable as they are faced with
higher winter flows and lower summer flows (Battin et al., 2007; Mantua, Tohver, &
Hamlet, 2010). Salmon have adapted and survived many fluctuations in global climate
throughout their existence, but current levels of anthropogenic climate change are
occurring at a much faster rate than the natural global climate cycles, and natural
adaptations will likely not be able to keep up with the current rate of change. These
changes are happening now, and we have already begun to see the results of a shifting
climate, further highlighting the need for increased understanding of the efficacy of our
restoration efforts.
Many methods have been developed to monitor the effectiveness of stream
restoration. Primarily these involve on the ground surveys, but remote sensing and
computer models have the potential to capture and predict the results of restoration, and
may be able provide valuable information when constraints on time, budget or access
prevent ground surveys and monitoring. As well, ground surveys can only capture the

2

conditions that exist on that day, while models—although not perfect—can provide a
snapshot of multiple theoretical conditions.
The goal of this thesis is to examine how new advancements in drone based lidar
and modeling technology can be utilized to quantify and visualize the results of stream
restoration efforts, particularly how well the Greenwater River Floodplain Restoration
Project, located north of Mount Rainier in Washington State, was able to achieve project
goals, with a focus on the reconnection of the floodplain and seasonal side channels in
order to reduce high flow velocities and increase flood resiliency by inserting woody
debris and spreading flow out across the floodplain. To accomplish this, a 2D hydraulic
model, which predicts two-dimensional, multi-directional flow across a threedimensional terrain, was utilized to compare the area of inundation of the floodplain and
the flow velocities at various flood stages before and after restoration. Topographic data
for the hydraulic model were acquired from lidar datasets, incorporating a pulsed laser
and receiver to measure distance and ultimately create a three-dimensional model of the
target area, flown both pre- and post-project. The post-project lidar acquisition was
collected by an unmanned aerial vehicle (UAV), more commonly known as a drone,
which can provide very high-resolution topographical data for analysis. Possible benefits
and uses of this higher resolution data are an additional goal explored in this research.
Project changes to the landscape in the Greenwater River restoration were found to be
successful in meeting project goals, and UAV lidar was found to provide a more costeffective option, yielding a more detailed terrain model, useful not only in modeling
floodplain inundation and flow velocities, but with the potential to provide insights into
vegetation cover and instream habitat as well.

3

The project area of the Greenwater River was identified and funded for restoration
primarily due to anthropogenic modifications to the landscape that separated the river
from its flood plain and degraded salmon habitat. The project area is located along the
border of Pierce and King Counties in the Mt Baker-Snoqualmie National Forest,
Washington, and is a tributary to the White River, which feeds into the Puyallup River
before emptying into the Puget Sound. See Figure 1.1. Historically the Greenwater
Watershed supported healthy populations of fish, and was one of the essential spawning
areas in the White River watershed for threatened Spring Chinook (Laurie, 2002). In the
1960s, clear-cut logging activities around the Greenwater River removed all but some
small stands of trees close to the river. In December of 1977, a rain-on-snow event
generated a record peak flow of 10,500 cubic feet per second (cfs). The flood flushed
large logs, landslide debris and remnant logging material downstream, with much debris
racking up on the Highway 410 Bridge, leading to record flooding in the town of
Greenwater. By 1979, reactions to the flooding led managers to remove all woody debris
from the river greater than 3-inches in diameter and 3-feet in length. The lack of riparian
forests and instream wood led to a decrease in fish habitat caused by increased water
velocities and shear stresses scouring the river bed, resulting in an incised main channel
further removed from its floodplain. Restoration of the Greenwater River would be
focused on improving aquatic and riparian habitat for currently threatened populations of
Spring Chinook (O. tshawytscha) and steelhead (O. mykiss) utilizing the project area
alongside Coho Salmon (O. kisutch) (Abbe, Beason, & Bunn, 2007; Ecology, 1998;
Marks et al., 2016). Restoration efforts within the project area were performed between
2010 and 2014.

4

Figure 1.1. Location of Greenwater River Floodplain Restoration

The Greenwater River restoration project involved a number of restoration
activities including large woody debris (LWD) placement in the form of 17 engineered
log jams (ELJ), the removal of an abandoned forest road posing a barrier between the
river and its floodplain, and riparian plantings. Successful restoration would increase
floodplain connectivity and off-channel habitat. ELJs provide increased roughness,
promote activation of relic side channels, encourage natural wood and sediment
recruitment, and increase pool frequency—all with the goal of improving salmon habitat
(Abbe et al., 2007; Cramer et al., 2012). Riparian plantings provide habitat complexity
and future thermal protection. In monitoring the results of this and other projects,

5

decision makers should be able to use that knowledge to help identify the best restoration
methods to use and the most beneficial areas to concentrate efforts.
This thesis first provides a review of the applicable literature. It then outlines
research methods before providing model results. Next, a discussion of the results and
their relevance is provided, followed by a conclusion of the findings. The literature
review first examines the efficacy of stream restoration projects and the need for
monitoring in order to be able to best adapt our practices to be most effective. It then
provides a more in-depth look at how climate change is affecting salmon populations in
the Pacific Northwest to illustrate the need for successful restoration to add resiliency to
salmon-bearing streams and rivers. Finally, previous research on methods and the use of
lidar-derived topographic models within a 2D hydraulic model is explored.
The methods chapter first provides additional details on the Greenwater River
study area. It then outlines the data used to drive the hydraulic model, including lidar,
flood discharge levels, and roughness values. Finally, specific model parameters are
discussed followed by a description of the methods of analysis. Results are then
presented, providing a quantified description of inundation area for various flood events
comparing pre- and post-project condition results. This is followed by modeled flow
velocities across the floodplain and within the spawning channel at the 10 and 100-year
events. Inundation results are then compared on the post-project terrain for the native 1foot resolution compared to the same terrain downsampled to a 3-foot grid cell.
The discussion chapter first considers the pre- to post-project comparisons of
inundation and flow velocities, highlighting the specific improvements accomplished by
the Greenwater River restoration. It then looks at specific fish life histories within the

6

Greenwater River and how those relate to recent increases in peak flood occurrences,
attributed to climate change. This presents a direct correlation between high flow events
and fish presence, showing the importance of this and other similar projects to reduce
flow velocities for incubating and rearing fish. The small effect of the studied terrain
resolution difference on model results is then discussed followed by the cost benefit of
UAV to conventional lidar for the Greenwater River restoration and other projects
covering a few 100 acres or less. Lastly, recommendations to improve future research are
explored, including use of the UAV lidar point cloud to determine detailed roughness
values, along with the capture of blue-green bathymetric lidar, to provide insight into
fine-scale habitat features generally captured through on the ground, instream surveys.

7

2.0 LITERATURE REVIEW
2.1 Introduction
In this thesis 2D hydraulic modeling is used to analyze how well the Greenwater
River Floodplain Restoration Project met projected goals related to floodplain and side
channel activation, and the reduction of flow velocities. This contributes to the general
knowledge regarding similar restoration efforts, as well as giving specific insight into the
gains of this and future restoration plans in the Greenwater Basin. The use of a lidarderived digital terrain model (DTM) as the primary input into the hydraulic model has
been utilized in many previous studies to assess restoration efforts or flood risk (Herrera
Environmental Consultants, Inc., 2010; Khattak et al., 2016; Quiroga, Kure, Udo, &
Mano, 2016; Yang, Townsend, & Daneshfar, 2006). Recent advancements in drone and
lidar technology are able to provide more detailed models of terrain than previously
available. The possible benefits of UAV lidar for both cost effectiveness and providing a
more precise terrain for use in hydraulic models and other analysis are also examined to
provide restoration practitioners with information on the advantages and uses of this
relatively new method of obtaining a very detailed DTM. Trends of a shifting climate
causing changes in discharge patterns are also analyzed to highlight the need for
restoration in the face of climate change.
Past studies were identified to give a better understanding of the need for
monitoring restoration projects, as well as the uses and capabilities of hydraulic models.
This chapter reviews the general effectiveness of restoration projects leading to the need
for monitoring, the impacts of climate change on Washington salmon, and describes

8

techniques and uses of lidar in 2D hydraulic modeling, all providing a framework of how
this research fits into the current knowledge base.

2.2 Monitoring the Effectiveness of Stream Restoration
In the late 20th century, the need to further improve salmon recovery efforts
became evident. Thorough monitoring and analysis of the results of stream habitat
restoration methods was not occurring, and their effectiveness was highly debated by the
scientific community (Reeves et al., 1997). As we move further into the 21st century,
most projects are still either not monitored or are poorly monitored (Bernhardt et al.,
2005; O’Neal, Roni, Crawford, Ritchie, & Shelly, 2016). By monitoring the outcomes of
restoration projects, management practices can be adapted to give the most desired results
based on scientific evaluation. New methods and tools continue to be developed that can
help practitioners in their monitoring efforts. This information can be used to plan future
projects and set meaningful project goals, which should increase success and maximize
effectiveness. Prior monitoring of restoration projects with similar aspects as the
Greenwater restoration gives some insight into the expected results of restoration.
O’Neal et al. (2016) statistically assessed the success and effectiveness of 65
projects in the Pacific Northwest, across multiple project categories including fish
passage, instream habitat, riparian planting, and floodplain enhancement. Elements of the
Greenwater River Floodplain Restoration Project included floodplain enhancement and
instream habitat improvement, through the use of ELJ placements, topographic
modifications, and riparian plantings. Although this thesis investigates metrics not
specifically addressed by O’Neal et al. (2016), the likely benefits of the Greenwater
restoration are identified. In the O’Neal et al. (2016) study, the timeline of post-project
9

monitoring was estimated based on how much time a given restoration category would
need to produce detectable results. For example, fish passage barrier removal projects
were expected to show an impact soon after implementation so they were monitored at 1,
2, and 5 years after completion. Habitat projects such as LWD installations were
expected to take longer before results could be seen and monitoring was scheduled to
occur at 1, 3, 5, and 10 years after implementation. This research on the Greenwater
River represents a 3-5 year post-project evaluation.
Instream habitat projects involving the placement of structures, such as ELJs,
generally show improvements in the habitat indicators being assessed such as pool area,
depth, sediment and wood volumes. In the O’Neal et al. (2016) study, the biologic
response of fish numbers reported a general negative trend with juvenile Chinook and
Coho Salmon being slightly negative but insignificant, and steelhead showing a
significant negative trend in relation to placement of instream structures. Along with
showing positive habitat indicators, structure placement is seen as successful when after
the fifth year 90% of the structures are still in place, which is still the case for the
Greenwater restoration. The lack of improvement in fish numbers, along with some
negative responses, could be because salmonid populations need longer to respond, adapt,
and recover from changing habitat conditions. Similar negative fish responses have been
noted in other studies (Stewart, Bayliss, Showler, Sutherland, & Pulin, 2009; Whiteway,
Biron, Zimmermann, Venter, & Grant, 2010). This may also be pointing to the possibility
of limiting factors that should be addressed elsewhere in the system, causing a general
negative trend in fish populations throughout the watershed. Even though the effect of
these structures on fish populations has not been positively correlated, resulting

10

improvements to habitat features continue to foster the popularity of instream structure
projects.
Floodplain enhancement projects have been found to increase the bank-full width,
flood-prone width and mean canopy density. Fish densities assessed for these projects
were fairly low across most of the sites assessed, with some increases in Coho Salmon
densities. Off-channel habitat found in floodplains is thought to provide a velocity refuge
for juvenile fish (Beechie, Liermann, Pollock, Baker, & Davies, 2006). The benefits of
floodplain enhancement and connectivity projects may also be able to minimize the
scouring effects of high flows during periods of flooding (O’Neal et al., 2016).
Riparian planting projects assessed show an increase in woody species cover and
exceeded plant survival criteria. The percent canopy cover did not change in the 5 years
of monitoring done by O’Neal et al. (2016) and will most likely need significantly more
time to show an increase. As well, no differences were noted in a reduction of active bank
erosion after 5 years and will likely also require more time to see results. Because
riparian planting projects require a longer timescale, fish densities were not looked at for
those projects. Riparian plantings have been shown to be ecologically beneficial but are
difficult to prove significant change due to the long timescale needed for planted
vegetation to mature. Even though an immediate ecological response to these projects is
not seen, they still provide a potential long-term benefit to future changes in flow and
stream temperature that are likely to occur due to climate change.
Overall the study performed by O’Neal et al. (2016) showed that instream habitat,
floodplain enhancements, and riparian plantings, which were all part of the Greenwater
restoration, led to significant improvements in physical habitat after 5 years, even though

11

increased fish densities did not necessarily correlate with these projects. These results
give us some insight into the effectiveness of these projects to meet goals of increasing
the overall ecological and functional health of our waterways, but the biological response
of salmonids to restoration is the primary factor that we are concerned with and is stated
to be the “ultimate measure of restoration effectiveness” (O’Neal et al., 2016). Due to the
large variability in the interannual abundance of salmonids, monitoring for 10 years or
more is recommended to truly observe the effectiveness of restoration (Bisson, Quinn,
Reeves, & Gregory, 1992; Reeves et al., 1997). As we plan for future projects, we should
consider these results along with our knowledge of salmon life histories and the ecology
of the rivers in which they incubate, rear, migrate through, and hopefully return to spawn
in. Continuing to hone and develop monitoring methods along with growing the database
of results should provide the tools needed to be most effective in adapting our
management of streams and river. This becomes especially important in order to increase
resilience of fish-bearing streams to a future, unknown climate.

2.3 Climate Change and Salmon
In the past century, human activities including overfishing and changes to the
landscape have led to reduced, threatened and endangered salmon populations. Many
goals of restoration target historic conditions at a time before modern human
disturbances. However, changes in our global climate that are predicted to occur in the
relatively near future may dramatically change how rivers function, and managers should
consider more than just restoring rivers to their historic state. Land use shifts and
unprecedented climate change are also leading to changes in biodiversity that can make
the goal of restoring to a past environment unrealistic and ineffective (Choi, 2007).
12

Hence, we must consider more “forward-looking” paradigms that include enhancing
ecosystem services and increasing resilience in the face of future climate change (Suding,
2011). This may be accomplished by focusing on the abundance of target species relative
to project areas, the composition of native species, and healthy ecological processes
(Thorpe & Stanley, 2011).
Shifting climates within the greater Pacific Northwest and specifically the
Greenwater River basin are following trends predicted by future climate models. The
Northwestern U.S. has warmed between 0.7° to 0.9° Celsius (C) during just the 20th
century, in contrast to the 1° C in warming over the previous millennium, and climate
models predict another 1.5° to 3.2° C in warming by the middle of the 21st century
(Mann et al., 2009). Results modeled by Battin et al. (2007) in the Snohomish River
Basin led to consistently negative impacts on freshwater salmon habitat, including higher
water temperatures, lower spawning flows, and increased winter peak flows. These
models predict a decline of Chinook salmon populations by 20-40% by 2050 in the
absence of further habitat restoration, with the greatest effect being seen during spawning
and incubation periods in the high-elevation areas, due to the impact on egg survival by
increased peak flows. This predicted negative effect of climate change may be
conservative as they did not model the impact of sea level rise and ocean warming that
will likely also decrease the salmon’s survival. When associating Geophysical Fluid
Dynamics Laboratory’s (GFDL) R30 climate model results with restoration plans, it was
shown that by completing a full suite of restoration efforts we could limit the population
declines to 5% with a possibility of increasing salmon abundance when using the Hadley
Center’s HadCM3 model (Battin et al., 2007).

13

A study by Mantua et al. (2010) assessed the hydrologic changes in watersheds
across Washington State, and how predicted changes would affect the reproductive
success of salmon. Averages based on 19 scenarios predicted increases in annual
temperature in the Pacific Northwest compared to the 1980s to be 1.2° C by the 2020s,
1.9° C by the 2040s, and 3.2° C by the 2080s. Averaged annual precipitation change was
small, but models predicted large seasonal changes towards wetter winters and drier
summers. Hydrologic modeling showed a complete loss of snowmelt dominant basins
across Washington by the 2080s, with only 10 basins in the North Cascades remaining as
transient basins, fed by a mix of rain and some snow. Many of Washington’s current
transient runoff basins, including the Greenwater River Basin, are predicted to be fed
primarily by rainfall, which will lead to a dramatically increased magnitude and
frequency of flooding in the months of December and January (Mantua et al., 2010).
Mantua et al. (2010) lists the effects on salmon as follows. Significant stream
temperature increases will lead to thermal stress for all salmon that have a life history that
puts them in freshwater during summer for spawning, rearing, or smolt migrations. This
will be most severe for salmon populations that have summertime migrations that rely on
thermal cues to initiate spawning migration. As well, the loss of adequate rearing habitat
caused by increased stream temperatures will negatively affect both summer and winter
runs of stream-type Chinook, Coho Salmon, and steelhead, which spend at least one
summer—typically two for steelhead—rearing in freshwater streams. The movement
away from snowfall to rain, increasing the magnitude of winter flooding, will have a
varying impact across species, depending on the depth of the gravel spawning nests, or
redds, they create. Deeper redds, generally made by bigger fish, will be less vulnerable in

14

these conditions. A lack of snowmelt will also affect smolt migrations that have evolved
to match the timing of cooler, snow-fed flows. Changes in these thermal timing events
could also lead to a mismatch with the ocean prey and/or predator fields. Cool season
stream temperature changes were not assessed by Mantua et al. (2010), but it is noted that
warming in winter and spring could lead to earlier and longer growing seasons,
increasing the aquatic food-web productivity, which could aid in more rapid juvenile
salmon development rates (Schindler & Rogers, 2009). Considering all the impacts of a
changing global climate on salmon, the resilience of restoration projects becomes even
more important.
These modeled effects of climate change all point to recovery targets becoming
increasingly difficult to meet, as environmental stress on salmon populations increases
(Battin et al., 2007). Ecological resilience will be key to ensuring that restoration is
sustainable and will not require intensive and ongoing intervention in the face of
environmental change (Suding, 2011). The Greenwater restoration has incorporated
methods that add increased resilience to the basin by reducing high flow velocities and
increasing thermal refuge habitat through LWD placement. Post-project lidar input into a
hydraulic model allow for the quantification of many of the benefits gained by the
Greenwater restoration and other stream restoration projects.

2.4 Hydraulic Modeling Using Lidar
One method of monitoring the effectiveness of wood placement and floodplain
reconnection projects is through the use of hydraulic models. If adequate topographic
data are available, from cross-sections or lidar bare-earth models, a hydraulic model can
be developed to examine water flow and floodplain attributes such as inundation and
15

velocity. These attributes are especially important to spawning and rearing salmonid
populations (Jeffres, Opperman, & Moyle, 2008). The results from the hydraulic model
can help us set meaningful goals pre-project and check the efficacy of the completed
restoration project to meet those goals.
There are a number of hydraulic modeling software packages available today.
Terrain data input into these models are generally 1D, 2D, or a combination of the two.
Modeling in 1D solves one-dimensional equations of flow using a sequence of crosssections connected by an interpolated surface on which flow is modeled. Onedimensional models are a more simplified representation of reality (Costabile,
Macchione, Natale, & Petaccia, 2015). When modeling in 1D, flow is solved only in one
dimension, perpendicular to the cross-sections. Hence, 1D modeling only provides a
single water level, velocity and flow rate for each cross-section in the model, while 2D
modeling may show significant variability across the same section. If there are enough
cross-sections available, the 1D model can provide a good representation of the
topography of the riverbed. One-dimensional models also have the advantage of running
computations relatively quickly. One-dimensional models, however, are limited by the
interval between cross-sections and their extent into the floodplain. They also require a
time investment in gathering enough cross-sections to accurately describe the channel.
One-dimensional modeling can be useful to identify detailed descriptions of flow through
the channel, but can find greater use when combined with 2D modeling (Brunner, 2016).
Two-dimensional flood modeling solves for 2D equations of flow, allowing for
flow in any direction across the terrain surface from higher to lower areas. The terrain
input into a 2D model is generally in the form of a DTM, which provides a three-

16

dimensional topographic surface of the entire floodplain. This type of modeling
calculates flow routes, velocity and depth distribution across the floodplain. Twodimensional models can be computationally slower, but are more useful when a detailed
description of the floodplain in required.
Data input into the 2D hydraulic model primarily include terrain data, a stream
discharge hydrograph, and roughness. The terrain is generally captured using lidar (light
detection and ranging) technology. Discharge is available from various USGS stream
gages, and roughness is discussed later in this section. Lidar is a remote sensing method
in which the combination of a pulsed laser, receiving scanner, and highly accurate GPS
receiver are used to accurately measure distances, resulting in a three-dimensional model
of the target environment. Lidar data are output in a point cloud of laser returns, which is
then converted into a Triangulated Irregular Network (TIN) or raster DEM. See Figure
2.1 for a two-dimensional representation of the three-dimensional lidar point cloud.

17

Figure 2.1. Oblique view of the three-dimensional lidar point cloud of all laser returns,
2007 Greenwater restoration area

When there is a need to use previously captured topographic data you are limited
by what is available in your study area. As technology has advanced, the resolutions of
available topographic data have increased over the years. The effect of topographic grid
sizes on hydraulic model outputs should be considered. It has been noted that a higher
resolution terrain does not necessarily output higher quality results (Charrie & Li, 2012;
Costabile et al., 2015). In the study performed by Charrier & Li (2012) a 1-meter lidar
digital elevation model (DEM) was downsampled to 3, 5, 10, 15, and 30 meters, and
hydraulic model outputs were compared. The 3-, 5-, and 10-meter DEMs produced
similar results, within 2%, 3.6%, and 2.8% respectively, to the 1-meter DEM. The 15and 30-meter DEMs both resulted in a 6.8% difference from the target 1-meter DEM.

18

When floodplain inundation from the 1-meter DEM was compared to models run on
USGS 5-, 10-, and 30-meter DEMs differences were -12.6%, -9.3%, and -1.2%
respectively. This suggested that different data sources produced more significant
changes in results than downsampling a single data source. This thesis explores effects
resulting from the next level of topographic resolution difference from 3-foot (approx. 1
meter) to 1-foot resolution. Besides providing the base terrain for the 2D hydraulic
model, lidar outputs can be used to inform roughness values.
Lidar has been shown to be useful in stream and riparian habitat analysis and
monitoring (Cavalli, Tarolli, Marchi, & Fontana, 2008; McKean, Isaak, & Wright, 2009).
Various outputs can be produced from analysis of lidar in the GIS environment. Some of
these outputs can be used in the development of accurate Manning’s n roughness
determinations. Roughness values reflect impedance to flow that occurs on and above the
terrain surface, and can have a significant impact on modeled velocity, depth, and extent
of inundation (Golshan, Jahanshahi, & Afzali, 2016). Vegetation plays a large part in the
roughness of the floodplain, and vegetation heights derived from the difference between
bare-earth and highest hit terrain models are useful in parameterizing roughness (Mason,
Cobby, Horritt, & Bates, 2003; Quang Minh & La, 2011). Lidar intensity and aerial
imagery are also useful in classifying roughness (Quang Minh & La, 2011). The methods
listed above were the primary processes used in this thesis for determining roughness.
Another method of assigning roughness that was not incorporated in this study is
through the inspection of the lidar point cloud. Research produced by Casas, Lane, Yu
and Benito (2010) describes a method for parameterizing roughness by analyzing the subgrid lidar data points above and below the bare-earth lidar surface. This method seems

19

very promising for describing highly detailed changes in roughness, and would be
recommended when modeling to determine fine-scale habitat utilizing subsurface
topography that can be acquired through bathymetric lidar.

2.5 Summary
The use of lidar within a 2D hydraulic model is seen to be a useful tool for
assessing the outcomes of floodplain restoration projects. As we move into a more
pronounced age of climate change, the need to assess these and other restoration projects
in order to adapt and manage our goals and techniques is becoming even more important
in the effort to slow and, hopefully, one day reverse declines in populations of Northwest
salmon. As seen in other successful projects, the Greenwater restoration incorporates
ELJs and topographical modifications, resulting in a reconnection of the floodplain and a
subsequent reduction in flow velocities. Native plantings and LWD placements also lead
to increased habitat and flood resilience. This research first assesses the effectiveness of
the Greenwater River Floodplain Restoration Project. Second, it analyzes the use of UAV
lidar. By monitoring this project as well as exploring the benefits of new drone based
lidar technology utilized in this research, restoration practitioners should be able to make
more informed decisions on when to incorporate these tools into future project
monitoring and planning efforts.

20

3.0 METHODS
3.1 Introduction
One goal of this research is to model the hydrology of the Greenwater River as it
flows through the Greenwater River Floodplain Restoration Project area in an effort to
analyze the effectiveness of restoration. Secondly, this research identifies the effects of a
higher resolution DEM, captured via UAV-mounted lidar, on hydraulic model outputs.
To attain the first goal, a comparison was made between pre- and post-project model
results at various flood stages to investigate the change in floodplain and side channel
connectivity and flow velocities within the channel. The results of this research roughly
mirror the pre-project assessment performed by Herrera Environmental Consultants, Inc.
(HECI) in 2010. This methodology was chosen so that a comparison to the pre-project
assessment’s projected outcomes could be made. Pre-project lidar data were collected in
2007. Post-project conditions were captured by lidar in late 2017.
Hydraulic modeling was done using the US Army Corps of Engineers Hydrologic
Engineering Center’s River Analysis System (HEC-RAS) software version 5.0.3. HECRAS was chosen for use in this research because it is a reputable 2D hydraulic model
provided free of charge by the USGS (Golshan et al., 2016; Khattak et al., 2016). HECRAS was recently updated in February of 2016 to include 2D modeling capabilities,
allowing for lidar data to be used as the primary terrain input into the hydraulic model.
Results from the analysis as described in this thesis are reported to the South Puget Sound
Salmon Enhancement Group (SPSSEG), Washington State Recreation and Conservation
office (RCO), Puyallup Tribe, Muckleshoot Tribe, King County Flood Control District,
Washington Department of Fish and Wildlife (WDFW), Forest Service, National Oceanic
21

and Atmospheric Administration (NOAA), Watershed Resource Inventory Areas
(WRIA) 10/12 Salmon Recovery Lead Entity, and the Puyallup River Watershed
Council.
In addition to determining the effectiveness of the Greenwater River Floodplain
Restoration Project, this research also explores the possible benefits of high-resolution
UAV lidar to improve the accuracy of hydraulic model outputs and other analysis, as well
as for cost efficiency. Drones are able to fly much lower and slower over the terrain,
capturing a denser laser return point cloud than typically achieved from lidar flown by
conventional manned aircraft. Lidar flown with a UAV can thus yield a higher resolution
terrain model as well as a more detailed representation of the vegetation and other
features above the surface. As similarly done by Charrier & Li (2012), downsampling the
2017 lidar data for this research explores the effects of using various resolution terrain
inputs on hydraulic model outputs and parameters. This also provides an understanding
what amount of error might be presented in comparing the lower resolution pre-project to
the higher resolution post-project terrain modeled results.
This chapter gives an overview of the study area, discusses model inputs, details
the methods used to run the hydraulic model, and outlines the methods of analysis
between different model runs. It also discusses how key decisions were made in the
process.

3.2 Study Area
The study area comprises a 1.5-mile reach of the Greenwater River located in
Washington State. See Figure 1.1. The Greenwater River is a fifth-order tributary to the
White River located along the border of Pierce and King Counties in the Cascade
22

Mountains north of Mount Rainier. The entire restoration site is federal land, managed by
the US Forest Service within the Mt. Baker-Snoqualmie National Forest. The Greenwater
River is documented to support spawning and rearing salmonid species, including Spring
Chinook, Coho Salmon, and steelhead (Abbe, Beason, & Bunn, 2007; Ecology, 1998;
Marks et al., 2016). Snorkel surveys of the project reach in 2014 and 2016 observed
rearing Coho Salmon and Chinook Salmon in pools and side channels, and Coho Salmon
were observed to be spawning in the upper reaches of the project area (Brakensiek,
2017). The project reach has seen many negative effects to the riverine ecosystem due to
past logging activities and the clearing of large wood from the river. In an effort to
restore the ecological health of the river, the Greenwater River Floodplain Restoration
Project was started in 2010 with the completion of Phase 3 in 2014. Primary aspects of
the restoration project impacting this research were the construction of 17 engineered log
jams and the removal of a section of Forest Road 70 (FR 70) that separated the river from
part of its floodplain. Riparian plantings also contributed to roughness of the floodplain
and provide future instream cover and habitat complexity.

3.3 Data
The primary data used to create the hydraulic model of pre-project conditions was
a 2007 bare-earth digital elevation model. Additional data used to inform both pre- and
post-project models incorporates river gage discharge, basin statistics, aerial imagery,
landcover, and lidar highest hit digital surface models (DSM). The elevation models of
post-project conditions were created using lidar flown in December, 2017. The Hydraulic
Assessment of Restoration Alternatives: Greenwater River Engineered Logjam Project
Report (HECI, 2010), which modeled 2007 lidar data using FLO-2D modeling software,
23

was used to verify HEC-RAS 2007 model inputs and results. The 2007 data were
remodeled for this research so a more direct and accurate analysis could be made between
the 2007 and 2017 model results. A detailed description of the data and sources is
presented below.

3.3.1 Lidar
In order to determine the effectiveness of the Greenwater River restoration, a
comparison is made between past and present conditions, represented primarily by
elevation models from 2007 and 2017 lidar acquisitions. Watershed Sciences, Inc.
collected 2007 lidar data between May 22-25 for the Washington State Department of
Transportation (WSDOT) and the SPSSEG. Lidar was obtained utilizing a Leica ALS50
Phase II laser system mounted in a Cessna Caravan 208, acquiring >105,000 laser pulses
per second. Lidar points were corrected with a root mean square error of 0.10 feet, a 1sigma absolute deviation of 0.10 feet and a 2-sigma absolute deviation of 0.20 feet
(Watershed Sciences, 2007). Both bare-earth and highest hit models were determined at
3-foot resolution. Data output used the Washington State Plane North Federal
Information Processing Standard area (FIPS) 4601 coordinate system in the 1983 North
American Datum/1988 North American Vertical Datum (NAD83/NAVD88), reported in
US survey feet (Watershed Sciences, 2007). The data were downloaded for this project
from the Washington Department of Natural Resources (DNR) lidar portal. Lidar from
DNR was provided in GeoTIFF format, which could be imported directly into the
hydraulic model as the primary terrain data.
Post-project lidar data were collected by Flight Evolved on December 7, 2017 for
SPSSEG. Lidar was obtained utilizing a Riegl VUX-1 LR mounted on a DJI Matrice 600
24

Pro drone, with the ability to acquire 750,000 laser pulses per second. Lidar points were
corrected with a root mean square error of 0.169 feet and a standard deviation of 0.206
feet. Both bare-earth and highest-hit models were determined at 1-foot resolution. It was
hoped that a higher resolution DTM could be produced but, due to dense canopy in the
project area limiting the amount of laser ground returns making it back to the lidar
device, the point spacing of the bare-earth lidar point cloud would not accurately support
raster resolutions finer than a 1-foot grid. Pictures taken of the 2017 lidar flight are shown
in Figure 3.1.

Figure 3.1. Photos of 2017 lidar drone flight
25

Lidar flights in both 2007 and 2017 collected data using standard near-infrared
(NIR) lidar. Blue-green lidar capable of capturing bathymetry, the terrain under the water
surface, was not available for pre- or post-project conditions. Without bathymetry, the
pool-riffle sequence and subsurface topographical features such as boulders, root wads
and other obstructions that cause friction to water flow must be represented in equations
that drive the hydraulic model through increased Manning’s roughness values (Crowder
& Diplas, 2000). This method can predict average depth and velocity, but is not able to
identify exact flow patterns or fine-scale ecological features in the vicinity of these
obstructions (Crowder & Diplas, 2000). Manning’s n roughness values were originally
tabulated according to numerous factors posing a resistance to flow by Chow (1959). In
essence, results from hydraulic models using channel roughness to replace the absence of
bathymetry data are adequate for reach-scale analysis of floodplain inundation and
average velocities needed for this analysis, but would not provide accurate representation
of fine-scale individual habitat features such as detailed pool/riffle sequences and their
metrics, which are typically captured through instream surveys.
Discharge at the time of the lidar flights was at relatively low flows, allowing for
some, but not all, in-channel features to be captured, and required appropriate roughness
values to accurately model velocities through the wetted channel. Lidar from 2007, flown
during slightly higher discharge than in 2017, and producing a lower resolution 3-foot
grid cell DTM, masked more of the fine-scale topographical features than the higher
resolution 2017 lidar data flown during lower discharge. The HECI (2010) hydraulic
report was faced with the same limitations of the 2007 data. Inspection of the 2007 lidar
data by HECI assessed that it provided a good representation of the topography of the

26

area and was appropriate for the level of detail needed for hydraulic modeling of local
floodplain inundation and velocity. The 2017 lidar, flown during relatively low discharge
should provide a more detailed description of topographical features within the channel.
Various resolutions of the 2017 DTM were modeled to determine possible changes in
these local flow patterns relative to terrain detail. As the size of the grid cell in the DTM
is increased, subgrid level features are lost to an average smoothing of the terrain surface
and variations in local flow patterns within the channel are expected to decrease. Pre- and
post-project bare-earth lidar are shown in Figure 3.2.

Figure 3.2. Pre- and post-project lidar terrain used for hydraulic modeling

27

3.3.2 Discharge and Basin Statistics
The U.S. Geological Survey (USGS) measures discharge flows at various gage
locations and reports these data through the National Water Information System: Web
Interface. USGS river gage station number 12097500, located on the Greenwater River at
Greenwater, Washington, is the closest river gage to the project site, approximately 5
miles downstream. The highest peak flood flow was recorded in November 1977, at a
discharge of 10,500 cfs. Gage daily mean discharge during the 2007 lidar flight window
ranged from 303 to 338 cfs, with an average daily flow of 317 cfs over the 4 day
acquisition period. This discharge is higher than the mean annual flow of 211 cfs,
averaged over 70 years, but well below the bankfull flow of 871 cfs, representing the
stage at which the water level tops the channel before it spills out into the floodplain
(Laurie, 2002). Gage discharge during the 2017 lidar flight recorded at a daily mean
discharge of 210 cfs.
The hydraulic model developed for this research required inflow discharge values
at the upstream end of the project area and two tributaries. Because there is no USGS
flow gage located within the project area, the discharge for the inflows into the model
must be adjusted from the Greenwater River gage at Greenwater, Washington. Basin area
characteristics were determined using data gathered from the USGS StreamStats web
application, which delineates drainage areas for selected locations along stream lines. The
required discharge inflows were determined using the ratio of basin drainage area at the
gage to the basin areas at the inflow points to approximate discharge flow at the inlets for
the various flood stages to be modeled. Basin areas are shown in Figure 3.3 and
calculated discharge values are shown in Table 3.1. This method provides a reasonable

28

estimation when discharge is required at a location upstream of a stream gage. There is a
margin of error in this method as it assumes the same contributing precipitation and
groundwater upwelling across the whole area, and does not take into account snowmelt
contributions primarily located in the upper watershed. Upon reviewing source discharge
values used in the pre-project hydraulic assessment performed by HECI (2010), this
discharge estimation method is observed to be the same method used to estimate previous
modeled values.

Figure 3.3. Greenwater basin drainage areas. Project area flow moves from east to west.
29

Table 3.1. Drainage areas and peak flood levels

It is important to note that models run by HECI (2010) used flood discharge
values calculated through 1996, reported in Abbe et al. (2007), which were partially
sourced from Laurie (2002). The analysis done for this research, using HEC-RAS,
utilized the most current flood values posted by the USGS, determined through 2014, and
adjusted for basin area. There is a significant difference between older peak flood levels
using data through 1996 and current discharge values using data through 2014. This is
most notable for the 100-year flood, which was reduced from 10,534 to 8,320 cfs—a
difference of 2,214 cfs. Combined with computational and procedural differences
between the two modeling software packages, this resulted in further variation in model
results for specific flood levels run on the 2007 lidar data in HEC-RAS versus the
previous model results run in FLO-2D, and should be considered when comparing
results. Because there is no bankfull discharge values at the 1.6-year flood occurrence

30

using the current reported USGS discharge, the HEC-RAS model used the 1.6-year
bankfull discharge reported by (Laurie, 2002). This value is considered acceptable as it
falls within the range between the current 1- and 2-year discharges, and allows for a more
direct comparison between the previous and current model runs on the 2007 terrain.
Another difference between the previous and current models is that previous
models used Greenwater project area outflow values as the inflow at the top of the
project, and did not model tributary inflows separately. There are two significant
tributaries that flow into the Greenwater River within the project area. The current model
used for this research applies discrete tributary flow inputs apart from the Greenwater
model inflow applied to the top of the project reach. Greenwater River discharge flow
values for the top of the project in the HEC-RAS model are based on accumulation from
the basin above the project along with local surface accumulation above the outflow point
not captured in the tributaries of Slide Creek and 28 Mile Creek. This local surface
accumulation is referred to as Project Reach (unnamed flow lines) in Table 3.1,
accounting for 0.82 square miles of area.

3.3.3 Aerial Imagery
Pre-project aerial imagery was gathered from the National Agriculture Imagery
Program (NAIP). Imagery was not available for 2007 so data from 2006 and 2009 were
acquired. Both 2006 and 2009 imagery showed no visible clearing or logging activity and
appeared virtually identical outside of the time of day the image was acquired. Both
images were considered to be a good representation of 2007 conditions. High-resolution
aerial imagery was collected by drone photography on the same day as the 2017 lidar

31

flight, with a resolution of 3-inch grid cells. Samples of the 2017 aerial imagery are
shown in Figure 3.4. Locations of ELJs are highlighted in the top two images.

Figure 3.4. 2017 Aerial imagery

32

3.3.4 Land Cover and Manning’s n Roughness Values
An accurate representation of land cover in the form of Manning’s n roughness
values is used in model calculations to parameterize the surface friction effects on water
flow. Existing land cover datasets are typically used to assign roughness values (Brunner,
2016). Upon inspection of the National Land Cover Database (NLCD) in the project area,
both accuracy and precision of the dataset were determined to be spatially unreliable and
at an unusable scale to inform roughness values for the hydraulic model. As seen in
Figure 3.5, resolution of the data is very low and depicts areas that do not accurately align
with ground features such as the road and river.

Figure 3.5. National Land Cover Dataset within the Greenwater project area
33

Manning’s roughness values across the project area were delineated for this
research using a combination of HECI (2010) roughness values, aerial imagery,
vegetation height models, DTM, and DTM slope. See Figure 3.6 for a sample of these
inputs. Manning’s n tables from Chow (1959) were referenced to validate final roughness
value determinations, while the above inputs were primarily used to digitize precise areas
of land cover. Manning’s n roughness values used in the HECI (2010) model—
determined through field investigations, expert experience and knowledge of roughness
values used in other modeling projects in the area—were utilized as a starting point for
roughness determinations. HECI roughness was provided as a static map that was
georeferenced into ArcGIS to be used as a starting template for delineation of the various
land cover areas. Areas were drawn within the ArcGIS environment. By toggling
between the above mentioned layers within ArcGIS, the new land cover areas were
drawn and appropriate Manning’s n roughness values were assigned. The following
describes the use of the various layers utilized to assign roughness areas and values.
Vegetation heights are useful for parameterizing surface roughness for use in
hydraulic models (Mason et al., 2003). Vegetation height maps were created by
subtracting the bare-earth DTM from the highest hit DSM in ArcGIS. This provides a top
surface height of vegetation and other features relative to ground level. Land cover type
and density were estimated from the vegetation height map in conjunction with aerial
imagery. The DTM and the visualization of DTM slope are useful to identify channels
and their banks when delineating these areas of land cover. Only features within the area

34

to be modeled were delineated, and features such as roads outside of the Greenwater
River floodplain and its tributary inflows were not identified.
Land cover designations for the 2017 model were created in a similar fashion to
those determined for the 2007. Manning’s n values associated with log jams was guided
by Dudley, Fischenich, & Abt (1998) which stated that Manning’s n values increase 39%
when woody debris is present. Channel roughness of n=0.55 was multiplied by this factor
to yield a roughness of n=0.076 for log jams.

Figure 3.6. Sample of data used to delineate Manning’s n roughness values

35

Once all of the necessary data were collected and prepared, hydraulic models of
the project reach were developed. A detailed description of the parameters of the HECRAS hydraulic model can be found in the next section.

3.4 Modeling
Hydraulic modeling for this research was done using HEC-RAS 5.0.3. Modeling
was performed using the 2007 and 2017 lidar at various flood stages and at different
resolutions of the 2017 data. This section discusses the modeling process and
specifications.
Preprocessing of model inputs was done in ArcGIS. To prepare the bare-earth
lidar DTMs for input into the hydraulic model, the area of interest, including the river,
major tributary inflows and the surrounding floodplain, were delineated. This
determination was assisted by following contours well above the floodplain to specify the
project area. The lidar DTM was then clipped to the project area and output in GeoTIFF
format for direct input into HEC-RAS.
When starting a new HEC-RAS project, a projection file must be designated
before any terrain data can be imported. The Greenwater project area is located within
two US State Plane projection zones, although the river and the majority of the floodplain
are located in the northern zone. The projection was set to the NAD83 Washington State
Plane North FIPS 4601 projected coordinate system. This is also the native projection of
the pre-project lidar data (Watershed Sciences, 2007). The lidar DTM could then be
imported into HEC-RAS as a new terrain layer.
HEC-RAS allows for the combination of 1D cross-section data, portraying the
topography under the water surface, with a 2D terrain. No cross-section data were
36

available from 2007, so the hydraulic model was run as a 2D model only and not a
combined 1D/2D model. Without blue-green bathymetry lidar to accurately describe the
subsurface terrain, the lidar DTM depicts increased smoothing and a generalized slope in
areas covered by water. This can be mitigated within the modeling environment by using
informed Manning’s roughness values (Crowder & Diplas, 2000). The land classification
data set with established Manning’s n values described in the above section was hence
imported into the model.
The 2017 data were also run as a 2D model to allow for comparisons to be made.
It should be noted that there was less effect on the 2017 data by the lack of bathymetry,
as the flows were lower during the 2017 lidar flight, exposing more of the channel
topography. Captured at a higher resolution, the 2017 data also reveal more fine-scale
topographic features within the channel that affect water movement.
Within HEC-RAS geometry editor, the 2D-flow area is delineated. High ground
contour lines well outside the possible flow area were again used to select an area that
included the entire floodplain. The 2D-flow area describes the boundaries of the model
and contains the computational mesh. HEC-RAS uses an implicit finite-volume solution
scheme to calculate flow within the 2D-flow area. The algorithm provides improved
stability and robustness over more traditional finite difference and finite element
techniques, and was developed to work with both a structured or unstructured
computation mesh (Brunner, 2016). This allows computations to be made across a
standard gridded mesh with 4-sided cells or one that contains a combination of polygons
with a mixture of 3–8 sides. Various polygonal shapes are created along the border of the
2D-flow area and along breaklines within the flow area. See Figure 3.7.

37

Figure 3.7. Depiction of combined structured and unstructured computational mesh.
Breakline along FR 70 road surface is in red.

Breaklines are added to insure that computational cell faces align properly to
capture high-ground features. It is recommended to add break lines along levees, roads,
and high-ground features (Brunner, 2016). Breaklines were added to the 2007 flow area
along the top surface of the road prism that was later removed during restoration.
The computational mesh was generated on regular intervals with all breaklines.
Cells of the computational mesh in HEC-RAS do not have a flat-bottom, single elevation,
as do some other hydraulic models. Instead HEC-RAS 2D modeling uses a highresolution, subgrid model where each cell face is similar to a detailed cross-section, and
38

cells and cell faces are preprocessed with detailed hydraulic property tables (elevation
versus area, wetted perimeter, and roughness) based on the underlying terrain (Brunner,
2016). See Figure 3.8. This allows for cells to be partially wetted, providing detailed
subgrid level precision to model outputs through the retention of terrain detail while
implementing larger computational cells, resulting in faster model run times. A
computation point spacing of 10 feet was used for this research. After the computational
mesh was defined, flow inputs and outputs were established along the boundary of the 2D
flow area.

Figure 3.8. Example of cell and cell face detailed hydraulic tables (Brunner, 2016)

Within the project area, the Greenwater River flows primarily from east to west,
with one tributary coming in from the north and another from the south. Boundary
conditions are identified to bring flow into and out of the modeled area. The model

39

inflows are identified across the channel at the upstream end of the Greenwater River and
the two tributaries, while the outflow is delineated across the entire floodplain at the
downstream end of the project area. Peak flood discharge values run through the model
inflows were adjusted from the Greenwater USGS gage number 12097500, as seen in
Table 3.1.
The model was run on the 2007, 2017, and downsampled 2017 DTM for the mean
annual minimum, mean annual, 1-year, 1.6-year (bankfull), 2-year, 5-year, 10-year, 25year, 50-year, and 100-year flow events. The default diffusion wave equations were used,
allowing the model to run faster and have increased stability (Brunner, 2016).

3.5 Results Analysis
Model results were exported from HEC-RAS and imported into ArcGIS for
analysis. Results between pre- and post-project conditions were compared at specific
flow events to determine the overall effectiveness as well as how the project held up to
predictions made in the HECI (2010) proposed conditions assessment. The inundation
area for each flow event was calculated in order to determine the gain in wetted area.
Visual inspection of specific events provides insight into morphological changes that
have caused an increase in floodplain and side channel activation. Mean flow velocities
as well as local patterns in flow velocity were investigated for the 10-year and 100-year
events. Finally, a comparison between the native resolution and downsampled 2017
terrain was made to explore the effects of lower resolution DTM on hydraulic model
outputs.

40

4.0 RESULTS
4.1 Introduction
The results of this research first answer the question of the effectiveness of the
Greenwater River Floodplain Restoration Project to achieve floodplain connectivity and
velocity reduction goals. This project removed a section of Forest Road 70, previously
separating the river from part of its floodplain, and added several ELJs within the river.
All ELJs are still intact and functioning at the time of this study. Hydraulic modeling
performed for this research analyzes the effect of project changes to the landscape on
water flow throughout the restoration area. A comparison is made between the preproject (2007) and the post-project (2017) inundation areas and velocities for a range of
flow events from the mean annual minimum to the 100-year flood.
Differences in inundation area and mean velocity are also compared between the
native 1-foot resolution of the post-project DTM and the same dataset downsampled to 3foot grid cells, the pre-project DTM resolution. This was done to answer whether there is
any difference in modeled floodplain delineation at the reach scale with this level of
resolution difference. It gives some insight into benefits of the higher resolution digital
elevation model produced from a drone flight as compared to the lower resolution
elevation model that would be output from lidar flown by manned aircraft.
Maps of the hydraulic model results are broken up into 4 zones labeled A, B, C,
and D, following flow direction from east to west. Zones were established to aid in the
discussion of specific features and changes in flow patterns. Zone A includes the upper
side channel/floodplain connection and 3 ELJs. See Figure 4.1. Zone B includes the side
channel, known as the “wall-based side channel” by project managers, and 2 ELJs. Zone
41

C includes the historic channel connection, the Slide Creek confluence with the
Greenwater River, and the majority of ELJs. Zone D includes the 28 Mile Creek
confluence, the lower floodplain, and the final ELJ.

Figure 4.1. Engineered log jam locations shown on the post-project bankfull inundation

Results are organized as follows. First, inundation area is compared across all
flow events on the pre- and post-project landscape. Inundation area of individual events
comparing pre- and post-project conditions for the mean annual minimum, 1.6, 5, 10, 25,
50, and 100-year flow events are examined to identify changes to flow patterns and
floodplain activation. Velocities are compared for the 100- and 10-year flow pre- and
post-project across the entire floodplain as well as specifically within the mean annual
minimum inundation area, where the scour effect on salmon redds would be greatest.
Finally, digital elevation model resolution effects on modeled floodplain delineation and
velocities are explored to identify effects of a higher resolution terrain on modeled
results.

42

4.2 Inundation Area
The inundation area is the area of land covered by water. Inundation area was
modeled for specific flow events including the mean annual minimum flow, mean annual
flow, 1, 1.6 (bankfull), 2, 5, 10, 25, 50, and 100-year floods. Inundation areas were
calculated within the Greenwater River floodplain only and do not include the 2
tributaries’ inflow areas.
A comparison of inundation area for all flow events is illustrated in Figure 4.2.
Wetted area within the project boundary ranges from 8.6 acres for the Mean Annual
Minimum flow, modeled on the 2007 landscape, to 47.7 acres for the 100-year flood
event, modeled on the 2017 landscape. All flow events show an increase in inundation
area on the 2017 post-project landscape. The area increase from 2007 to 2017 showed a
relatively narrow range of change across all events, from 1.5 acres for the 25-year flood
to 3.6 acres for the 100-year flood. The percent area increased ranged from 4.5% for the
25-year flood to 31.2% for the mean annual minimum flood. Events below the 1.6-year
bankfull flow showed area changes averaging 2.6 acres, but revealed the greatest percent
increase in area ranging from 21.7% to 31.2% and averaging 26.7%. Events at bankfull
flow and above showed similar area increases, averaging 2.4 acres, but revealed smaller
percent area increases ranging from 4.5% to 12.4% and averaging 9.0%.

43

Figure 4.2. Inundation areas for all flood events

44

4.2.1 Inundation Area - Mean Annual Minimum
The mean annual minimum flow is representative of flow rates during spawning
months of Spring Chinook and Coho Salmon in the Greenwater River (Ecology, 1998;
Marks et al., 2016). The inundation area for the mean annual minimum flow, shown in
Figure 4.3, increased 2.7 acres, from 8.6 to 11.2 acres. This represents a 31.2% increase
in wetted area. Zone A shows some local changes in flow patterns likely due to log jam
placements. A comparison between the mean annual minimum inundation areas show the
approximately 1,500-foot side channel in zone B is now fully connected during low flow
conditions post-project. During pre-project conditions this side channel did not connect
until somewhere between the 10- and 25-year events. This is one of the most notable and
significant changes in post-project conditions for the mean annual minimum flow. Zones
C and D reveal a wider, more braided channel post-project, presenting more potential
spawning area for fish and increased potential for wood and sediment recruitment.

Figure 4.3. Mean annual minimum inundation areas pre- and post-project

45

4.2.2 Inundation Area – Bankfull
Bankfull flow represents the stage at which the water level tops the channel
before it spills out into the floodplain. The inundation area for the 1.6-year bankfull flow,
shown in Figure 4.4, increased 1.8 acres, from 15.8 to 17.6 acres. This represents an
11.3% increase in wetted area. Zone A shows a connection of the smaller side channels in
the upper project area. In zone B, the channel has become slightly wider as it fills in.
Zones C and D see more braiding across the floodplain of the tributary inflows of Slide
and 28 Mile Creeks along, with a filling in of the main Greenwater channel.

Figure 4.4. 1.6-year bankfull inundation areas pre- and post-project

46

4.2.3 Inundation Area – 5-Year
The inundation area for the 5-year flow, shown in Figure 4.5, increased 2.6 acres,
from 22.4 to 25.0 acres. This represents an 11.8% increase in wetted area. Zone A shows
a filling in of the upper channel. Flows in zone B have started to spill slightly into the
floodplain. The connection with the historic channel, that joins the Greenwater River to
Slide Creek higher up and more to the north, is seen in zone C. This connection is more
consistent with the original channel on the new post-project landscape. The floodplain in
zone D has begun to fill in at the 5-year flood.

Figure 4.5. 5-year inundation areas pre- and post-project

47

4.2.4 Inundation Area – 10-Year
Occurrences of the 10-year flood have been observed to have increased in the last
10 years (USGS, 2018). The inundation area for the 10-year flow, shown in Figure 4.6,
increased 2.4 acres, from 26.9 to 29.3 acres. This represents a 9.0% increase in wetted
area. All zones are observed to be spreading out into the floodplain. The most notable
difference between pre- and post-project conditions, not already noted, is a more defined
connection within the historic channel in zone C.

Figure 4.6. 10-year inundation areas pre- and post-project

48

4.2.5 Inundation Area – 25-Year
The inundation area for the 25-year flow, shown in Figure 4.7, increased 1.5
acres, from 33.4 to 34.9 acres. This represents a 4.5% increase in wetted area. Zone A
shows the beginning of the connection of an upper side channel on the post-project
terrain. In Zone B, the pre-project, wall-based side channel would now be activated
through small openings in Forest Road 70 made in a previous, smaller scale restoration
effort that removed sections of the road where culverts had previously been. Before the
Greenwater River Floodplain Restoration Project, this was the only method for water to
reach that section of floodplain. Pre-project flow patterns are now spilling into the
historic channel in zone C. Post-project conditions are resulting in more flow spilling out
of the channel onto the floodplain across zones B, C, and D.

Figure 4.7. 25-year inundation areas pre- and post-project

49

4.2.6 Inundation Area – 50-Year
The inundation area for the 50-year flow, shown in Figure 4.8, increased 2.4 acres
from 39.1 to 41.5 acres. This represents a 2.4% increase in wetted area. Zone A shows
the connection of the upper side channel. More activation around the side channels of the
southern floodplain is seen in both zones A and B, post-project. Zone C and D both have
similar floodplain activation at the 50-year flood pre- and post-project.

Figure 4.8. 50-year inundation areas pre- and post-project

50

4.2.7 Inundation Area – 100-Year
The inundation area for the 100-year flow, shown in Figure 4.9, increased 3.6
acres from 44.1 to 47.7 acres. This represents an 8.2% increase in wetted area. Zone A
and B again display more of the southern floodplain activation post-project. Zone C and
D show similar floodplain activation at the 100-year flood pre- and post-project.

Figure 4.9. 100-year inundation areas pre- and post-project

51

4.3 Flow Velocity
Flow velocities were analyzed for the 10 and 100-year events. The 10-year event
represents a relatively frequent flooding event. The 100-year event represents extreme
flooding. Flow velocities were analyzed only within the Greenwater River floodplain and
do not include the 2 tributaries’ inflow areas. Velocities were compared both across the
entire wetted area as well as within the theoretical spawning channel, represented by the
mean annual minimum inundation area, in an effort to explore the possible effect of more
frequent flood events on salmon redds.

4.3.1 Flow Velocities – 10-Year
The 10-year flow velocities, shown in Figure 4.10, modeled on the post-project
landscape, show a decrease in velocities by 1.57% across the entire channel and
floodplain, while a 9.16% decrease is modeled within the spawning channel. See Table
4.1. In a Welch’s two sample t-test, the difference between the means is significantly
different, p < 0.001, for both areas.

Table 4.1. 10-year velocities pre- to post-project

52

Figure 4.10. 10-year velocities pre- and post-project
53

4.3.2 Flow Velocities – 100-Year
The 100-year flow velocities, shown in Figure 4.11, modeled on the post-project
landscape, show a decrease in velocities by 3.83% across the entire channel and
floodplain, while an 8.46% decrease is modeled within the spawning channel. See Table
4.2. In a Welch’s two sample t-test, the difference between the means is significantly
different, p < 0.001, for both areas. The visualization of flow velocities shows more
continuous high velocity flows in the pre-project channel, while in post-project
conditions velocities are seen to occur in sequence broken up by lower velocity regions
which would give rearing fish some refuge from higher velocities, and theoretically help
protect salmon redds in the area from scour.

Table 4.2. 100-year velocities pre- to post-project

54

Figure 4.11. 100-year velocities pre- and post-project
55

4.4 DTM Resolution Effect on Hydraulic Modeling
Digital terrain data for the 2017 post-project conditions were downsampled from
a 1-foot to a 3-foot raster grid, and hydraulic modeling was then performed on the
downsampled 3-foot gridded terrain, to determine any difference in modeled outputs at
this magnitude of difference. Very small variations were observed in the spatial patterns
of the modeled inundation areas for all events. Similar to the other event results, the 100year flood inundation area, seen in Figure 4.12, displays almost the entire inundation area
modeled on the 1-foot and 3-foot terrain as overlapping, with only small areas of
difference peeking out in the off-channel floodplain areas, resulting in a 1.34% change in
area. A comparison between inundation areas for all events is presented in Table 4.3.

Figure 4.12. 100-year inundation areas on post-project 1-foot and 3-foot gridded terrain

56

Table 4.3. Inundation areas on post-project 1-foot and 3-foot gridded terrain

All flow events show a decrease in inundation area on the downsampled 3-foot
landscape. The area decrease was similar across all events, ranging from 0.57 acres for
the 1.6-year flood to 0.91 acres for the 50-year flood, with a standard deviation of 0.09
acres. The percent area decrease ranged from 1.28% for the 100-year flood to 6.13% for
the mean annual minimum and the mean annual flows, consistently decreasing as
discharge and inundation area increases. Even though the percent decrease was larger for
the lower flow events the spatial patterns remain very similar, as seen in the inundation
areas for the mean annual minimum event, shown in Figure 4.13

57

Figure 4.13. Mean annual minimum inundation areas on post-project 1-foot and 3-foot
gridded terrain
A comparison between velocities modeled for the 10 and 100-year events on the
1-foot and 3-foot terrain is presented in Table 4.4. Results show no greater than a 0.5%
change in mean velocities, with a slightly negative change observed over the entire
floodplain, and a slightly positive change seen within the spawning channel.
Table 4.4. Mean velocities on post-project 1-foot and 3-foot gridded terrain

58

5.0 DISCUSION
“all models are wrong, but some are useful”
-George E. P. Box, 1976

5.1 Introduction
Results of the hydraulic modeling show that the Greenwater River Floodplain
Restoration Project had a positive effect on floodplain reconnection and reduction of
velocities within the channel across all flow events modeled. Combined with other
floodplain enhancement projects in the area, the Greenwater River project should add
resiliency to this watershed, increasing available ecologically functioning habitat for
salmon while reducing flooding risk for downstream communities. As climate changes
lead to shifting flow patterns, the need for an understanding of and continued effort
towards floodplain restoration becomes ever more important.
Results from the analysis of terrain resolution effect on the hydraulic model, at the
reach scale, show very little change in inundation and velocity outputs at the studied
resolution difference. This provides evidence that UAV lidar should not be selected for
project monitoring based on the hope for more accurate hydraulic model outputs alone.
There are, however, other advantages to this type of acquisition, such as cost savings for
smaller project areas, and the resulting higher resolution output has potential for other
analysis, especially if blue-green bathymetric lidar can be acquired.
This chapter first discusses the results of the modeled inundation and velocity
comparisons. The ecological importance of floodplain restoration as it relates to trending
changes in flood occurrences in the Greenwater Basin is then analyzed. Next, the

59

comparison between modeled results on different terrain resolutions is discussed, a lidar
cost comparison is presented, and finally recommendations for future research are made.

5.2 Pre- to Post-Project Comparisons
Results from this study were analyzed on their own as well as being compared to
those from the pre-project HECI, 2010 assessment. Modeled results from this study were
analyzed for numerical area and velocity differences pre- to post-project along with
spatial flow pattern shifts resulting from changes to the morphology of the river and
floodplain. When comparing models run for this study to the 2010 HECI assessment,
three significant differences should be noted. One, the proposed project area at the time
of the HECI assessment extended approximately an additional 1,700 feet downstream of
the current project reach and lidar acquisition. Two, the HECI model input all of the
calculated flow at the bottom of the project area to the inflow at the upper end of the
project, while this research modeled the contributing tributary flow inputs separately
rather than adding their values to the Greenwater River inflow at the top of the project.
Three, the HECI model utilized peak flood levels calculated in 1996 and reported in
Abbe et al. (2007), while this research used current peak flood levels reported through the
USGS, that were recalculated in 2014. See Table 3.1. Results from HECI, 2010 were
primarily presented graphically in the form of maps, and only the 100-year flood
inundation area values were reported on. Taking into account the differences between the
models, specific values would not be directly comparable. Instead, only the normalized
percent increases of the 100-year event, observed floodplain and side channel activation,
and velocity observations are compared between the HECI, 2010 assessment and this
study.
60

5.2.1 Inundation Area
The removal of Forest Road 70 combined with the addition of engineered log
jams throughout the project area had a notably positive effect on overall floodplain
inundation area and activation of side channel habitat. Results from this study showed an
8.2% increased area for the 100-year event, with an average gain of 14.3% across all
events, and a 9% gain for events above bankfull flow. These results aligned with the
reported 10% increase in floodplain inundation for the 100-year event (HECI, 2010). An
inspection of spatial changes to flow also revealed similar changes on the post-project
terrain. Although the 2010 HECI model of the 100-year proposed conditions did show
more filling in of the floodplain in zones A and B, this is likely due to higher modeled
discharge inflows at the top of the project area from both a greater 100-year peak flood
value and the allocation of tributary flow to the upper project area.
The most notable change to post-project inundation in this study is the activation
of the wall-based side channel occurring for all events, including the mean annual
minimum flow and above, which added approximately 1,500 feet of side channel habitat
in zone B. This side channel did not activate on the pre-project landscape until flows
exceeded the 10-year event. As expected, higher flow events also showed much more
utilization of the upper floodplain in zones A and B due to removal of FR 70. Log jams
installed throughout Zone C added much variation to the channel in that area, increasing
potential spawning and rearing habitat throughout the main channel in that zone. As well,
inspection of inundation in zone D reveals a more complex main channel due to local
anabranches likely caused by the deposition of sediment leading to aggradation.

61

It should be noted that inundation areas reported for this research, as well as those
in the HECI, 2010 assessment, are all slightly exaggerated for the specific discharge
values used. This is due to the lack of bathymetry accurately depicting the topography of
the channel below the water surface. Since NIR lidar does not penetrate the water surface
the lidar point cloud shows no data wherever water is covering the terrain (see Appendix
A & B) and the resulting DTM simply shows a flat surface between the lowest recorded
points on opposite banks. This is why it was important to capture the NIR lidar during
seasons of low flow, when as much of the channel as possible is visible. Discharge values
used in this research represent theoretical events, which have been noted to change
periodically as more gage data are collected, so a slight exaggeration of inundation is not
as vital across flows modeled on the same terrain. However, this could have a slight
effect on comparisons between pre- and post-project models made for this research. The
2007 Lidar was captured during a slightly higher discharge, which could skew the 2007
inundation areas to be slightly larger due to the displacement of more area under the
water surface. Taking this into account would only increase the change in inundation
areas, hence revealing an even more positive value for inundation area gained.

5.2.2 Flow Velocities
Modeled mean flow velocities for this study were observed to decrease for all
events on the post-project 2017 terrain. Results were presented for the 10 and 100-year
events. No specific values were reported for comparison in the HECI (2010) assessment,
but it was stated that “velocities in the main channel significantly decrease” on the
proposed project landscape, which could lead to positive post-restoration responses such
as sediment deposition. Current results show that mean velocities across the entire
62

floodplain showed a reduction of 1.57% and 3.83% respectively. Outside of the channel,
velocities are seen to be much lower than within the channel. Because there is a larger
area of inundation outside of the channel than within, the difference in mean velocities
across the entire inundation area are skewed to a lower value. In order to compare the
changes in the higher velocities within the channel, where reductions would likely be
more pronounced due to ELJ placements and morphology changes, and there would be
the most effect on incubating redds, velocities were also compared for the 10 and 100year event in the pre- and post-project mean annual minimum inundation area. An
inspection of the monthly mean gage discharges (1929–2017) during the Spring Chinook
and Coho Salmon spawning months of September through November yielded an average
mean discharge of 102 cfs, σ = 58. Hence, the mean annual minimum event of 92 cfs at
the gage was designated as an acceptable representation of the spawning channel for this
research. Mean velocities in this area were observed to decrease 9.16% and 8.46%
respectively. Velocity reduction was calculated to be significantly different for these
events, p < 0.001. The ecological significance of the observed decrease in flow velocities
in the mean annual minimum channel, designated as the spawning channel for this
analysis, is discussed in the next section.

5.3 Ecological Importance
The importance of understanding the hydrology of the Greenwater River is made
more apparent when analyzing changes in discharge trends over the last 10 years and how
it relates to fish use. This section first outlines the different salmon species and their
potential use of the project reach. Trends in peak flood occurrences are then examined.

63

Finally, timing of fish use is compared with flood events to highlight possible effects of
flooding events on fish populations.

5.3.1 Salmon Populations
The Greenwater River is reported to support populations of Spring Chinook (O.
tshawytscha), Coho Salmon (O. kisutch), and winter steelhead (O. mykiss) (Ecology,
1998; Marks et al., 2016). Chinook stocks in Puget Sound were federally listed as
endangered in 1999, and are currently listed as threatened along with Puget Sound
steelhead. Coho Salmon are common throughout the Puyallup/White River Watershed
(Marks et al., 2016.). All three of these species have various life histories throughout the
Greenwater River system.
Spring Chinook return to the freshwater as early as May and typically spawn in
September through October (Ecology, 1998; Marks et al., 2016; WDFW et al.,1996). Egg
to fry emergence occurs 90–110 days later in February through March (Ecology, 1998;
Marks et al., 2016; Smith & Wampler, 1995). Most (80%) of juvenile Spring Chinook
migrate as sub-yearlings out into salt water (Marks et al., 2016).
Coho Salmon enter the river system in early August with peak spawning
occurring in October and November (Marks et al., 2016). These fish generally rear in the
system for over a year (18 months) before entering marine waters as yearlings.
Winter steelhead, an anadromous form of rainbow trout, in the Greenwater River
system generally return in November and December with peak spawning occurring in
April and May (Marks et al., 2016). Egg-to-fry emergence of winter steelhead occurs 28–
56 days later, depending on water temperature, and fish will rear in the freshwater river

64

system for 1–4 years before migrating out to salt water in the spring (Marks et al., 2016).
See figure 5.1 for StreamNet fish distributions by species, within the project reach.

Figure 5.1. StreamNet fish distributions by species. (StreamNet GIS DATA, 2013)
65

5.3.2 Changes in Peak Flood Occurrence
Discharge has been regularly collected for the Greenwater River gage, no.
12097500, since 1929. Upon inspection of the USGS gage data it is observed that there
has been an increase in the past decade of the number of flood events greater than the
estimated 10-year event. See Figure 5.2 and Appendix C. The 10-year flood event is the
peak flood level that has a 10% probability of occurring each year, with an average
recurrence interval of 10 years. Within the past 10 years, there were 5 years with flood
events that equaled or exceeded 10-year flood levels, with 2 years seeing multiple events
greater than the 10-year flood. Prior to 2008, the average reoccurrence of the 10-year
flood was once per decade, as would generally be expected. As all of these flood events
occur in the winter from November to February, this is likely due to a shifting climate in
which warmer temperatures are transitioning winter snowfall into rain. When rain falls on
existing snowpack, both the rain and meltwaters flow downstream, amplifying the
flooding effect. Alternately, as less snowpack is left in the upper watershed, decreased
summer flows generally result, which can then lead to increased water temperatures
during the summer months. How this relates to fish populations is discussed in the next
section.

66

Figure 5.2. Annual peak streamflow 1912–2017 (top), Streamflow 2008–2017 (middle),
Annual peak flow events > 10-year flood per decade (bottom)

67

5.3.3 Ecological Interaction
The interaction between Greenwater River salmon stocks and an increase in peak
flood occurrences is visualized in Figure 5.3. Past studies have shown that decreased eggto-fry survival rates and smolt production have been correlated with larger flood events
(Beamer & Pess, 1999). Higher flows increase risk of bed scour on incubating salmon
redds and can prematurely flush rearing fish out of the system.
With significant flooding events becoming more prevalent in last 10 years,
particularly during Spring Chinook and Coho Salmon egg incubation months, there is a
greater risk for bed scour to destroy these redds. This shift in discharge patterns is the
predicted result of a changing climate transforming primarily snowmelt fed basins into
transient, rain/snowmelt fed, basins, and transient basins into rainfall dominant basins
(Mantua et al., 2010). Less summer snowmelt, leading to lower summer flows and
increased temperatures, also increases risk to winter steelhead redds and all rearing
salmon species.
In the face of a changing climate it becomes ever more important and beneficial to
restore as many natural floodplains as possible to naturally increase flood storage and
ease the effect of high-flow events. Restoring vegetation along riparian zones can provide
needed shade, habitat complexity and eventually contribute to wood recruitment (Mantua
et al., 2010). Engineered log jam placements in these projects can lead to positive wood
and sediment recruitment and create more pools and protected habitat for salmon species
(Cramer et al., 2012). The Greenwater River Floodplain Restoration Project adds
increased resilience to the system by effectively increasing floodplain and side channel
habitat while reducing flow velocities in the main channel.

68

Figure 5.3. Relationship between peak flow events and salmon life histories in the
Greenwater River

5.4 DTM Resolution Effect
Lidar data for post-project conditions was captured using UAV-mounted lidar,
with the potential to achieve 1 cm precision with 1.5 cm accuracy. Because of heavy
canopy and vegetation cover in the project area limiting the density of lidar ground
returns, a 1-foot raster DTM was yielded from the bare-earth point cloud. The initial
intent of this part of the research, comparing results yielded from different resolutions of
the post-project data, was to examine the advantages of higher-resolution lidar data, but
69

this research also provided insight into the level of expected error when comparing the
pre- and post-project data, with the same resolution difference.
Resolution downsampling of the 2017 terrain resulted in a relatively small change
in modeled inundation area and flow velocities, with consistently smaller inundation
areas modeled on the 3-foot terrain. Most of the variation between modeled inundation
areas on the native and downsampled post-project terrains is observed along the banks
and edges of the wetted area, with an increase in variation across the floodplain, in
relatively flat areas of shallow water, where the averaging of sub-grid high and low
ground features in the downsampled terrain is likely blocking shallow water flow into
these areas. An increase in the percent variation is seen for the lower flow events. This is
likely due to a higher ratio of effected perimeter to area of the inundation shape.
Modeled velocity comparisons for the 10 and 100-year floods reveal even less
change than inundation area, showing a positive change across the whole floodplain, and
a negative change within the spawning channel area. The absolute values were very
small, with no change observed greater than 0.5 percent, and an average change of 0.32
percent.
While this method of downsampling the DEM is the same method used in
research done by Charrier & Li (2012), it is somewhat flawed when used to output a
relatively small resolution change. The downsampled DTM is still based on the same
underlying higher density point cloud and the magnitude of resolution change is evidently
not large enough to show a change in model results based on the averaged terrain. A
more notable difference may occur if the 3-foot DTM was based on a randomly
downsampled point cloud, or a separate lidar acquisition all together. Although the output

70

results do not warrant the higher resolution data used in this study on model output gains
alone, the small change observed represents the likely small amount of error when
comparing the pre- and post-project terrain data, with the same resolution difference.
The higher resolution data still have notable advantages. If captured with
bathymetry data, high-resolution data could be much more useful in identifying subthree-foot grid habitat features within the channel. The higher resolution data could also
provide a more accurate representation of vegetation and other roughness factors along
and above the terrain surface when analyzing these fine-scale ecological features.

5.5 Cost Benefit
A comparison of the cost difference between lidar captured by UAV or manned
aircraft was also done. It can be said that for smaller project areas of only a few hundred
acres the UAV lidar is less costly, although bathymetry was not available through the
vender used. Larger project areas captured with a UAV would be more difficult, require
multiple battery changes and ground control locations. For larger project areas, a
conventional manned aircraft lidar flight would likely be more cost effective. See Table
5.1. It should be noted that the resolution of the bare-earth digital elevation model
delivered from the manned aircraft lidar flight was still quoted at 3-foot resolution
compared to the 1-foot resolution captured in the UAV flight.

71

Table 5.1. Lidar cost comparison
UAV Lidar - 1-foot resolution DEM
LiDAR Acquisition $8,750 + Correction Points Survey $4,227 = $12,977 Total
Manned Aircraft Lidar Flight (quoted for the project) – 3-foot resolution DEM
Green Water River, WA
AOI 1 – NIR
AOI 1 - Topobathy
AOI 2 - NIR

Area (Acres)
1,473
1,473
146

Total Cost
$24,550
$31,430
$23,110

5.6 Future Research
The results from this research, while they are not an exact representation of the
hydrology and habitat patterns in the Greenwater River, provide a level of accuracy
useful in monitoring the successes of this restoration. Typically there are methods that
could improve the models and metrics used to analyze restoration efforts. Below are a
few points that could further both the precision and accuracy of the predictions we make
on future projects.
Roughness values, also referred to as Manning’s n values in this research, are an
important part of the equations that drive the hydraulic model. Currently HEC-RAS only
assigns a single roughness value to each computational grid cell face, although future
versions are expected to allow for sub-grid subtlety in roughness to be expressed, similar
to how sub-grid terrain is currently taken into account (Brunner, 2016). This would
warrant more precise roughness determinations. The method presented by Casas et al.
(2010) could then be used to more precisely parameterize roughness using the lidar data.
By capturing bathymetry data within the high-resolution terrain model a more
detailed representation of instream habitat could likely be modeled and classified. This

72

could provide a more remote method of analysis when sending field crews in to collect
habitat data is too difficult or costly.
Results from this study contribute to the body of knowledge around the
effectiveness of floodplain restoration. To prioritize the best type and location of future
restoration sites it would be recommended that future studies look into other river
systems for similar or changing discharge trends or separation of vital functioning
habitat.

73

6.0 CONCLUSION
Two-dimensional hydraulic modeling, utilizing HEC-RAS 5.0.3, was used in this
study to evaluate the effectiveness of the Greenwater River Floodplain Restoration
Project, and to provide insight into the effects and uses of higher resolution lidar-based
terrain data, acquired with a UAV. Modeled results presented an increase in floodplain
inundation and reduction of velocities during high flow events, providing evidence to the
success of this restoration in meeting proposed restoration goals, and providing increased
flood protection within the system. Collecting lidar data for this study using a UAV
provided a more detailed cost-effective product than using the manned aircraft option.
Although the comparison of results modeled on different resolutions of the post-project
terrain yielded only small variations in spatial patterns, it is recommended that the
potential uses of the higher resolution data is explored in future research, which includes
the capture of bathymetric data. Results from this and other restoration monitoring
projects provide a source of insight into the effectiveness of similar future projects, so
that management practices can be best adapted to increase the gains of restoration.
The Greenwater River Floodplain Restoration Project was largely focused on the
restoration of salmon habitat. The need for this and future restoration has become more
apparent when examining the hydrology of the Greenwater River in the past ten years
compared to the previous eight decades of gage flow data. In particular, it is observed
that there has been an increase in 10-year peak flood occurrences in the last ten years of
over 5 times the expected and historic occurrence. Shifts in the global climate, that are
seen to be occurring now, have resulted in higher and more frequent flooding during

74

months of salmon egg incubation and rearing, and pose a risk to populations of Pacific
salmon.
Continued efforts to restore and conserve rivers and streams are vital in insuring
the health and survival of not only the salmon, but the rest of the ecosystem that they are
an integral part of. These types of projects add both resilience and flood protection,
benefiting the riverine ecosystem as well as the people who live near them. It is
imperative that we continue to conduct and support the monitoring and study of these
projects so that we have the information to make the most informed restoration and
management decisions into the future.

75

REFERENCES
Abbe, T., Beason, S., & Bunn, J. (2007). Geomorphic Basis of Conceptual Design
Greenwater River Restoration Project. Prepared for South Puget Sound Salmon
Enhancement Group.
Battin, J., Wiley, M. W., Ruckelshaus, M. H., Palmer, R. N., Korb, E., Bartz, K. K., &
Imaki, H. (2007). Projected Impacts of Climate Change on Salmon Habitat
Restoration. Proceedings of the National Academy of Sciences of the United States
of America, 104(16), 6720–6725.
Beamer, E. M., & Pess, G. R. (1999). Effects of peak flows on Chinook Oncorhynchus
tshawytscha spawning success in two Puget Sound river basins. In Watershed
Management to Protect Declining Species. American Water Resources Association.
Beechie, T. J., Liermann, M., Pollock, M. M., Baker, S., & Davies, J. (2006). Channel
pattern and river-floodplain dynamics in forested mountain river systems.
Geomorphology, 78(1), 124–141.
Bernhardt, E. S., Palmer, M. A., Allan, J. D., Alexander, G., Barnas, K., Brooks, S., …
Follstad-Shah, J. (2005). Synthesizing US river restoration efforts. American
Association for the Advancement of Science.
Bisson, P. A., Quinn, T. P., Reeves, G. H., & Gregory, S. V. (1992). Best management
practices, cumulative effects, and long-term trends in fish abundance in Pacific
Northwest river systems. In Watershed management (pp. 189–232). Springer.
Retrieved from http://link.springer.com/chapter/10.1007/978-1-4612-4382-3_7
Box, G. E. P. (1976). Science and statistics. Journal of the American Statistical
Association, 71(356), 791–799.
Brakensiek, K. (2017). Greenwater River, Washington, Restoration and Monitoring -Survey Year 2016 Update -- Summary Report. Prepared for South Puget Sound
Salmon Enhancement Group.
Brunner, G. W. (2016). HEC-RAS River Analysis System, 2D Modeling User’s Manual
Version 5.0. US Army Corps of Engineers Institute for Water Resources Hydrologic
Engineering Center.
Casas, A., Lane, S. N., Yu, D., & Benito, G. (2010). A method for parameterising
roughness and topographic sub-grid scale effects in hydraulic modelling from
LiDAR data.
76

Cavalli, M., Tarolli, P., Marchi, L., & Dalla Fontana, G. (2008). The effectiveness of
airborne LiDAR data in the recognition of channel-bed morphology. Catena, 73(3),
249–260.
Charrier, R., & Li, Y. (2012). Assessing resolution and source effects of digital elevation
models on automated floodplain delineation: a case study from the Camp Creek
Watershed, Missouri. Applied Geography, 34, 38–46.
Choi, Y. D. (2007). Restoration ecology to the future: a call for new paradigm.
Restoration Ecology, 15(2), 351–353.
Costabile, P., Macchione, F., Natale, L., & Petaccia, G. (2015). Flood mapping using
LIDAR DEM. Limitations of the 1-D modeling highlighted by the 2-D approach.
Natural Hazards, 77(1), 181–204.
Chow, V. (1959). Open-channel hydraulics (Vol. 1). McGraw-Hill New York.
Cramer, Michelle L. (managing editor). (2012). Stream Habitat Restoration Guidelines.
Co-published by the Washington Departments of Fish and Wildlife, Natural
Resources, Transportation and Ecology, Washington State Recreation and
Conservation Office, Puget Sound Partnership, and the U.S. Fish and Wildlife
Service. Olympia, Washington.
Crowder, D. W., & Diplas, P. (2000). Using two-dimensional hydrodynamic models at
scales of ecological importance. Journal of Hydrology, 230(3–4), 172–191.
Dudley, S. J., Fischenich, J. C., & Abt, S. R. (1998). EFFECT OF WOODY DEBRIS
ENTRAPMENT ON FLOW RESISTANCE 1. JAWRA Journal of the American
Water Resources Association, 34(5), 1189–1197.
Golshan, M., Jahanshahi, A., & Afzali, A. (2016). Flood hazard zoning using HEC-RAS
in GIS environment and impact of manning roughness coefficient changes on flood
zones in Semi-arid climate. Desert, 21(1), 24–34.
Gresh, T., Lichatowich, J., & Schoonmaker, P. (2000). An Estimation of Historic and
Current Levels of Salmon Production in the Northeast Pacific Ecosystem: Evidence
of a Nutrient Deficit in the Freshwater Systems of the Pacific Northwest. Fisheries,
25(1), 15–21. https://doi.org/10.1577/15488446(2000)025<0015:AEOHAC>2.0.CO;2
Herrera Environmental Consultants, Inc. (2010). HYDRAULIC ASSESSMENT OF
RESTORATION ALTERNATIVES, Greenwater River Engineered Logjam Project.
Prepared for South Puget Sound Salmon Enhancement Group.
77

Jeffres, C. A., Opperman, J. J., & Moyle, P. B. (2008). Ephemeral floodplain habitats
provide best growth conditions for juvenile Chinook salmon in a California river.
Environmental Biology of Fishes, 83(4), 449–458.
Khattak, M., Anwar, F., Sheraz, K., Saeed, T., Sharif, M., & Ahmed, A. (2016).
Floodplain Mapping Using HEC-RAS and ArcGIS: A Case Study of Kabul River.
Arabian Journal for Science & Engineering (Springer Science & Business Media
BV), 41(4).
Laurie, G.J., (2002), A Natural Channel Design to Restore the Greenwater River: Mt.
Baker-Snoqualmie National Forest: U.S.D.A. Forest Service, 69 p.
Mann, M. E., Zhang, Z., Rutherford, S., Bradley, R. S., Hughes, M. K., Shindell, D., …
Ni, F. (2009). Global Signatures and Dynamical Origins of the Little Ice Age and
Medieval Climate Anomaly. Science, 326(5957), 1256–1260.
https://doi.org/10.1126/science.1177303
Mantua, N., Tohver, I., & Hamlet, A. (2010). Climate change impacts on streamflow
extremes and summertime stream temperature and their possible consequences for
freshwater salmon habitat in Washington State. Climatic Change, 102(1–2), 187–
223. https://doi.org/10.1007/s10584-010-9845-2
Marks, E. L., R.C. Ladley, B.E. Smith, A.G. Berger, J.A. Paul, T.G. Sebastian and K.
Williamson. (2016). 2015-2016 Annual Salmon, Steelhead, and Bull Trout Report:
Puyallup/White River Watershed--Water Resource Inventory Area 10. Puyallup
Tribal Fisheries, Puyallup, WA.
Mason, D. C., Cobby, D. M., Horritt, M. S., & Bates, P. D. (2003). Floodplain friction
parameterization in two-dimensional river flood models using vegetation heights
derived from airborne scanning laser altimetry. Hydrological Processes, 17(9),
1711–1732.
McKean, J., Isaak, D., & Wright, W. (2009). Stream and riparian habitat analysis and
monitoring with a high-resolution terrestrial-aquatic LiDAR.
O’Neal, J. S., Roni, P., Crawford, B., Ritchie, A., & Shelly, A. (2016). Comparing Stream
Restoration Project Effectiveness Using a Programmatic Evaluation of Salmonid
Habitat and Fish Response. North American Journal of Fisheries Management,
36(3), 681–703. https://doi.org/10.1080/02755947.2016.1165773
Quang Minh, N., & La, H. (2011). Land cover classification using LiDAR intensity data
and neural network. Korean Journal of Geomatics, 29, 429–438.
https://doi.org/10.7848/ksgpc.2011.29.4.429
78

Quiroga, V. M., Kure, S., Udo, K., & Mano, A. (2016). Application of 2D numerical
simulation for the analysis of the February 2014 Bolivian Amazonia flood:
Application of the new HEC-RAS version 5. RIBAGUA-Revista Iberoamericana Del
Agua, 3(1), 25–33.
RCO. (2009, July). Salmon Recovery - Species Listed Under the Federal Endangered
Species Act - Washington State Recreation and Conservation Office. Retrieved
November 28, 2017, from
http://www.rco.wa.gov/salmon_recovery/listed_species.shtml
Reeves, G. H., Hohler, D. B., Hansen, B. E., Everest, F. H., Sedell, J. R., Hickman, T. L.,
& Shively, D. (1997). Fish habitat restoration in the Pacific Northwest: fish creek of
Oregon. Watershed Restoration: Principles and Practices., 335–359.
Schindler, D. E., & Rogers, L. A. (2009). Responses of Pacific salmon populations to
climate variation in freshwater ecosystems. In American Fisheries Society
Symposium (Vol. 70, pp. 1127–1142).
Smith, C. and P. Wampler, eds. (1995). Dungeness River chinook salmon rebuilding
project progress report 1992-1993. Northwest Fishery Resource Bulletin No. 3. pp 172.
Stewart, G. B., Bayliss, H. R., Showler, D. A., Sutherland, W. J., & Pullin, A. S. (2009).
Effectiveness of engineered in-stream structure mitigation measures to increase
salmonid abundance: a systematic review. Ecological Applications, 19(4), 931–941.
StreamNet GIS DATA. (2013). StreamNet – Fish Distribution by Species. Retrieved
from https://www.streamnet.org/data/interactive-maps-and-gis-data/
Suding, K. N. (2011). Toward an era of restoration in ecology: successes, failures, and
opportunities ahead. Annual Review of Ecology, Evolution, and Systematics, 42,
465–487.
Thorpe, A. S., & Stanley, A. G. (2011). Determining appropriate goals for restoration of
imperilled communities and species. Journal of Applied Ecology, 48(2), 275–279.
USGS Current Conditions for USGS 12097500 GREENWATER RIVER AT
GREENWATER, WA. (2018). Retrieved from
https://waterdata.usgs.gov/usa/nwis/uv?12097500
Waples, R. S., Pess, G. R., & Beechie, T. (2008). Evolutionary history of Pacific salmon
in dynamic environments. Evolutionary Applications, 1(2), 189–206.

79

Washington Department of Fish and Wildlife, Puyallup Indian Tribe and Muckleshoot
Indian Tribe. (1996). Recovery plan for White River spring chinook. Washington
Department of Fish and Wildlife. Olympia, WA 81p.
Washington State Department of Ecology, Water Quality Program. (1998). White River
Spring Chinook Habitat Guidance, A Water Quality Management Approach for the
Upper White River.
Watershed Sciences (2007). LiDAR Remote Sensing Data Collection: SR 410 &
Greenwater River: Point, Raster and Vector Data and Data Report submitted to
WSDOT Photogrammetry, Herrera Environmental Consultants and the South Puget
Sound Salmon Enhancement Group, 26 p.
Whiteway, S. L., Biron, P. M., Zimmermann, A., Venter, O., & Grant, J. W. A. (2010).
Do in-stream restoration structures enhance salmonid abundance? A meta-analysis.
Canadian Journal of Fisheries and Aquatic Sciences, 67(5), 831–841.
Yang, J., Townsend, R. D., & Daneshfar, B. (2006). Applying the HEC-RAS model and
GIS techniques in river network floodplain delineation. Canadian Journal of Civil
Engineering, 33(1), 19–28.

80

APPENDICES
Appendix A. 2017 three-dimensional lidar point cloud of bare-earth laser returns
Top: oblique view. Bottom: top down perspective

Low

High

Point Elevation

81

Appendix B. 2017 three-dimensional lidar point cloud of all laser returns
Top: oblique view. Bottom: top down perspective

Low

High
Point Elevation

82

Appendix C. Greenwater gage discharge 2008-2017

83

Media

Thesis_MES_2018_ZierdtB.pdf