EFFECTIVENESS OF EXTRACTING WATER SURFACE SLOPES FROM LIDAR
DATA WITHIN THE ACTIVE CHANNEL: SANDY RIVER, OREGON, USA
JOHN THOMAS ENGLISH
Presented to the Department of Geography
and the Graduate School of the University of Oregon
in partial fulfillment of the requirements
for the degree of
Master of Science
"Effectiveness of Extracting Water Surface Slopes from LiDAR Data within the Active
Channel: Sandy River, Oregon, USA," a thesis prepared by John Thomas English in
partial fulfillment of the requirements for the Master of Science degree in the Department
of Geography. This thesis has been approved and accepted by:
Committee in Charge: W. Andrew Marcus, Chair
Patricia F. McDowell
Dean of the Graduate School
© 2009 John Thomas English
An Abstract of the Thesis of
John Thomas English
in the Department of Geography
for the degree of
to be taken
Master of Science
Title: EFFECTIVENESS OF EXTRACTING WATER SURFACE SLOPES FROM
LIDAR DATA WITHIN THE ACTIVE CHANNEL: SANDY RIVER, OREGON,
W. Andrew Marcus
This paper examines the capability ofLiDAR data to accurately map river water
surface slopes in three reaches of the Sandy River, Oregon, USA. LiDAR data were
compared with field measurements to evaluate accuracies and determine how water
surface roughness and point density affect LiDAR measurements. Results show that
LiDAR derived water surface slopes were accurate to within 0.0047,0.0025, and 0.0014
slope, with adjusted R2 values of 0.35, 0.47, and 0.76 for horizontal intervals of 5, 10, and
20m, respectively. Additionally, results show LiDAR provides greater data density
where water surfaces are broken. This study provides conclusive evidence supporting
use ofLiDAR to measure water surface slopes of channels with accuracies similar to
field based approaches.
NAME OF AUTHOR: John Thomas English
PLACE OF BIRTH: Eugene, Oregon
DATE OF BIRTH: January 1st, 1980
GRADUATE AND UNDERGRADUATE SCHOOLS ATTENDED:
University of Oregon, Eugene, Oregon
Southern Oregon University, Ashland, Oregon
Master of Science, Geography, March 2009, University of Oregon
Bachelor of Science, Geography, 2001, Southern Oregon University
AREAS OF SPECIAL INTEREST:
LiDAR Database Coordinator, Oregon Department of Geology & Mineral
Industries, June 2008 - present.
LiDAR & Remote Sensing Specialist, Sky Research Inc., 2003 - 2008
GRANTS, AWARDS AND HONORS:
Gamma Theta Upsilon Geographic Society Member, 2006
Gradutate Teaching Fellowship, Social Science Instructional Laboratory, 20062007
I wish to express special thanks to Professors W.A. Marcus and Patricia McDowell
for their assistance in the preparation of this manuscript. In addition, special thanks are due
to Mr. Paul Blanton who assisted with field data collection for this project. I also thank the
members ofmy family who have been encouraging and supportive during the entirety of
my graduate schooling. I wish to thank my parents Thomas and Nancy English for always
being proud of me. Special thanks to my son Finn for always making me smile. Lastly,
special thanks to my wife Kathryn for her unwavering support, love, and encouragement.
Dedicated to my mother Bonita Claire English (1950-2004).
TABLE OF CONTENTS
I. INTRODUCTION 1
II. BACKGROlTND 5
Water Surface Slope 5
LiDAR Measurements of Active Channel Features 7
III. STUDY AREA 10
IV. METHODS 22
LiDAR Data and Image Acquisition 23
Field Data Acquisition 24
LiDAR Processing 25
Calculation of Water Surface Slopes 27
Evaluating LiDAR Slope Accuracies and Controls 33
V. RESULTS 35
Comparison of Absolute Elevations from Field and LiDAR Data in
Reach 1 35
Slope Comparisons 41
Surface Roughness Analysis 46
VI. DiSCUSSiON 51
VII. CONCLUSION 57
APPENDIX: ARCGIS VBA SCRIPT CODE 58
LIST OF FIGURES
1. Return Factor vs. LiDAR Scan Angle 2
2. Angle of Incidence 3
3. Wave Action Relationship to LiDAR Echo 3
4. Site Map 11
5. Annual Hydrograph of Sandy River 13
6. Oregon GAP Vegetation within Study Area 15
7. Photo of Himalayan Blackberry on Sandy River 16
8. Reach 1 Site Area Map with photo 18
9. Reach 2 Site Area Map 20
10. Reach 3 Site Area Map 21
11. LiDAR Point Filtering Processing Step 26
12. Field DEM Interpolated using Kriging 29
13. Reach 1 LiDAR Cross Sections and Sample Point Location 31
14. Differences Between LiDAR and Field Based Elevations 37
15. Regression ofLiDAR and Field Cross section Elevations 38
16. Comparison of LiDAR and Field Longitudinal Profiles
(5, 10,20 meters) 40
17. Regression ofField and LiDAR Based Slopes (5, 10,20 meters) 42
18. Differences Between LiDAR and Field Based Slopes (5, 10,20 meters) 44
19. Relationship of Water Surfaces to LiDAR Point Density 47
20. Marmot Dam: Orthophotographyand Colorized Slope Model 50
21. LiDAR Point Density versus Interpolation 53
LIST OF TABLES
1. Reported Accuracies of 2006 and 2007 LiDAR 24
2. Results of LiDAR and Field Elevation Comparison 38
3. Results ofLiDAR and Field Slope Comparison (5, 10,20 meters) 45
4. Results of Reach 1 Slope Comparison 46
5. Water Surface Roughness Results for Reach 1,2, and 3 48
6. Results of Reach 1 Water Surface Roughness Comparison 49
7. Subset of Reach 3 Water Surface Roughness Analysis Near
Marmot Dam 50
LiDAR (Light Detection and Ranging) has become a common tool for mapping
and documenting floodplain environments by supplying individual point elevations and
accurate Digital Terrain Models (DTM) (Bowen & Waltermire, 2002; Gilvear et aI.,
2004; Glenn et aI., 2005; Magid et aI., 2005; Thoma, 2005; Smith et aI., 2006;
Gangodagamage et aI., 2007). Active channel characteristics that have been extracted
using LiDAR include bank profiles, longitudinal profiles (Magid et aI., 2005; Cavalli et
aI., 2007) and transverse profiles of gullies under forest canopies (James et aI., 2007). To
date, however, no one has tested if LiDAR returns from water surfaces can be used to
measure local water surface slopes within the active channel.
Much of the reason that researchers have not attempted to measure water surface
slopes with LiDAR is because most LiDAR pulses are absorbed or not returned from the
water surface. However, where the angle of incidence is close to nadir (i.e. the LiDAR
pulse is fired near perpendicular to water surface plane), light is reflected and provides
elevations off the water surface (Figure 1, Maslov et aI., 2000). Where LiDAR pulses
glance the water surface at angles of incidence greater than 53 degrees, a LiDAR pulse is
more often lost to refraction (Figure 2) (Jenkins, 1957). In broken water surface
conditions the water surface plane is angled, which produces perpendicular angles of
incidence allowing for greater chance of return (Maslov et al. 2000). Su et al. (2007)
documented this concept by examining LiDAR returns off disturbed surfaces in a
controlled lab setting (Figure 3). LiDAR returns off the water surface potentially provide
accurate surface elevations that can be used to calculate surface slopes.
~ 0.6 o
2000 4000 60.00
sensing angle, degree
Figure 1. Return Factor vs. LiDAR Scan Angle. Figure shows relationship
between water surface return and scan angle. Return Factor versus sensing angle at
different levels of the waving d (d = scan angle). Figure shows the relationship of
scan angle of LiDAR to return from a water surface. Return factor is greatest at low
scan angles relative to the nadir region of scan. (Maslov, D. V. et. al. (2000). A
Shore-based LiDAR for Coastal Seawater Monitoring. Proceedings ofEARSeL-SIGWorkshop,
Figure 1, pg. 47).
reflected\\ :.;/ incident
1 I 1 .
'\ I lAIR
•••••••• ••••••••••••• •••••• ••••••••••••••••••••• • •• eo ••••••••••• o •••••••••••• _0 •••••••••• 0 •••
.•.•.•.•.•.•00 ,••••• ' 0•••• 0 ••••••••••• 0 ••I' .•.•.•.•.•.,................. .".0 ••••••••••••• , •••••••••••• , ••••••••••0••••.
.....................................~ . ••••••••••••••••••••••••••••••••••••• • •••••••••••••••••••••••••• 0 •••••••••••••••••••• 0 ••••• 0 ••
Figure 2. Angle of Incidence. Figure displays concept of reflection and refraction
of light according to angle of incidence. The intensity of light is greater as the angle
of incidence approaches nadir. (Jenkins, F.A., White, RE. "Fundamentals of
Optics". McGraw-Hili, 1957, Chapter 25)
-500 17.5 35 52.5 70
horizonral scanning dislancC(lllm)
Figure 3. Wave Action Relationship to LiDAR Echo. "LiDAR measurements of
wake profiles generated by propeller at 6000 rpm (a) and 8000 rpm (b). Su's work
definitively showed LiDAR's ability to measure water surfaces, and the relationship of
wave action to capability of echo. From Su (2007) figure 5, p.844 .
This study examines whether LiDAR can accurately measure water surface
elevations and slopes. In order to address this topic, I assess the vertical accuracy of
LiDAR and the effects of water surface roughness on LiDAR within the active channel.
Findings shed light on the utility of LiDAR for measuring water surface slopes in
different stream environments and methodological constraints to using LiDAR for this
Water Surface Slope
Water surface slope is a significant component to many equations for modeling
hydraulics, sediment transport, and fluvial geomorphic processes (Knighton, 1999, Sing
& Zang, in press). Traditional methods for measuring water surface slope include both
direct and indirect methods. Direct water surface slope measurements typically use a
device such as a total station or theodolite in combination with a stadia rod or drop line to
measure water surface elevations (Harrelson, et ai., 1994, Western et ai., 1997).
Inaccuracies in measurements stem from surface turbulence that makes it difficult to
precisely locate the water surface, especially in fast water where flows pile up against the
measuring device (Halwas, 2002). Direct survey methods often require a field team to
occupy several known points throughout a reach. This is a time consuming process,
especially if one wanted to document water surface slope along large portions of a river.
This method can be dangerous in deep or fast water.
Indirect methods of water surface slope measurement consist of acquiring
approximate water surface elevations using strand lines, water marks, secondary data
sources such as contours from topographic maps, or hydraulic modeling to back calculate
the water depth (USACE, 1993; Western et aI., 1997). Variable quality of data and
modeling errors can lead to inaccuracies using these methods. The use of strand lines and
water marks may not necessarily represent the peak flows or the water surface. Contours
may be calculated or interpolated from survey points taken outside the channel area. The
most commonly used hydraulic models are based on reconstruction of I-dimensional
flow within the channel and do not account for channel variability between cross section
LiDAR water surface returns have a great deal of promise for improving
measurement of water surfaces in several significant ways. LiDAR measurements
eliminate hazards associated with surveyors being in the water. LiDAR also captures an
immense amount of elevation data over a very short period of time, with hundreds of
thousands of pulses collected within a few seconds for a single swath. Within this mass of
pulses, hundreds or thousands of measurements off the water's surface may be collected
depending on the nature of surface roughness, with broken water surfaces increasing the
likelihood of measurements (Figure 3). In addition, most terrestrial LiDAR surveys
collect data by flying multiple overlapping flight lines, thus increasing the number of
returns in off nadir overlapping areas and the potential for returns from water surfaces.
The accuracy of high quality LiDAR measurements is comparable to field
techniques. The relative variability of quality LiDAR vertical measurements typically
ranges between 0.03-0.05 meters (Leica, 2007), where relative variability is the total
range of vertical error within an individual scan on surface of consistent elevation.
Lastly, LiDAR has the ability to collect water surface elevations over large stretches of
river within a single flight of a few hours.
LiDAR Measurements of Active Channel Features
Recent studies evaluating the utility of LiDAR in the active channel environment
have documented the effectiveness of using LiDAR DTMs to extract bank profiles.
Magid et al. (2005) examined long term changes of longitudinal profiles along the
Colorado River in the Grand Canyon. The study used historical survey data from 1923
and differenced topographic elevations with LiDAR data flown in 2000. LiDAR with
three meter spot spacing was used to estimate water surface profiles based on the LiDAR
elevations nearest to the known channel. Cavalli et al. (2007) extracted longitudinal
profiles of the exposed bed of the Rio Cordon, Italy using 0.5 meter LiDAR DEM cells.
This study successfully attributed LiDAR DEM roughness within the channel to instream
habitats. Bowen and Waltermire (2002) found that LiDAR elevations within the
floodplain were less accurate than advertised by vendors and sensor manufacturers.
Dense vegetation within the riparian area prevented LiDAR pulses from reaching the
ground surface resulting in accuracies ranging 1-2 meters. Accuracies within
unvegetated areas and flat surfaces met vendor specifications (l5-20cm).
James et al. (2007) used LiDAR at 3 meter spot spacing to map transverse profiles
of gullies under forest canopies. Results from this study showed that gully morphologies
were underestimated by LiDAR data, possibly due to low density point spacing and
biased filtering of the bare earth model. Today, point densities of 4-8 points/m2 are
common and would likely alleviate some of the troubles found in this study.
Additional studies have used LiDAR to extract geomorphic data from channel
areas. Schumann et al. (2008) compared a variety of remotely sensed elevation models
for floodplain mapping. The study used 2 meter LiDAR DEMs as topographic base data
for floodplain modeling, and found that modeled flood stages based on the LiDAR DEM
were accurate to within 0.35m. Ruesser and Bierman (2007) used high resolution LiDAR
data to calculate erosion fluxes between strath terraces based on elevation.
Gangodagamage et al. (2007) used LiDAR to extract river corridor width series, which
help to quantify processes involved in valley formation. This study used a fixed water
surface elevation and did not attempt to demonstrate the accuracy of LiDAR derived
Green LiDAR also has been used to examine riverine environments. Green
LiDAR functions much like terrestrial LiDAR (which uses an infrared laser) except that
green LiDAR systems use green light that has the ability to penetrate the water surface
and measure the elevation of the channel bed. Green LiDAR is far less common than
terrestrial LiDAR and the majority of studies have been centered on studies of ocean
shorelines. Wang and Philpot (2007) assessed attenuation parameters for measuring
bathymetry in near shore shallow water, concluding that quality bathymetric models can
be achieved through a number of post-processing steps. Hilldale and Raft (2007)
assessed the accuracy and precision of bathymetric LiDAR and concluded that although
the resulting models were informative, bathymetric LiDAR was less precise than
traditional survey methods. In general, it is often difficult to assess the accuracy of
bathymetric LiDAR given issues related to access of the channel bed at time of flight.
The study area is the Sandy River, Oregon, which flows from the western slopes
ofMount Hood northwest to the Columbia River (Figure 4). Recent LiDAR data and
aerial photography capture the variety of water surface characteristics in the Sandy River,
which range from shooting flow to wide pool-riffle formations. The recent removal of
the large run-of-river Marmot Dam upstream of the analysis sites has also generated
interest in the river's hydraulics and geomorphology.
550000 556000 560000
-. Portland Sandy River
Clack. fna County
IHillshaded area represents 2006 LiDAR
extent. Ol1hophotography was collected
only along the Sandy River channel
within the LiDAR extent.
545000 550000 555000 560000 565000 570000
Figure 4. Site Map. Site area map showing location of analysis reaches within the 2006
and 2007 LiDAR coverage areas. Olihophotography was also collected for the 2006
study, but was collected only along the Sandy River channel.
Floodplain longitudinal slopes along the Sandy River average 0.02 and reach a
maximum of 0.04. The Sandy River has closely spaced pool-riffles and rapids in the
upper reaches, transitioning to longer sequenced pool-riffle morphology in the middle
and lower reaches. The Sandy River bed is dominated by sand. Cobbles and small
boulders are present mostly in areas of riffles and rapids. Much of the channel is incised
with steep slopes along the channel boundaries. The flow regime is typical of Pacific
Northwest streams, with peak flows in the winter months ofNovember through February
and in late spring with snowmelt runoff (Figure 5). Low flows occur between late
September and early October. The average peak annual flow at the Sandy River station
below Bull Run River (USGS 14142500) is 106cms. Average annual low flow for the
same gauge is 13.9cms.
USGS 14142500 SRNDY RIVER BL~ BULL RUN RIVER, NR BULL RUN, OR
200 k.===_~~~=~~~=.......==",,=~-........==~ ~....J
Jan 01Feb Ollar 01Rpr O:t1ay 01Jun 01Jul 01Rug OJSep 010ct 01Nov O:IJec 01
2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006
~I\\ ,1\ 1\ j\
1"J'fn I\. I,
) \ , ,;'
I I" 'I'•.,
] 30000 ~~-~----~-------------~-------,
Ql to 1000
Hedian daily statistic <59 years)
Daily nean discharge
--- Estinated daily nean discharge
Period of approved data
Period of provisional data
Figure 5, Annual Hydrograph of Sandy River. US Geological Survey gaging
station annual hydrograph of Sandy River, Oregon at Bull Run River. Data from
Vegetation is mostly a mixture of Douglas fir and western red hemlock (Figure 6).
Other vegetation includes palustrine forest found in the upper portions of the study area,
and agricultural lands found in the middle and lower portions. Douglas fir and western
red hemlock make up 87% of vegetated areas, palustrine forest 5%, and agricultural lands
5%, the remaining 3% is open water associated with the channel and reservoirs (Oregon
GAP Analysis Program, 2002). The city of Troutdale, OR abuts the lower reaches of the
Sandy River. Along this stretch of river Himalayan blackberry, an invasive species,
dominates the western banks (Figure 7). The presence of Himalayan blackberry is
significant because LiDAR has trouble penetrating through the dense clusters of vines.
When this blackberry is close to the water's edge it is difficult to accurately define the
545000 550000 555000 560000 565000 570000
Oregon GAP Vegetation
0 g 0
~ Reach 1 I '~"
• Oregon Trout ale
0 8 00
Douglas Fir-W. Hemlock-W. Red Cedar Forest
Mixed Conifer/Mixed Deciduous Forest
Red Alder-Big Leaf Maple Forest
545000 550000 555000 560000 565000 570000
Figure 6. Oregon GAP Vegetation within Study Area. 1999 Oregon GAP
Analysis data for Sandy River area. Map shows how the Sandy River area is
dominated by Douglas fir forest with areas of palustrine forest and agricultural lands
(Oregon Natural Heritage Program, 1999).
Figure 7. Photo of Himalayan Blackberry on Sandy River. Himalayan blackberry
near mouth of the Sandy River March, 25th 2007. Photo by John English.
This study focuses on three reaches of channel that represent a range of water
surface conditions along the river. Reach 1 is a I80-m long pool-riffle reach located 3.7
river kilometers upstream from the mouth, and is where we collected field data shortly
after the 2007 LiDAR flight (Figure 8a). The bed is sandy in this reach and can change
dramatically during high flows. The bank full width of Reach 1 is approximately 108
meters at its widest point. At the downstream end of the riffle, the channel is constricted
by riprap placed along the banks as the river flows under a bridge. Vegetation comprises
deciduous and conifer trees such as Douglas fir, hemlock, and cottonwoods. Blackberry
is present along the channel, but is not so dense that it obscures the active channel
Figure 8. Reach 1 Site Area Map with Photo. Reach 1 site area. Top figure (a)
shows approximate width at bank full and length of field data collections. Yellow
circles represent points along stream margins where water surface elevations were
surveyed. Bottom photo (b) looks downstream from total station location.
Reach 2 (Figure 9) is located approximately 23.5 kIn upstream from the mouth of
the Sandy River and is 1,815 meters in length. The widest portion of channel at
approximate bank full is 116m. The channel consists of a large meander with sinuosity
of 1.38 and consists of six riffles and five pools spaced at regular intervals. The substrate
consists of sands with small boulders and large cobbles dominating riffle areas. Cobbles
and boulders have likely been introduced to the channel as a result of mass wasting.
Douglas fir dominates along banks.
0° 200 MetersO 0
Figure 9. Reach 2 Site Area Map. Site map of Reach 2. Reach 2 contains 359 cross
sections derived from LiDAR and 3,456 sample points. Inset map shows cross section
sample locations derived from LiDAR and smooth/rough water surface delineations
used in analysis.
Reach 3 is located 40.7km upstream from the mouth of the Sandy and is 2,815
meters in length (Figure 10). The widest portion of this section at approximate banle full
is 88 meters. The upstream extent of the channel includes the supercritical flow of
Marmot Dam. The channel is incised and relatively straight with a sinuosity of 1.08. Fine
sands dominate the channel bed with some boulders likely present from mass wasting
along valley walls. As with Reach 2, Douglas fir dominates bank vegetation along.
Inset mAp displays UDAR point I densily alol1g willl cross seellon
Sanlpleing dala LiDAR cross section
SAmple locations were used to
eX1mcl poinl density values.
503 fOC I 000 '.1..Hrs
. Reach 3
Figure 10. Reach 3 Site Area Map. Site map of Reach 3. Inset map shows point
LiDAR water surface points. Reach 3 contains 550 cross sections and 3,348 sample
points. Visual examination of this map allows one to see how point density varies
within the active channel.
LiDAR data and orthophotography were collected in 2006 and additional LiDAR
data were collected over the same area in 2007. Field measurements were obtained five
days after the 2007 LiDAR flight in order to compare field measurements of water
surface slope to LiDAR-based measurements. Time of flight field measurements of
water surface elevations were not obtained for the 2006 flight, but the coincident
collection of LiDAR data and orthophotos provide a basis for evaluating variability of
LiDAR-based slopes over different channel types as identified from aerial photos.
Following sections provide more detail regarding these methods.
LiDAR Data and Image Acquisition
All LiDAR data were collected using a Leica ALS50 Phase II LiDAR system
mounted on a Cessna Caravan C208 (see Table 1 for LiDAR acquisition specifications).
The 2006 LiDAR data were collected October 2211d and encompassed 13,780 hectares of
high resolution (2':4 points/m2
) LiDAR data from the mouth of the Sandy River to Marmot
Dam. Fifteen centimeter ground resolution orthophotography was collected September
, 2006 along the riparian corridor of the Sandy River from its mouth to just above the
former site ofMarmot dam (Figure 4). The 2007 LiDAR were collected on October 8th
and covered the same extent as the 2006 flight, but did not include orthophotography.
Data included filtered XYZ ASCII point data, LiDAR DEMs as ESRI formatted grids at
0.5 meter cell size. Data were collected at 2':8 points per m2 providing a data set with
significantly higher point density than the 2006 LiDAR data.
The 2006 LiDAR data were collected in one continuous flight. 2006
orthophotography was collected using an RC30 camera system. Data were delivered in
RGB geoTIFF format. LiDAR data were calibrated by the contractor to correct for IMU
position errors (pitch, roll, heading, and mirror scale). Quality control points were
collected along roads and other permanent flat features for absolute vertical correction of
data. Horizontal accuracy ofLiDAR data is governed by flying height above ground
with horizontal accuracy being equal to 1I3300th of flight altitude (meters) (Leica, 2007).
Table 1. Reported Accuracies of 2006 and 2007 LiDAR. Reported Accuracies
and conditions for 2006 and 2007 LiDAR data. (Watershed Sciences PGE LiDAR
Delivery Report, 2006, Watershed Sciences DOGAMI LiDAR Delivery Report,
2007). Relative Accuracy is a measure of flight line offsets resulting from sensor
2006 LiDAR 2007 LiDAR
Flying height above ground level meters (AGL) 1100 1000
Absolute Vertical Accuracy in meters 0.063 0.034
Relative Accuracy in meters (calibration) 0.058 0.054
Horizontal Accuracy (l/3300th * AGL) meters 0.37 0.33
Discharge @ time of flight (cms) 13.05 20.8 - 21.8
LiDAR data collection over the Reach 1 field survey location was obtained in a
single flight on October 8, 2007 between 1:30 and 6:00 pm. During the LiDAR flight,
ground quality control data were collected along roads and other permanent flat surfaces
within the collection area. These data were used to adjust for absolute vertical accuracy.
Field Data Acquisition
A river survey crew was dispatched at the soonest possible date (October 13,
2007) after the 2007 flight to collect ground truth data within the Reach 1. The initial aim
was to survey water surface elevations at cross sections of the channel, but the survey
was limited to near shore measurements due to high velocity conditions. We collected
187 measurements of bed elevation and depth one to fifteen meters from banks along
both sides of the channel (Figure 8a) using standard total station longitudinal profile
survey methods (Harrelson, 1994). Seventy-six and 98 measurements were collected
along the east and west banks, respectively, at intervals of approximately 1 to 2 meters.
Thirteen additional measurements were collected along the east bank at approximately
ten meter intervals. Depth measurements were added to bed elevations to derive water
surface elevations. Discharge during the survey ranged between 22.5 and 22.7 cms
during the survey of the east bank and remained steady at 22.5 cms during the survey of
the west bank (USGS station 14142500).
The goal ofLiDAR processing for this project was to classify LiDAR point data
within the active channel as water and output this subset data for further analysis. The
LiDAR imagery was first clipped to the active channel using a boundary digitized from
the 2006 high resolution orthophotography. LiDAR point data were then reclassified to
remove bars, banks, and overhanging vegetation (Figure 11).
Figure 11. LiDAR Point Filtering Processing Step. LiDAR processing steps.
Top image shows entire LiDAR point cloud clipped to active channel boundary.
Lower image shows the final processed LiDAR points representing only those
points that reflect off the water surface. All bars and overhanging vegetation
have been removed as well.
Water points were classified using the ground classification algorithm in
Terrascan© (Soininen, 2005) to separate water surface returns from those off of
vegetation or other surfaces elevated above the ground. The classification routine uses a
proprietary mathematical model to accomplish this task.
Once the ground classification was finished, classified points were visually
inspected to add or remove false positives and remove in-channel features such as bar
islands. A total of 11,593 of 1,854,219 LiDAR points were classified as water. Points
classified as water were output as comma delimited x,y,z ASCII text files (XYZ), then
converted to a 0.5 meter linearly interpolated ESRI formatted grid using ESRI
geoprocessing model script.
Calculation of Water Surface Slopes
Water surface slopes were calculated using the rise over run dimensionless slope
equation where the rise is the vertical difference between upstream and downstream
water surface elevations and run is the longitudinal distance between elevation locations.
LiDAR data is typically used in grid format. For this reason grid data were used
for calculation of water surface slopes. We used linear interpolation to grid the LiDAR
point data as this is the standard method used by the LiDAR contractor. In order to
compare the LiDAR and field data it was also necessary to interpolate field
measurements to create a water surface for the entire stream. The field data-based DEM
was created using kriging interpolation within ArcGIS Desktop Spatial Analyst (Figure
12). No quantitative analysis was performed to evaluate the interpolation method of the
field-based water surface. The kriging interpolation was chosen because it producex the
smoothest water surface based on visual inspection when compared to linear and natural
neighbor interpolations, which generated irregular fluctuations that were unrealistic for a
water surface. The kriged surface provided a water surface elevation model for
comparative analysis with LiDAR.
Figure 12. Field DEM Interpolated using Kriging. Field DEM interpolated
from field survey points using kriging method found in ArcGIS Spatial Analyst.
DEM has been hiIlshaded to show surface characteristics. The very small
differences in water surface elevations generate only slight variations in the
To compare LiDAR and field-based water surface slopes, water surface elevations
from the LiDAR and field-based DEMS were extracted at the same locations along
Reach I. To accomplish this, 37 cross sections were manually constructed at
approximately Sm spacings (Figure 13). Cross sections comparisons were used rather
than point-to-point comparisons between streamside field and LiDAR data points because
the cross sections provide water surface slopes that are more representative of the entire
channel. The Sm interval spacing was considered to be a sufficient for fine resolution
slope extraction. Because cross section center points were used to calculate the
longitudinal distance and because the stream was sinuous, the projection of the cross
sections from the center line to the banks led to stream side distances between cross
sections that differed from Sm.
Cross Section Data Roughness Delineation
Cross Section Sample Locations _ Rough
~ each 1
Figure 13. Reach 1 LiDAR Cross Sections and Sample Point Locations. Reach I
LiDAR-derived cross section sample locations and areas of smooth and rough water
surface delineations. 37 cross section and 444 sample points lie within Reach 1.
Cross sections were extracted using a custom ArcObjects VBA script (Appendix
A). This script extracted 1 cell nearest neighbor elevations along the transverse cross
sections at 5 meter intervals creating 444 cross section sample locations (Figure 13).
Cross section averages were calculated using field-based and LiDAR-based elevation
water surface grids. The average cross sectional elevation value for field and LiDAR data
were then exported to Excel files, merged with longitudinal distance between cross
section, and used to calculate field survey-based and LiDAR-based slopes between cross
Reaches 2 and 3, for which only LiDAR data were available, were sampled using
the same cross sectional approach used in Reach 1. The data extracted from these
reaches were used to characterize how LiDAR-based elevations, slopes and point
densities interact with varying water surface roughness. Within Reach 2, 359 cross
sections were drawn and elevations were sampled every five meters along each cross
section creating 3,456 cross section sample locations (Figure 9). Reach 3 contained 550
cross sections and 3,348 cross section sample locations (Figure 10). Slopes were
calculated between each cross section.
Evaluating LiDAR Slope Accuracies and Controls
The accuracy of elevation data is the major control on slope accuracy, so a
comparative analysis was performed using field survey and LiDAR elevations. First,
field-based and LiDAR slopes were calculated at distance intervals of five, ten and
twenty meters using average cross section elevations to test the sensitivity of the slopes to
vertical inaccuracies in the LiDAR data. The field and LiDAR elevations were
differenced using the same points used to create average cross section elevations.
Differences were plotted in the form of histogram and cumulative frequency plot after
transforming them into absolute values. Descriptive statistics were calculated to examine
the range, minimum, maximum, and mean offset between data sets. Finally LiDAR and
field-based values were compared using regression analysis.
This study also examined the effects of water surface roughness on LiDAR
elevation measurements, LiDAR point density, and LiDAR derived water surface slopes.
Each reach was divided into smooth and rough sections based on visual analysis of the
orthophoto data. One-meter resolution slope rasters were created from the LiDAR water
surface grids using ArcGIS Spatial Analyst. One meter resolution point density grids
were created from LiDAR point data (ArcGIS Spatial Analyst). Using the cross section
sample points, values for water surface type, elevation, slope, and point density were
extracted within each reach. Point sample data were transferred to tabular format, and
average values were generated for each cross section. These tables were used to calculate
descriptive statistics associated with water surfaces such as elevation variance, average
slope variance, average point density, and average slope.
It is assumed in this study that smooth water surfaces are associated with pools
and thus ought to have relatively low slopes. Conversely rough water surfaces are
assumed to be representative of riffles and rapids, and thus ought to have relatively
steeper slopes. Reach 1 contains field data, so slopes from LiDAR and field data were
compared with respect to water surface conditions as determined from the aerial photos.
Results of this study encompass three analyses. Elevation analysis describes the
statistical difference between LiDAR and field-based water surface elevations for Reach
1. Slope analysis compares LiDAR derived and field-based slopes calculated at 5, 10,
and 20m longitudinal distances. These analyses aim to quantify both slope accuracy and
slope sensitivity. Lastly, water surface analysis examines the relationship between
LiDAR measured water surface slopes, point density, and water surface roughness.
Comparison of Absolute Elevations from Field and LiDAR Data in Reach 1
The difference between water surface elevations from LiDAR affects the
numerator within the rise over run equation, which in tum affects slope. This elevation
analysis evaluation quantifies differences between field and LiDAR data. LiDAR-based
cross section elevations were differenced from field-based cross section elevations.
Difference values were examined through statistical analysis.
In terms of absolute elevations relative to sea level, the majority of LiDAR-based
water surface elevations were lower than field-based elevations, although the LiDAR
elevations were higher in the upper portion ofReach 1. Differences ranged between -0.04
and 0.05m with a mean absolute difference between field and LiDAR elevations of
0.02m (Figure 14 and Table 2). The range of differences is within the expected relative
accuracies of LiDAR claimed by the LiDAR provider. Elevations for field and LiDAR
data are significantly correlated with an R2 of 0.94 (Figure 15).
The negative offset was expected given that discharge at time of LiDAR
acquisition was lower than discharge at time of field data acquisition. Discharge during
field acquisition ranged between 22.5 and 22.7 cfs, while discharge during LiDAR
acquisition was between 20.8 and 21.8cfs. The portion of Reach 1 where LiDAR water
surface measurements were higher than field measurements may be related to difference
in discharge or change in bed configuration. Overall results showed that LiDAR data and
field-based water surface measurements are comparable.
Distribution of Elevation Differences Between Field
and LiDAR Water Surfaces
r:: ell 5
-0.05 -0.04 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 0.05 More
Elevation Difference, Field - L1DAR (m)
Figure 14. Differences Between LiDAR and Field Based Elevations. Elevation
difference statistics between cross sections derived from field and LiDAR elevation
data. Positive differences indicate that field-based elevations were higher than
LiDAR; negative differences indicate LiDAR elevations were higher. Values on x
axis represent minimum difference within range. For example, the 0.01 category
includes values ranging from 0.01 to 0.0199.
R2 =0.94 ""..,;
./... ./ .- ./ •
./ • ./.
Table 2. Results of LiDAR and Field Elevation Comparison.
Descriptive and regression statistics for absolute difference lField - LiDARI
values between cross section elevations. All units in meters. Sample size is
Standard Deviation 0.013
Range of difference 0.093
Minimum difference 0.002
Absolute maximum difference 0.047
Confidence Level(95.0%) (m) 0.004
Elevation Comparison of Field and LiDAR Water Surface
> iii 5.62
<C 5.48 o
5.46 5.48 5.50 5.52 5.54 5.56 5.58 5.60 5.62 5.64 5.66 5.68 5.70
Field Water Surface Elevation (m)
Figure 15. Regression of LiDAR and Field Cross Section Elevations.
Regression of field-based (x) and LiDAR-based (y) cross section elevations.
Comparison of longitudinal profiles offield and LiDAR water surfaces shows a
clear relationship in overall shape (Figure 16), capturing similar trends in longitudinal
profiles. Figure 16 shows field and LiDAR profiles become more similar in shape as
distance between cross sections increases. In terms of overall shape, the greatest
differences occur in the upper 30 m, where LiDAR-based profiles demonstrate a higher
slope than do field-based measurements. Because of the five day lag between LiDAR
and field measurements in this mobile bed stream, it is impossible to know the degree to
which this difference represents error in measurements or real change in the system.
5 meter Longitudinal Profile Comparison
20 40 60 80 100 120 140 160 180
5.75 .s 5.70 ~" _ • •• • :. 5 <reter Field Profile II::
5.65 ...- .=....:....:l..,... H.T• tI.:!..~.....~.Io-,•..-..;.....-.------j. 5rreterL,DARprof,lel-
..0.._. 5.60 .. • •• ~ 5.55 -1------------ .~•.~•.-.-.--------
~ 5.50 +---------------"''-.--'~~ ~.. ,~ "yT1I:!'-'---
W 5.45 -1---------------.-::..---'.1-.........-- ...:I:C"IL'J-"----
Longitudinal Distance Down Stream (m)
10 meter Longitudinal Profile Comparison
5.70 . • [,.10 rreler Field Profile! I:
I 5.65 • • , . • • • 10 rreter LiDAR Profile • • • I:: 5.60 • • 0
:;:; • • >Cll 5.55 • • ~ • • w 5.50 • • • • • • • • • 5.45
0 20 40 60 80 100 120 140 160 180
Longitudinal Distance Down Stream (m)
20 meter Longitudinal Profile Comparison
5.70 • ,. 20 <reter Field Profile .s 5.65 • . • • • 20 rreler LiDARProfile • I:: 5.60 • 0
:;:; >Cll 5.55 •• Q) W 5.50 •• • , 5.45 .
0 20 40 60 80 100 120 140 160 180
Longitudinal Distance Down Stream (m)
Figure 16. Comparison of LiDAR and Field Longitudinal
Profiles (5, 10, 20 meters). Longitudinal profiles of a) 5 meter, b)
10 meter, and c) 20 meter cross section elevations.
Slope in this study is calculated as the dimensionless ratio of rise over run. As
noted in the Methods section, slopes were calculated over three different horizontal
intervals to test the sensitivity of the LiDAR's internal relative accuracy.
Differences in Sm LiDAR and field-based slopes derived from cross sections
reveal substantial scatter (Figure l7a), although they clearly covary. Ten meter interval
slopes show a stronger relationship (Figure 17b), while slopes based on cross sections
spaced 20 m apart have the strongest relationship (Figure l7c). The slope associated with
regression of field and LiDAR elevation data is not approximately 1 as one might expect.
This is because LiDAR elevations are higher than field elevations at the upstream end of
the reach, and lower at the downstream end.
5m Slope Comparison
= 0.58x - 0.001
R2 = 0.38
Field Slope (Rise/Run)
10 meter Slope Comparison
y = 0.63x - 0.001
R2 = 0.51
Field Slope (Rise/Run)
20 meter Slope Comparison
(/l i2 ~.01 -Q) c. o
=0.66x - 0.001
R2 = 0.80
Field Slope (Rise/Run)
Figure 17. Regression of Field and LiDAR Based Slopes (5,10,20
meters). Scatter plots showing comparisons between slope values calculated
at distance intervals of a) 5 meters, b) 10 meters, and c) 20 meters.
Figure 18 shows how the range of differences between LiDAR and field-based
water surface slopes decrease as longitudinal distance increases. Five meter slope
differences ranged between -0.004 and 0.004 (Figure 18a). Ten meter slope differences
ranged between -0.002 and 0.003 (Figure 18b). Twenty meter slope differences ranged
between 0 and 0.002 (Figure 18c).
Differences of Slope at 5m Between Field and
0" 4 .Q..l u.
SIll>< SIl"> SIll\-
~<::J <;:><::J <;:><::J
SIl" ~ SIl" SIll\- SIl"> SIll>< ~/l,
r;:,<::J ~'::; ~'::; ~'::; ~'::; ~o
Slope Difference (Field-LiDAR)
Differences of Slope at 10m Between Field and
Slope Difference (Field-LiDAR)
Differences of Slope at 20m Between Field and
~~I\- ~~" ~ ~~" ~~I\- ~~"> ~~I>< o"/l,
<;:>.~. ~.~.~.~. ~
Slope Difference (Field-LiDAR)
Figure 18. Differences Between LiDAR and Field Based Slopes (5,
10,20 meters). Histogram charts showing difference values between
field and LiDAR derived slopes at a) 5 meter slope distances, b) 10
meter slope distances, and c) 20 meter slope distances.
The mean difference between slopes decreases from 0.0017 to 0.0007 as slope
distance interval is increased. Maximum slope difference and standard deviation of
offsets decrease from 0.001 to 0.0005 and 0.0047 to 0.0014 respectively. Regression
analysis of these data show a significant relationship for all three comparisons, and
adjusted R2 increased from 0.357 to 0.763 with slope distance interval (Table 3).
Table 3. Results of LiDAR and Field Slope Comparison (5, 10,20
meters). Descriptive and regression statistics for offsets between field and
LiDAR derived slope values (Field minus LiDAR). Slope values are
dimensionless rise / run. All data is significant at 0.01.
Distance Interval 5m 10m 20m
Mean 0.0017 0.0012 0.0007
Standard Deviation 0.0010 0.0007 0.0005
Range of Difference 0.0080 0.0047 0.0024
Minimum difference 0.0000 0.0000 0.0001
Maximum difference 0.0047 0.0026 0.0015
Count 36 16 8
Adjusted R squared 0.36 0.47 0.76
Water surface slope for the entire length of Reach 1 (l59.32m) was compared and
yielded a difference of 0.0005. This difference is smaller (by 0.0002) than the difference
between 20 meter slope (Table 4). Slope was calculated by differencing the most
upstream and downstream cross sections and dividing by total length of reach.
Differences between LiDAR and field-based slopes may represent real change due to the
five day lag between data sets and difference in discharge.
Table 4. Results of Reach 1 Slope Comparison. Comparison of
slopes calculated using the farthest upstream and downstream cross
section elevation values. Slope values have dimensionless units
stemming from rise over run.
Upper Lower Reach
Elevation (m) Elevation (m) Len2th (m) Slope
Field 5.652 5.491 159.32 -0.0010
LiDAR 5.697 5.455 159.32 -0.0015
Surface Roughness Analysis
Water surface condition was characterized as smooth or rough based on 2006
aerial photography (Figure 19). Surface roughness was examined to understand its effect
on LiDAR data within the active channel, as well as LiDAR's ability to potentially
capture difference in water surface turbulence. Table 5 shows statistics with relation to
water surface condition for all three reaches.
Figure 19. Relationship of Water Surfaces to LiDAR Point Density. 2006
aerial photos were used to delineate rough and smooth water surfaces. Image on
left shows a transition between rough water surface (seen as white water) and
smooth water surface (seen as upstream pool). Image on right shows LiDAR point
density in points per square meter.
In all reaches point density, variance of elevations, and water surface slopes were
significantly higher in rough surface conditions. These results indicate that LiDAR point
density is directly related to the roughness of a water surface and that is capturing the
rough water characteristics one would expect in areas where turbulence generates surface
Table 5. Water Surface Roughness Results for Reach 1,2, and 3. Water
surface statistical output for rough and smooth water surface of Reaches 1, 2, and
3. Results within table represent average values for each Reach. Slope values
have dimensionless units from rise over run equation derived from ESRI
generated slope grid. Point density values based on points/m2
variance in meters.
Reach 1 Reach 2 Reach 3
No. of Sample Points 153 1981 1968
Avg Slope -0.013 -0.011 -0.007
Point Density (pts/mL
) 1.195 1.002 1.217
Elevation Variance (m) 0.003 0.018 0.041
No. of Sample Points 290 1474 1378
Avg Slope 0.0075 -0.0006 -0.0033
Point Density (pts/mL
) 0.149 0.550 0.480
Elevation Variance (m) 0.001 0.0077 0.024
Within Reach 1, cross section elevations were separated into rough and smooth
water conditions and slopes were calculated using field and LiDAR data sets (Table 6).
Again, results showed that rough water surfaces have greater slopes than smooth water
surfaces. The smooth water surface of Reach 1 yielded a larger discrepancy between
field and LiDAR derived slopes compared to rough water surface. This is because small
differences between LiDAR and field elevations generate larger proportional error in the
rise / run equation when total elevation differences between upstream and downstream
Table 6. Results of Reach 1 Water Surface Roughness Comparison. Reach 1
water surface roughness slope analysis. Reach 1 was divided into smooth and
rough water surfaces based upon visual characteristics present in aerial
photography. Slopes were calculated for each area and compared with field data
to examine accuracy.
Surface Reach Upper Lower Slope
Type Lenl!th (m) Elevation (m) Elevation (m) Slope Difference
Field Smooth 83.11 5.652 5.642 -0.0001 N/A
LiDAR Smooth 83.11 5.697 5.612 -0.0010 0.0009
Field Rough 71.73 5.635 5.491 -0.0020 N/A
LiDAR Rough 71.73 5.592 5.455 -0.0019 -0.0001
Prior to collections of the 2007 data, Reach 3 contained the former Marmot Dam
that was dismantled on October 19th
, 2007 (Figure 20). The areas at and directly below
the dam are rough water surfaces. The super critical flow at the dam yielded a slope of -
0.896 (Table 7). The run below the dam contained low slope values of less than -0.002.
Both the dam fall and adjacent run yielded high point densities of greater than 2 points
per square meter.
o Cross Section Sample Locations
L1DAR derived Slope Model
25 50 75 100 125 150 ~.',eters
I I I I I I
Figure 20. Marmot Dam: Orthophotography and Colorized Slope Model.
Mannot Dam at far upstream portion of Reach 3. Image on left shows dam site in
2006 orthophotography. Image on right shows the increase in slope associated with
the dam. Marmot Dam was removed Oct. 19th
Table 7. Subset of Reach 3 Water Surface Roughness Analysis Near
Marmot Dam. Subset of Reach 3 immediately surrounding Marmot Dam
roughness analysis containing values for Mannot Dam. The roughness results
fell within expectations showing increases in slope at the dam fall and high point
densities at the dam fall and immediate down stream run.
Habitat Type Avg Slope Point Density Point Density Variance
Dam Fall -0.896 2.284 1.003
Dam Run -0.001 2.085 5.320
The elevation analysis portion of this study shows that LiDAR can provide water
surface profiles and slopes that are comparable to field-based data. The differences
between LiDAR and field based measurements can be attributed to three potential
sources. The first is the relative accuracy of the LiDAR data which has been reported
between O.05m and O.06m by the vendor. The second source can be associated with the
accuracy of field based measurements which are similar to the relative accuracy of the
LiDAR (O.03m-O.05m). Lastly, the discharge differed between field data collection and
LiDAR collection by O.02cms. It is possible that much of the O.05m difference observed
through most of the Reach 1 profile (Figure 16) could be attributed to the difference in
discharge and changes in bed configuration, but without further evidence, the degree of
difference due to error or real change cannot be identified. Even if one attributes all the
difference to error in LiDAR measurements, the overall correspondence ofLiDAR and
field measurement (Figure 15 and 16) indicates that LiDAR-based surveys are useful for
many hydrologic applications.
In the upper portion of the reach, the profiles display LiDAR elevations that are
higher than the field data elevations, whereas the reverse is true at the base of the reach.
This could be a function of difference in discharge between datasets, change in bed
configuration, or an artifact of low point density. Low density of points forces greater
lengths of interpolation between LiDAR points leading to a coarse DEM (Figure 21).
Overall, the analysis Reach 1 profile indicates that LiDAR was able to match the fieldbased
elevation measurements within ±O.05m.
Rough & Smooth Wa~t:e:-r~S~u=rf;:a~c:e:s~rz~~J,;~~
Grid Interpolation in Low Point Density
Figure 21. LiDAR Point Density versus Interpolation. Side by side image
showing long lines of interpolation associated with smooth water surfaces (right
image). Smooth water surfaces tend to have low LiDAR point density. The image on
the right shows a hillshade ofthe LiDAR DEM. The DEM has been visualized using
a 2 standard deviation stretch to highlight long lines of interpolation.
The comparability of LiDAR and field-based slopes showed a significant trend
with increasing downstream distances between cross sections. Adjusted R2 values
increased from 0.36 to 0.76 and the range of difference between field and LiDAR based
slopes decreased from 0.0047 to 0.00 14 as longitudinal distance increased from 5 to 20-
m. This suggests that the 0.05m of expected variation of LiDAR derived water surface
elevation has less effect on water surface slope accuracy as distance between elevation
measurements points increases. Likewise, slopes accuracies along rivers with low
gradients will improve as the longitudinal distance between elevation points increases.
Overall, data has shown that LiDAR can measure water surface slopes with mean
difference relative to field measurements of 0.017, 0.012, and 0.007 at horizontal
distances of 5, 10, and 20 meters respectively. Although the discrepancy between field
and LiDAR-based slopes is greatest at 5-m intervals, the overall slopes (Fig 17) and
longitudinal profiles (Fig 16) even at this distance generally correspond. The use of a 5m
interval water surface slope as a basis for comparison is really a worst case example, as
water surface slopes are usually measured over longer reach scale distances where the
discrepancy between LiDAR and field-based measurements is lower. The continuous
channel coverage and accuracies derived from LiDAR represent a new level of accuracy
and precision in terms of spatial extent and resolution of water surface slope
Analysis of surface roughness found that rough water surfaces had significantly
higher point densities than smooth water surfaces. Rough water surfaces averaged at
least 1 point/m2
, while smooth water surfaces averaged less than 1 point/2m2
Longitudinal profiles of Reach 1 indicate the most accurate water surface measurements
occur in areas of higher point density (Fig. 16). Future applications that attempt to use
LiDAR to measure water surface slope ought to sample DEM elevations from high point
density areas of channel.
Water surface analysis also showed trends relating water surface roughness and
slope. Rough water surfaces for all three analysis reaches averaged larger average slope
values than smooth water surfaces. This is because rough water surfaces are commonly
associated with steps, riffles, and rapids. All three of these habitat types are areas have
higher slopes than smooth water habitats. Smooth water surfaces are commonly
associated with pools or glides, which would be areas of lower slope. Future research
should examine the potential for using LiDAR to characterize stream habitats based on
in-stream point density and slope.
This study is not without its limitations. The field area used to test the accuracy of
LiDAR is only representative of a small portion of the Sandy River. Comparisons of
field and LiDAR data would be improved by having mid-channel field data. One might
also question the use of field based water surface slopes as control for measuring
"accuracy". Water surface slope is difficult to measure for reasons stated earlier in this
paper. One might make the argument that there is no real way to truly measure LiDAR
accuracy of water surface slope, and that LiDAR and field based measurements are
simply comparable. In this context, LiDAR holds an advantage over field based
measurements given its ability to measure large sections of river in a single day.
LiDAR has a distinct advantage over traditional methods of measurement in that
measurements are returned from the water surface, and consequently not subject to errors
associated with variability of surface turbulence piling up against the measuring device.
LiDAR can also capture long stretches of channel within a few seconds reducing the
influence of changes in discharge. LiDAR data in general does have its limitations.
LiDAR data are only as accurate as the instrumentation and vendor capabilities. LiDAR
must be corrected for calibrations and GPS drift to create a reliable data set, and not all
LiDAR vendors produce the same level of quality.
LiDAR data may be more accurate in some river reaches than others. The study
reaches of this study contained well defined open channels, which made identifying
LiDAR returns off the water surface possible. Both LiDAR data sets were collected at
low flows. Flows that are too low or channels that are too narrow may limit ability to
extract water surface elevations because of protruding boulders or dense vegetation that
hinders accurate measurements. In some cases vegetation within and adjacent to the
channel may interfere with LiDAR's ability to reach the water surface. Researchers
should consider flow, channel morphology, and biota when obtaining water surface
slopes from LiDAR.
This paper examined the ability of LiDAR data to accurately measure water
surface slopes. This study has shown that LiDAR data provides sufficiently accurate
elevation measurements within the active channel to accurately measure water surface
slopes. Measurement of water surface slope with LiDAR provides researchers a tool
which is both more efficient and cost effective in comparison with traditional field-based
survey methods. Additionally, analysis showed that LiDAR point density is significantly
higher in rough surface conditions. Water surface elevations should be gathered from
high point density areas as low point density may hinder elevation accuracy. Channel
morphology, gradient, flow, and biota should be considered when extracting water
surface slopes as these attributes influence water surface measurement. Further study
should examine accuracy of LiDAR derived water surface slopes in channel
morphologies other than those in this study. Overall, the recognition that LiDAR can
accurately measure water surface slopes allows researchers an unprecedented ability to
study hydraulic processes for large stretches of river.
APPENDIX ARCGIS VBA SCRIPT CODE
Public g---.pStrmLayer As ILayer ' stream centerline layer selected by user (for step 1)
Public g_StrearnLength As Double ' stream centerline length (for step 1)
Public g_InputDistance As Integer 'As Double 'distance entered by user (for step 1)
Public g_NumSegments As Integer I number of sample points entered by user (for step 1)
Public gyPointLayer As ILayer I point layer created from stream centerline (for step 1)
Public g]ntShpF1Name As String I point layer pathname (for step 1)
Public gyMouseCursor As IMouseCursor 'mouse cursor
Public g_LinearConverson As Double I linear conversion factor
Public gyDEMLayer As IRasterLayer I DEM layer (for steps 3 and 4)
Public g_DEMConvertUnits As Double I DEM vertical units conversion factor (for steps 3 and 4)
Public g_MaxSearchDistance As Double 'maximum search distance (for step 4)
Public L NumDirections As Integer I number of directions to search in (for step 4)
Public g_SampleDistance As Double 'sample distance (for step 5)
Public g_SampleNumber As Double ' total sample points (for step 5)
Public g_VegBeginPoint As Boolean I where to start the calucaltion (for step 5)
Public g_VegCaclMethod As Boolean 'which method for Vegetation Calculation (for step 5)
Public gyContribLayer As ILayer ' contributing point layer (for step 6)
Public gyReceivLayer As ILayer 'receiving point layer (for step 6)
Public gyOutputLayerName As String I output shapefile (for step 6)
Function VerifyField(fLayer As ILayer, fldName As String) As Boolean
I verify that topo fields are in the stream centerline point layer
Dim pFields As IFields
Dim pField As IField
Dim pFeatLayer As IFeatureLayer
Dim pFeatClass As IFeatureClass
Set pFeatLayer = fLayer
Set pFeatClass = pFeatLayer.FeatureClass
Set pFields = pFeatClass.Fields
For i = 0 To pFields.FieldCount - 1
Set pField = pFields.Field(i)
IfpField.Name = fldName Then
VerifyField = True
VerifyField = False
Function Ca1cPointLatLong(inPnt As IPoint, inLayer As ILayer) As IPoint
, in point layer
Dim pFLayer As IFeatureLayer
Set pFLayer = inLayer
, spatial reference environment
Dim pInSpatialRef As ISpatialReference
Dim pOutSpatialRef As ISpatialReference
Dim pGeoTrans As IGeoTransformation
Dim pInGeoDataset As IGeoDataset
Set pInGeoDataset = pFLayer
Dim pSpatRefFact As ISpatialReferenceFactory
, get map units of shapefile spatial reference
Dim pPCS As IProjectedCoordinateSystem
Set pPCS = pInGeoDataset.SpatialReference
'set spatial reference environment
Set pSpatRefFact = New SpatialReferenceEnvironment
Set pInSpatialRef= pInGeoDataset.SpatialReference
Set pOutSpatialRef= pSpatRefFact.CreateGeographicCoordinateSystem(esriSRGeoCS_WGS1984)
Set pGeoTrans =
Dim pOutGeom As IGeometry2
Set Ca1cPointLatLong = New Point
Set CalcPointLatLong.SpatialReference = pInSpatialRef
Ca1cPointLatLong.PutCoords inPnt.X, inPnt.Y
Set pOutGeom = Ca1cPointLatLong
pOutGeom.ProjectEx pOutSpatialRef, esriTransformForward, pGeoTrans, 0, 0, °
'MsgBox inPnt.X &" "& inPnt.Y & vbCrLf& Ca1cPointLatLong.X &" "& Ca1cPointLatLong.Y
Dim pGxdial As IGxDialog
Set pGxdial = New GxDialog
pGxdial.ButtonCaption = "OK"
pGxdial.Title = "Create Stream Centerline Point Shapefile"
pGxdial.RememberLocation = True
Dim pShapeFileObj As IGxObject
Dim pGxFilter As IGxObjectFilter
Set pGxFilter = New GxFilterShapefiles 'e.g shp
Set pGxdial.ObjectFilter = pGxFilter
If pGxdial.DoModaISave(ThisDocument.Parent.hWnd) Then
Dim pLocation As IGxFile
Dim fn As String
Set pLocation = pGxdial.FinalLocation
fn = pGxdial.Name
If Not pLocation Is Nothing Then
LPntShpFlName = pLocation.Path & "\" & fn
frmlB.tbxShpFileName.Text = g]ntShpFlName
frmlB.cmdOK.Enabled = True
Function GetAngle(pPolyline As IPolyline, dAlong As Double) As Double
Dim pi As Double
pi = 4 * Atn(l)
Dim dAngle As Double
Dim pLine As ILine
Set pLine = New Line
pPolyline.QueryTangent esriNoExtension, dAlong, False, 1, pLine
, convert from radians to degrees
dAngle = (180 * pLine.Angle) / pi
I adjust angles
, ESRI defines 0 degrees as the positive X-axis, increasing counter-clockwise
I Ecology references 0 degrees as North, increasing clockwise
If dAngle <= 90 Then
GetAngle = 90 - dAngle
GetAngle = 360 - (dAngle - 90)
Function FeatureExists(strFeatureFileName As String) As Boolean
On Error GoTo ErrHandler:
Dim pWSF As IWorkspaceFactory
Set pWSF = New ShapefileWorkspaceFactory
Dim pFeatWS As IFeatureWorksiJace
Dim pFeatDS As IFeatureClass
Dim strWorkspace As String
Dim strFeatDS As String
strWorkspace = SplitWorkspaceName(strFeatureFileName) & "\"
strFeatDS = SplitFileName(strFeatureFileName)
If PWSF.IsWorkspace(strWorkspace) Then
Set pFeatWS = pWSF.OpenFromFile(strWorkspace, 0)
Set pFeatDS = pFeatWS.OpenFeatureClass(strFeatDS)
FeatureExists = (Not pFeatDS Is Nothing)
Set pWSF =Nothing
Set pFeatWS = Nothing
Set pFeatDS = Nothing
FeatureExists = False
'Returns a Workspace given for example C: \temp\dataset returns C:\temp
Function SplitWorkspaceName(sWholeName As String) As String
On Error GoTo ERH
Dim pos As Integer
pos = InStrRev(sWholeName, "\")
If pos > 0 Then
SplitWorkspaceName = Mid(sWholeName, 1, pos - 1)
MsgBox "Workspace Split" & Err.Description
'Returns a filename given for example C:\temp\dataset returns dataset
Function SplitFileName(sWholeName As String) As String
On Error GoTo ERH
Dim pos As Integer
Dim sT, sName As String
pos = InStrRev(sWholeName, "\")
Ifpos > 0 Then
sT = Mid(sWholeName, 1, pos - 1)
Ifpos = Len(sWholeName) Then
sName = Mid(sWholeName, pos + 1, Len(sWholeName) - Len(sT))
pos = InStr(sName, ".")
If pos > 0 Then
SplitFileName = Mid(sName, 1, pos - 1)
SplitFileName = sName
MsgBox "Workspace Split:" & Err.Description
Public Sub BusyMouse(bolBusy As Boolean)
'Subroutine to change mouse cursor
If g---'pMouseCursor Is Nothing Then
Set g---'pMouseCursor = New MouseCursor
Function MakeColor(lRGB As Long) As IRgbColor
Set MakeColor =New RgbColor
MakeColor.RGB = lRGB
Function MakeDecoElement(pMarkerSym As IMarkerSymbol, _
dPos As Double)_