>c>EPA
United States
Environmental Protection
Agency
Using Repeated LIDAR to
Characterize Topographic Change
in Riparian Areas and Stream
Channel Morphology in Areas
Undergoing Urban Development:
An Accuracy Assessment Guide
for Local Watershed Managers
APM 286
RESEARCH AND DEVELOPMENT
-------
-------
EPA/600/R-10/120
October 2010
www.epa.gov
Using Repeated LIDAR to Characterize
Topographic Change in Riparian Areas
and Stream Channel Morphology in
Areas Undergoing Urban Development:
An Accuracy Assessment Guide for
Local Watershed Managers
APM 286
Prepared by
S. Taylor Jarnagin, Ph.D.
U.S. Environmental Protection Agency
National Exposure Research Laboratory
Environmental Sciences Division
Landscape Ecology Branch
Research Triangle Park, NC
Although this work was reviewed by EPA and approved for publication, it may not necessarily reflect official
Agency policy. Mention of trade names and commercial products does not constitute endorsement or
recommendation for use.
U.S. Environmental Protection Agency
Office of Research and Development
Washington, DC 20460
-------
-------
Disclaimers:
The United States Environmental Protection Agency through its Office of Research and
Development funded and managed the research described here. Some of the information in this
document has been funded wholly by the United States Environmental Protection Agency under
Contract number EP-D-05-088 to Lockheed Martin. It has been subjected to the Agency's peer
and administrative review and has been approved for publication as an EPA document. Mention
of trade names or commercial products does not constitute endorsement or recommendation by
EPA for use.
-------
Table of Contents page
Title Page i
Disclaimers ii
Table of Contents iii
List of Figures and Tables v
Abstract/Executive Summary x
Acknowledgements xii
Introduction 1
Accuracy and Precision 3
Study Area and LiDAR Coverages 3
Absolute versus Averaged Error 6
Part 1: LiDAR Accuracy and Precision Assessment:
Ground Truth Measurements. 6
2006 Ground Survey: Overall LiDAR Accuracy and LiDAR
Accuracy by Ground Cover Type 6
2006 Ground Survey: LiDAR Precision: Accuracy at Ground
Survey Elevations for Repeated LiDAR Coverages by
Vegetation Cover Class 9
Multi-Year LiDAR Coverages: LiDAR Accuracy by Slope Percent 12
LiDAR Precision Assessment: Differences in LiDAR-Derived
Elevations at Surveyed Ground-Truth Points between
Subsequent Sets of Multi-Year LiDAR Coverages. 15
Stream Transects 16
Part 2: LiDAR Precision Assessment Using Multi-Year
LiDAR Coverages Where No Ground Truth Measurements Exist. 18
Large-Scale Elevation Changes Shown in LiDAR 18
-------
LiDAR-Derived DEM Differences by Land Use/Land Cover (LULC) 18
LiDAR-Derived DEM Precision by Slope Gradient by LULC 22
Stream Channels 29
Stream Sinuosity 38
LiDAR-Derived DEM Watershed Boundaries Precision 41
Discussion 50
References 52
Appendices - Table of Appendices 54
-------
List of Figures and Tables
Figure 1: LiDAR schematic. 2
Figure 2: Accuracy versus Precision. 3
Figure 3: Location of the Clarksburg Special Protection Area (CSPA). 4
Table 1: LiDAR Overflights information 5
Figure 4: Ground survey points (n = 604). 7
Table 2: 2006 Ground Survey Results - LiDAR Accuracy for all ground control points. 7
Figure 5: "Good", "OK", and "Bad" vegetative and surface condition classes. 8
Figure 6: Differences in absolute accuracy between Vegetation Cover Classes. 9
Table 3: Mean Absolute Difference between Ground Truth Elevation and
LiDAR-Derived Elevation for 2002, 2004, 2006, 2007, and 2008 coverages. 10
Figure 7: Sum of absolute differences between ground truth elevation and
LiDAR-derived elevations for repeat LiDAR coverages. 11
Figure 8: Slope quintiles created from slope derived from the 2007 LiDAR-derived
3-foot DEM calculated at the 604 ground-truth survey points. 12
Figure 9: Sum (All Years) Mean Absolute Accuracy by Slope Quintile. 13
Table 4: Sum (All Years) Mean Absolute Accuracy by Slope Quintile. 13
Figure 10: All Years Mean Absolute Year-to-Year Differences Between
LiDAR-Derived Elevations at Ground-Truth Points by Slope Quintile. 14
Table 5: All Years Mean Absolute Year-to-Year Differences Between LiDAR-Derived
Elevations at Ground-Truth Points by Slope Quintile. 14
Table 6: Modified Anderson Level One Land Cover Classes used in the study. 15
Table 7: Ground-Truth vs. LiDAR-Derived Sum Inter-Annual Delta Differences. 16
Figure 11: Stream Transects: Ground Truth vs. LiDAR-Derived Transect Elevations. 17
Figure 12: LiDAR-Derived DEM Differences by LULC Class for Sopers Branch. 19
Table 8: LiDAR-Derived DEM Differences by LULC Class for Sopers Branch. 19
-------
Figure 13: LiDAR-Derived DEM Differences by LULC Class for T104. 20
Table 9: LiDAR-Derived DEM Differences by LULC Class for T104. 20
Figure 14: LiDAR-Derived DEM Differences by LULC Class for T109. 21
Table 10: LiDAR-Derived DEM Differences by LULC Class for T109. 21
Figure 15: No-Change Land Cover, Mean Differences in Elevation by
Slope Gradient Quintiles - Sopers Branch, Forest LULC Class. 23
Table 11: No-Change Land Cover, Mean Differences in Elevation by
Slope Gradient Quintiles - Sopers Branch, Forest LULC Class. 23
Figure 16: No-Change Land Cover, Mean Differences in Elevation by
Slope Gradient Quintiles - T104, Forest LULC Class. 24
Table 12: No-Change Land Cover, Mean Differences in Elevation by
Slope Gradient Quintiles - T104, Forest LULC Class. 24
Figure 17: No-Change Land Cover, Mean Differences in Elevation by
Slope Gradient Quintiles - T109, Forest LULC Class. 25
Table 13: No-Change Land Cover, Mean Differences in Elevation by
Slope Gradient Quintiles - T109, Forest LULC Class. 25
Figure 18: No-Change Land Cover, Mean Differences in Elevation by
Slope Gradient Quintiles - Sopers Branch, Impervious Surfaces LULC class. 26
Table 14: No-Change Land Cover, Mean Differences in Elevation by
Slope Gradient Quintiles - Sopers Branch, Impervious Surfaces LULC class. 26
Figure 19: No-Change Land Cover, Mean Differences in Elevation by
Slope Gradient Quintiles - T104, Impervious Surfaces LULC class. 27
Table 15: No-Change Land Cover, Mean Differences in Elevation by
Slope Gradient Quintiles - T104, Impervious Surfaces LULC class. 27
Figure 20: No-Change Land Cover, Mean Differences in Elevation by
Slope Gradient Quintiles - T109, Impervious Surfaces LULC class. 28
Table 16: No-Change Land Cover, Mean Differences in Elevation by
Slope Gradient Quintiles - T109, Impervious Surfaces LULC class. 28
-------
Figure 21: ArcGIS-Derived Stream Channels from LiDAR-Derived
3-foot OEMs from the 2002, 2004, 2006, 2007, and 2008 LiDAR coverages. 29
Figure 22: Fligh-Variance and Low-Variance LiDAR-Derived Stream Channel
Areas for the Sopers Branch watershed. 30
Figure 23: Fligh-Variance and Low-Variance LiDAR-Derived Stream Channel
Areas for the T104 watershed. 31
Figure 24: Fligh-Variance and Low-Variance LiDAR-Derived Stream Channel
Areas for the T109 watershed. 32
Figure 25: T104 Low-Variance Stream Channel Area Example: LiDAR-Derived
Stream Channels Overlaid on 2007 Shaded Relief Imagery.
Figure 26: T104 Low-Variance Stream Channel Area Example: 2007 Shaded Relief
Imagery - the stream channel is relatively well delineated.
Figure 27: T104 High-Variance Stream Channel Area Example: LiDAR-Derived
Stream Channels Overlaid on 2007 Shaded Relief Imagery.
Figure 28: T104 High-Variance Stream Channel Area Example: 2007 Shaded Relief
Imagery - the stream channel is relatively not apparent.
Figure 29: Mean LiDAR Ground Point Density for High vs. Low-Variance Stream
Channel Areas.
33
34
35
36
37
Figure 30: Weighted Mean Average Rosgen Sinuosity over Time for Sopers Branch,
T104, and T109 for 2002, 2004, 2006, 2007, and 2008. 38
Table 17: Weighted Mean Average Rosgen Sinuosity over Time for Sopers Branch,
T104, and T109 for 2002, 2004, 2006, 2007, and 2008. 39
Figure 31: LiDAR-Derived Rosgen Sinuosity vs. LiDAR-Derived Slope Gradient. 40
Figure 32: LiDAR-Derived Rosgen Sinuosity vs. LiDAR-Derived Flow
Accumulation Gradient. 40
Figure 33: LiDAR-Derived 3-foot DEM Stream Channels and Watershed
Pour Points for 2002, 2004, 2006, 2007, and 2008 for Sopers Branch. 42
Figure 34: LiDAR-Derived 3-foot DEM Stream Channels and Watershed
Pour Points for 2002, 2004, 2006, 2007, and 2008 for T104. 43
Figure 35: LiDAR-Derived 3-foot DEM Stream Channels and Watershed
Pour Points for 2002, 2004, 2006, 2007, and 2008 for T109. 44
-------
Figure 36: 2002, 2004, 2006, 2007, and 2008 LiDAR-Derived DEM Watershed
Boundaries for Sopers Branch. 45
Figure 37: 2002, 2004, 2006, 2007, and 2008 LiDAR-Derived DEM Watershed
Boundaries for! 104. 46
Figure 38: 2002, 2004, 2006, 2007, and 2008 LiDAR-Derived DEM Watershed
Boundaries for! 109. 47
Table 18: Watershed Area Percent Differences based on LiDAR-Derived OEMs
for 2002, 2004, 2006, 2007, and 2008. 48
Figure 39: 2007-2008 watershed difference in T109. 49
-------
Appendices
Table of Appendices 54
Appendix One: Year-by-year mean absolute accuracy by slope quintile. I
Appendix Two: Year-by-year mean absolute difference by slope quintile. VI
Appendix Three: Sequence of modified Anderson Level One LULC classes mapped
from 1-foot (or better) digital orthoimages. X
Appendix Four: Sequence of figures that display the ground-truth stream transect
measurements and LiDAR-derived stream transect elevation values. XXVI
Appendix Five: Large-Scale Elevation Changes Shown in LiDAR. Sequence of
images that display the year-to-year LiDAR-derived elevation differences
between 3-foot DEMs. XL
Appendix Six: LiDAR-Derived DEM Precision by Slope Gradient by LULC. LIII
Appendix Seven: Metrics calculated for stream buffer High vs. Low-Variance
Stream Channel Areas. LXXVII
Appendix Eight: Year-to-year differences in watersheds calculated from
subsequent years of LiDAR coverages. Gray = Area Present in both coverages;
Red = Area Present in Year 1 but not Year 2;
Blue = Area Present in Year 2 but not Year 1. LXXXV
Appendix Nine: Examples of Vegetation Classes used in the LiDAR Accuracy
Assessment Ground Truth Survey Area. XCVII
-------
Abstract/Executive Summary
Urban development and the corresponding increases in impervious surfaces associated
with that development have long been known to have adverse impacts upon urban riparian
systems, water quality and quantity, groundwater recharge, streamflow, and aquatic ecosystem
integrity. The ability of Best Management Practices (BMPs) to mitigate the impact of urban
development is an emerging area of research and a central component of the Clean Water science
mission of the US EPA. The ability to monitor and characterize urban development and
corresponding stream channel changes due to urban development with remote sensing, high
spatial and temporal resolution mapping, and Geographic Information Systems (GIS) technology
is an area of research focus in the geospatial sciences and the US EPA Landscape Ecology
Branch.
A key component in the geospatial assessment of urban development is the analysis of
changes in ground surface topography, contributing watershed area, and stream channel
geomorphology using Digital Elevation Models (DEMs). LiDAR (Light Detection And
Ranging) is an active remote sensing technology that uses light pulses to measure distance and
other characteristics (texture, hardness, etc.) of terrain and objects. LiDAR can be used to
construct DEMs that are much finer in spatial scale over much larger areas than have been
previously possible. Repeated LiDAR-derived DEMs can be used to characterize the changes
over time associated with urban development.
This study is an attempt to categorize and characterize the accuracy and precision of
repeated LiDAR-derived DEMs. This study used ground truth measurements in conjunction
with repeated airborne LiDAR coverages to assess LiDAR accuracy and assess the precision of
LiDAR over different land use/land cover (LULC) types. This study is a portion of a larger
research program, the Clarksburg Monitoring Project, which combines remotely sensed imagery,
GIS, and LIDAR to map and monitor urban development and ground-based measurements of
streamflow, precipitation, aquatic biota, and water quality to measure stream response to urban
development and BMP effectiveness.
This study used a ground survey and repeated LiDAR overflights to assess overall
LiDAR accuracy, LiDAR accuracy by ground cover type and slope, and the precision
(repeatability) of LiDAR-derived DEMs and elevation measurements. The surveyed ground
control network consisted of 604 surveyed locations acting as ground-truth for the LiDAR-
derived measurements. We found significant differences in the absolute accuracy of LiDAR by
vegetation and surface condition class, with a decline in accuracy with increased interference
from vegetation, surface condition, and increasing slope. The precision of repeated LiDAR-
derived elevations across vegetation cover classes and slope also was significantly different
among vegetation cover classes and with increased slope; with declining precision found for both
increasing slope and interference intensity of vegetation cover classes.
We also looked at larger study areas to measure the precision of LiDAR-Derived DEM
differences by LULC. Each comparison yielded significantly different means for each group
with Forest showing the greatest mean difference and the Agricultural LULC class showing the
least. We compared stream transect ground-truth measurements to LiDAR-derived stream
-------
transects and found that the LiDAR-derived inter-annual transect differences measured were
from 2 to 12 times larger than the ground-truth-derived differences. The relationship between
the ground-truth transect measurements and the LiDAR-derived transects was noisy and the
ability of repeated LiDAR to reveal stream channel morphology changes at a 1-foot spatial scale
appears to be quite poor. The ability to use LULC to generally predict the precision of repeat
LiDAR-derived OEMs and elevation estimates also was weak. Changes in vegetation condition
and slope that adversely affect LiDAR accuracy and precision occur at a spatial scale much less
than that captured by the LULC variable. The use of repeat LiDAR to extract stream channel
paths and predict watershed boundaries was also found to be highly variable between LiDAR
coverages.
The results of this study should be taken as a warning to watershed modelers and
managers that LiDAR-derived stream channels, stream transects and morphology, elevation
changes, and watershed boundaries and areas may not be nearly so certain as one might
otherwise assume from the fine spatial scale of the LiDAR output maps. The final conclusion of
this study is that a user of LiDAR who attempts to delineate changes at or near to the spatial
level of resolution of the base data (about 1 meter or 3 feet in our study) cannot rely on LiDAR
alone to assess those changes. Ground truth is needed to verify changes predicted by repeat
LiDAR measurements. The loose relationship seen by the LiDAR transects and ground-truth
transects should act as a warning to the LiDAR user: use LiDAR to identify where change
appears to have occurred and then use ground-truth to verify.
-------
Acknowledgements
I thank my EPA reviewers Dave Williams and Maliha Nash for their ides and suggestions. I also
thank Rick Van Remortel and Ed Evanson of Lockheed Martin for their ideas and discussions
regarding LiDAR processing and land use/land cover classification/compilation. Their ideas,
suggestions, and assistance went beyond the call of duty called for by their contract. I would
also like to thank Mark S. Murphy (CSC ArcGIS Support Staff) and Tim Wade (EPA LEB-RTP)
for their ArcGIS suggestions, tools, and patient advice. Their counsel proved invaluable (and
much better than the on-line support at esri.com). I also would like to thank Asa Eckert-
Erdheim, a high school student volunteer from the Durham School of Arts, Durham, NC. Asa
spent many hours heads-up digitizing while compiling land use/land cover change. Hopefully,
he learned something from his volunteer internship other than how boring science is. I also
would like to thank Dr. Kaye Brubaker and her Intro to GIS class who also labored in the heads-
up digitizing mines. Again I hope the learning experience was as valuable to them as the data
were to me. Finally I would like to thank Eric Naibert, Keith Van Ness, and the other folks at
the Montgomery County Maryland Department of Environmental Protection for their invaluable
assistance in the gathering of ground truth data for the accuracy assessment of the County-
supplied stream transect data.
-------
Introduction
Changes in topography, impervious surfaces, and land use and land cover associated with
urban development have been shown to alter streamflow and stream geomorphology and
topography (Schueler, 1994; USEPA, 1994; Arnold and Gibbons, 1996; Caraco et al., 1998;
Jennings and Jarnagin, 2002; Jarnagin, 2004; Jarnagin, 2007). Measuring these changes in
streams is a difficult and labor-intensive task (Gardina, 2008). If LiDAR could reliably measure
these changes over a wide area over time, this would be a great benefit to watershed manages
and interested stakeholders and help to evaluate the effectiveness of Best Management Practices
(BMPS) in mitigating the adverse effects of urban development. .
LiDAR (Light Detection And Ranging) is a remote sensing technology that uses light
pulses to measure distance and other characteristics (texture, hardness, etc.) of terrain and
objects. LiDAR is an active remote sensing technique where the light pulse is sent from the
system and the length of time for the return signal is recorded, allowing for the distance between
the sensor and the object imaged to be calculated. LiDAR systems can be either mobile (such as
airborne LiDAR) or stationary. Both the position and orientation of the LiDAR sensor must be
known in order to accurately measure distance. GPS (Global Positioning System) receivers are
used to accurately determine the position of the aircraft and the environmental surface sensed
with LiDAR. One of the final results of airborne LiDAR is a very accurate and high-resolution
Digital Elevation Model (DEM) of the environmental surfaces remotely sensed with LiDAR.
See Figure 1 for a schematic drawing of an airborne LiDAR system.
Aerial topographic LiDAR is obtained in a series of flightlines that collect overlapping
data points. These are merged by complex software into the LiDAR All-Points data cloud.
Typically, what is delivered to the user is a computer-algorithm filtered Ground-Points dataset
converted into a DEM. The user may also request and receive First Return and Last Return Data
(the first and last set of pulses returned to the sensor, respectively). Intensity data may also be
provided. The computer algorithms and software used to create these data sets are proprietary
and exist as a black box with respect to the user. The typical user of LiDAR data will only be
using the end products of the LiDAR overflights, LiDAR-derived OEMs, not the raw data used
to create those products.
This study is an attempt to categorize and characterize the accuracy and precision of
LiDAR-derived OEMs. This study used ground truth measurements in conjunction with an
airborne LiDAR coverage to assess LiDAR accuracy and repeat airborne LiDAR coverages to
assess the precision of LiDAR over different land use/land cover (LULC) types. This study is a
portion of a larger research program, the Clarksburg Monitoring Project, that combines remotely
sensed imagery, GIS (Geographic Information Systems), and LIDAR to map and monitor urban
development and stream response to urban development. For further information about the
Clarksburg Monitoring Project see:
< http://www.epa.gov/esd/land-sci/epic/clarksburg01-05.htm >;
< http://www.montgomerycountymd.gov/dectmpl.asp?url=/content/dep/water/spaclarksburg.asp
>; < http://egsc.usgs.gov/currentscienceprojects.html >: "Best Management Practices Designed
to Improve Developing Landscapes"; and < http://egsc.usgs.gov/clarksburghighlights.html > (all
links last accessed 9/21/2010).
-------
LASER-SCANNING
LASER-
SCANNER
m
/III
/I I \
y
OBJ
Figure 1: LiDAR schematic. Image Source: Spencer B. Gross Inc., Portland OR.
-------
Accuracy and Precision
The accuracy of a set of measurements describes how close they are to the true or actual
value of the measured quantity. The precision of a set of measurements is the degree to which
the measurements agree with each other. Reproducibility, replicability, and repeatability are
synonyms for measurement precision. Figure 2 is an excellent graphic from an online Math
Skills Review at the Texas A&M Department of Chemistry Website that visually illustrates
Accuracy versus Precision:
accurate
(the average is accurate)
not precise
precise
not accurate
accurate
and
precise
Accuracy refers to how closely a measured value agrees with the correct value.
Precision refers to how closely individual measurements agree with each other.
Figure 2: Accuracy versus Precision: from Texas A&M website <
http://www.chem.tamu.edu/class/fyp/mathrev/mr-sigfg.html >. Last accessed 09/08/2010.
Study Area and LiDAR Coverages
The study area for this project is the Clarksburg Special Protection Area (CSPA) in
Montgomery County Maryland. Repeat LiDAR coverages were obtained as a part of the
ongoing research project: "Collaborative Research: Streamflow, Urban Riparian Zones, BMPs,
and Impervious Surfaces" (see: < http://www.epa.gov/esd/land-sci/epic/clarksburg01-05.htm >
for an overview. Last accessed 09/08/2010. Figure 3 shows the location of the CSPA in relation
to the state of Maryland and the Washington DC metro area.
Five airborne LiDAR coverages were obtained in the CSPA in the 2002-2008 time
period. The vendors, instruments, mean LiDAR raw point spacing, and reported accuracies for
those overflights are listed in Table 1.
-------
Image: Jarnagin/Jennings, EPA, 200-
j = 1970's NALC urban
I «1992 NLCD urban
• =2000 NLCD urban
Figure 3: Location of the Clarksburg Special Protection Area (CSPA).
-------
Year: 2002
Month: December
Vendor: Airborne 1
Instrument: Optech ALTM-2025
Mean LIDAR raw point spacing: sub 0.8 meter
Reported Accuracy:
Average vertical difference: 0.07 meter (0.23 foot)
Average horizontal difference: 0.04 meter (0.13 foot)
Max vertical difference: 0.12 meter (0.39 foot)
Max horizontal difference: 0.07 meter (0.23 foot)
Year: 2004
Month: March
Vendor: Laser Mapping Specialists Inc.
Instrument: Optech ALTM-203 3
Mean LIDAR raw point spacing: sub 0.8 meter
Reported Accuracy:
Vertical RMSE: 0.05 meter (0.15 foot)
Average horizontal difference: < 0.30 meter (< 1 foot)
Max vertical difference: 0.08 meter (0.28 foot)
Max horizontal difference: not reported
Year: 2006
Month: March
Vendor: Canaan Valley Institute (CVI)
Instrument: Optech ALTM-3100
Mean LIDAR raw point spacing: sub 0.1 meter
Reported Accuracy:
Vertical RMSE: 0.04 meter (0.15 foot)
Average horizontal difference: 0.13 meter (0.43 foot)
Max vertical difference: 0.08 meter (0.28 foot)
Max horizontal difference: not reported
Year: 2007
Month: March
Vendor: Canaan Valley Institute (CVI)
Instrument: Optech ALTM-3100
Mean LIDAR raw point spacing: sub 0.15 meter
Reported Accuracy:
Vertical RMSE: 0.03 meter (0.10 foot)
Average horizontal difference: 0.05 meter (0.16 foot)
Max vertical and horizontal differences: not reported
Year: 2008
Month: March
Vendor: Sanborn
Instrument: Leica ALS-50
Mean LIDAR raw point spacing: 1.4 meters
Reported Accuracy: (not reported - specifications below were met):
Vertical RMSE (Bare Earth): 0.15 meter (0.49 foot)
Horizontal RMSE: 1 meter (3.29 feet)
Max vertical and horizontal differences: not reported
Table 1: LiDAR Overflights information: Vendors, instruments, mean LiDAR raw point spacing,
and reported accuracies for the five LiDAR overflights used in this study.
-------
Absolute versus Averaged Error
Typically, LiDAR vendors express vertical error using an averaged estimate of error: root
mean square error (RMSE). For an unbiased estimator, the RMSE is the square root of the
variance, known as the standard error. If you want to know the confidence in an overall estimate
of an elevation value for a flat surface, the RMSE provides a good estimate of precision.
However, what we are often interested in is the error associated with a single point
measurement or a limited set of measured points (as in a stream transect). Calculating the mean
value of the error of a set of points yields a smaller number since positive and negative error
values tend to cancel each other out for an unbiased estimator. For this study, I have used the
absolute value of the error unless otherwise stated. This tends to maximize the reported value
but yields a better accuracy estimate for a single point or limited set of points.
Part 1: LiDAR Accuracy and Precision Assessment: Ground Truth Measurements.
2006 Ground Survey: Overall LiDAR Accuracy and LiDAR Accuracy by Ground
Cover Type
In March of 2006, students and faculty from the University of Maryland, College Park
joined EPA and USGS scientists, Montgomery County Maryland Department of Environmental
Protection scientists and staff, and employees from Johnson Mirmiran & Thompson PA, Sparks
Maryland in surveying a set of ground control locations near the Sopers Branch stream gauge in
the Clarksburg Special Protection Area (CSPA), near Clarksburg Maryland. EPA funded the
placement of a highly accurate geodetic ground survey monument network and the ground
survey crew used Total Stations (an electronic transit integrated with an electronic distance
meter) to read slope distances from the monuments to the ground control network. EPA also
funded a LiDAR overflight concurrent with the ground survey to allow a direct comparison of
LiDAR-derived locations with a high-accuracy ground survey. The ground control network
consisted of 604 surveyed locations (Figure 4)
The overall accuracy of the LiDAR-derived elevations is found in Table 2. Both mean
and absolute elevation differences are given. Using a mean value yields an accuracy estimate of
0.07 feet (2.1 cm) while the absolute value is an order of magnitude larger at 0.46 feet (14.1 cm).
-------
Figure 4: Ground survey points (n = 604).
Actual Values, All Points: n = 604
Mean elevation difference = 0.069
Std Dev = 0.706
± 95% C.I. = 0.056
Maximum elevation difference = 3.863
Minimum elevation difference = -2.396
Absolute Values, All Points: n = 604
Mean (Abs) elevation difference = 0.461
Std Dev = 0.539
± 95% C.I. = 0.043
Max (Abs) elevation difference = 3.863
Table 2: 2006 Ground Survey Results - LiDAR Accuracy for all ground control points. All
distances are in feet. Actual (±) values and absolute values of differences between LiDAR-
derived and ground truth elevations.
-------
I classified the ground control points according to the amount of overhanging branches
and ground litter and vegetation (Figure 5). Three vegetation condition classes were used:
"Good" (near-stream ground conditions have open, level, and firm-to-hard surfaces with few or
no overhanging branches), "OK" (near-stream vegetative conditions have relatively little ground
vegetation with relatively more overhanging branches and ground litter), and "Bad" (near-stream
vegetative conditions have more irregular and softer ground conditions with moderate-to-heavy
amounts of ground litter and dense underbrush and relatively dense overhanging branches).
Figure 5: "Good", "OK", and "Bad" vegetative and surface condition classes (represented by
green, blue, and red symbols respectively).
Comparison of the absolute accuracy by vegetation and surface condition class yields
significant differences with a decline in accuracy with increased interference with vegetation and
surface condition (Figure 6). See Appendix Nine for a visual depiction of the vegetation classes.
-------
1.0
0.9
0.8
0.7 H
0)
o
2 0.6
I
Q 0.5 H
< 0.4 H
8 0.3 H
Error Bars = ± 95% C.I.
All Points
n = 604
ANOVAF= 39.090
df = 2
p< 0.001
Good
Mean = 0.201
n = 91
Std Dev = 0.092
Max = 0.463
Bad
Mean = 0.708
n = 197
Std Dev = 0.723
Max = 3.863
OK
Mean = 0.382
n = 316
Std Dev = 0.405
Max = 2.594
Good
OK
Vegetation Class
Bad
Mean (Abs) Difference (ft) Between Vegetation Cover Classes
Figure 6: Differences in absolute accuracy between Vegetation Cover Classes. Good ground
conditions yield significantly better accuracy.
2006 Ground Survey: LiDAR Precision: Accuracy at Ground Survey Elevations for
Repeated LiDAR Coverages by Vegetation Cover Class
The accuracy of repeat LiDAR-derived elevations were calculated at the 604 surveyed
ground control truth elevation points for the LiDAR coverages for 2002, 2004, 2006, 2007, and
2008. The mean absolute difference between the ground truth elevation and the LiDAR-derived
elevation for each year was significantly different among years (ANOVA F = 23.381, df= 4, p <
0.001) but that was more a measure of the large sample size (n = 3020) than large differences
between mean elevations, which ranged from 0.4 - 0.7 feet (Table 3). The relationship of
significantly better accuracy for the "Good" vs. "OK" vs. "Bad" vegetation cover classes held for
all five years of LiDAR coverages. The precision of repeat LiDAR-derived elevations across
vegetation cover classes was measured by using the sum of differences for each ground truth
elevation point as the metric. A smaller sum of differences implies a greater precision. Figure 7
displays the significant differences found among vegetation cover classes, with increasing
difference found for the "Good" vs. "OK" vs. "Bad" vegetation cover classes.
-------
Year
2002
2004
2006
2007
2008
Mean Abs Diff
0.598
0.633
0.427
0.410
0.682
n
604
604
604
604
604
Std Dev
0.644
0.678
0.505
0.448
0.806
± 95% C.I.
0.051
0.054
0.040
0.036
0.064
Minimum
0.001
0.001
0.000
0.000
0.001
Maximum
3.383
4.005
3.996
2.563
4.350
Year
2002
2004
2006
2007
2008
2002
0.000
0.035
-0.171
-0.189
0.083
2004
0.000
-0.206
-0.223
0.049
2006
0.000
-0.017
0.255
2007
0.000
0.272
2008
0.000
Year
2002
2004
2006
2007
2008
2002
1.000
1.000
0.000
0.000
0.213
2004
1.000
0.000
0.000
1.000
2006
1.000
1.000
0.000
2007
1.000
0.000
2008
1.000
Table 3: Mean Absolute Difference between Ground Truth Elevation and LiDAR-Derived
Elevation for 2002, 2004, 2006, 2007, and 2008 coverages. Matrix of pairwise mean differences
using least squares means: Post Hoc test of ABS_DIFF; using model MSE of 0.396 with 3015 df
Bonferroni Adjustment: Matrix of pairwise comparison probabilities.
10
-------
g
0)
4.5
4.0
3.5
3.0
2.5
1.5
1.0
0.5
0.0
All Points
n = 604
ANOVAF= 64.513
df = 2
p < 0.001
Error Bars = ±95% C.I.
OK
Bad
Mean = 3.940
n = 197
Good
Mean = 0.785
n = 91
Std Dev = 0.294
Max = 2.447
Good
Mean = 2.574
n = 316
Std Dev = 2.115
Max =13.689
OK
Vegetation Class
Bad
Mean Sum of Absolute Difference (ft) Between Vegetation Cover
Classes for 2002, 2004, 2006, 2007, and 2008 LiDAR Coverages
Figure 7: Sum of absolute year-to-year differences between ground truth elevation and LiDAR-
derived elevations (n = 604) for repeat LiDAR coverages: 2002, 2004, 2006, 2007, and 2008.
ll
-------
Multi-Year LiDAR Coverages: LiDAR Accuracy by Slope Percent
Slope has been shown to account for differences in LiDAR accuracy, with increasing
error found in areas of increasing slope. Our repeat LiDAR and ground truth elevation dataset
offered an excellent opportunity to test this relationship. The 2007 LiDAR data over the ground
survey area was used as the 'best' data and processed to a 3-foot DEM (the 'best' spatial
resolution based upon the point spacing and the DEM rules stated in Maune, 2001). The Arclnfo
Grid 'slope' function was used to generate a %slope gradient and 'Slope Quintiles' were created
by using a combination of equal sample sizes and natural breaks in the data (Figure 8).
3.600001 - 8.500000
8.500001 - 19.300000
19.300001 -31.700OOO
31.700001 -77.833412
Figure 8: Slope quintiles created from slope derived from the 2007 LiDAR-derived 3-foot DEM
calculated at the 604 ground-truth survey points.
For this exercise, the accuracy value is calculated as the absolute value of the difference
between the survey ground truth points and the LiDAR-derived elevation values. 'Mean Sum
Accuracy' (Figure 9 and Table 4) is the mean of the sum for 2002, 2004, 2006, 2007, and 2008
for each point in the quintile. 'MeanDiff is the absolute value of the difference between the
LiDAR-derived elevation values for 2002-2004, 2004-2006, 2006-2007, and 2007-2008.
12
-------
'MeanDiff All Years' is the absolute value of the difference between the LiDAR-derived
elevation values for all four between-year pairs. The accuracy of the LiDAR-derived elevation
at the surveyed ground truth points significantly declined with increasing percent slope.
u
Ifl
4 -
3
2 -
1
ANOVA
F-ratio = 38.621
n = 604
df=4
P < 0.001
0-3.6 3.6-8.5 8.5-19.3 19.3-31.7
Slope Quintiles (Percent Change)
31.7-77.9
Error Bars = ± 95% C.I.
Sum (All Years) Mean Absolute Accuracy by Slope Quintile
Figure 9: Sum (All Years) Mean Absolute Accuracy by Slope Quintile.
Slope Quintiles
0-3.6
3.6 - 8.5
8.5 - 19.3
19.3 - 31.7
31.7 - 77.9
Mean Sum
Accuracy
1.26
1.84
2.77
3.41
4.41
n
116
123
121
122
122
stdev
1.08
1.81
2.48
2.10
2.97
Table 4: Sum (All Years) Mean Absolute Accuracy by Slope Quintile.
± 95% C.I.
0.20
0.51
0.42
0.53
0.83
Year-by-year mean absolute accuracy by slope quintile is shown in Appendix One.
These data are noisier than the sum of years but show the same general trend: increasing
accuracy with lower slope.
13
-------
LiDAR Precision Assessment: Differences in LiDAR-Derived Elevations at
Surveyed Ground-Truth Points between Subsequent Sets of Multi-Year LiDAR Coverages.
Another method of assessing precision (without respect to how accurate the
measurements are) is to compare year-to-year differences between the LiDAR-derived elevations
for subsequent coverages in the multi-year LiDAR coverage dataset.
1.5
0)
o
I
Q
ro
o>
1.0
0.5
ANOVA
F-ratio = 69.760
n = 2416
df = 4
P < 0.001
0-3.6 3.6-8.5 8.5-19.3 19.3-31.7 31.7-77.9
Error Bars = + 95% C.I. sl°Pe Qu'ntiles (Percent Change)
All Years: Mean Absolute Difference by Slope Quintile
Figure 10: All Years Mean Absolute Year-to-Year Differences Between LiDAR-Derived
Elevations at Ground-Truth Points by Slope Quintile.
Slope Quintiles
0-3.6
3.6 - 8.5
8.5 - 19.3
19.3 - 31.7
31.7 - 77.9
MeanDiffAH
Years
0.22
0.33
0.44
0.57
0.71
n
464
500
476
488
488
stdev
0.25
0.43
0.52
0.56
0.65
± 95% C.I.
0.02
0.04
0.05
0.05
0.06
Table 5: All Years Mean Absolute Year-to-Year Differences Between LiDAR-Derived
Elevations at Ground-Truth Points by Slope Quintile.
14
-------
Figure 10 shows the means of differences by slope quintile and the ANOVA statistics
between groups. There was a significant difference between groups with increasing differences
(less precision) as the percent slope increased. Table 5 displays the resulting numeric values.
Year-by-year mean absolute difference by slope quintile is shown in Appendix Two.
These data are noisier than the sum of years but show the same general trend: increasing
precision with lower slope.
LiDAR Precision Assessment: Multi-Year LiDAR Coverages by Land Use/Land
Cover (LULC)
In most real-life applications of LiDAR-derived elevations, DEMs, and topography, there
are few or no ground-truth elevations other than those used to calibrate the LiDAR overflight.
The multi-year LiDAR coverage dataset offered an opportunity to indirectly assess LiDAR
precision by using those multiple measurements sorted by Land Use/Land Cover (LULC) classes
and slope. We first needed to determine which areas did not undergo development and change in
LULC over the time period covered. Modified Anderson Level One and Level Two
classification of LULC was done for the study watersheds for the 1998, 2002, 2004, 2006, 2007,
and 2008 periods. Table 6 shows the modified Anderson Level One classification used.
LULC Code LULC Class
1 Agricultural
2 Barren
3 Forest
4 Impervious Surface (Urban)
5 Natural Clearing
6 Urban Grasses Cultivated
7 Urban Grasses Fallow
8 Water
9 Wetland
Table 6: Modified Anderson Level One Land Cover Classes used in the study.
Appendix Three: Sequence of modified Anderson Level One LULC classes mapped from
1-foot (or better) digital orthoimages (used as the ground-truth for LULC mapping) from the
1998, 2002, 2004, 2006, and 2008 aerial overflights by Montgomery County in the Clarksburg
Special Protection Area. 1998 was used as the base year. LULC coverages were done for the
Sopers Branch, Tributary 104 (T104), and Tributary 109 (T109) watershed areas, with the LULC
mapped to a 500-foot buffer and the County-mapped stream channels overlaid. For comparisons
of LiDAR precision (repeatability of LiDAR-derived elevation), only areas with no LULC
change over time were used as comparison locations.
15
-------
Stream Transects
The Montgomery County, Maryland Department of Environmental Protection (DEP)
does yearly measurements of streams in the CSPA as a part of their geomorphology assessments
of stream health. These repeated measurements are taken annually at a linear spatial scale of one
foot or less across stream transects chosen to yield information about changes in stream
morphology as a result of the urban development in the CSPA. I used a Trimble GeoXT2003
handheld Global Positioning System (GPS) to find the geographic locations of the DEP transects
and compared LiDAR-derived DEM transects at the same locations over time. I used a 1-foot
LiDAR-derived DEM to match the spatial scale of analysis used by the DEP and interpolated
both results to a uniform 1-foot spatial scale. Due to positional uncertainty (location error) in the
GPS locations and the temporal disconnect between stream transect dates and LiDAR overflight
dates, no attempt was made to directly compare ground-truth transects with LiDAR-derived
transects (a measure of accuracy). Instead, as a measurement of precision, year-to-year
differences between interpolated stream transect depths and LiDAR-derived transect elevations
were compared. The metric used for comparison was the "Sum Inter-Annual Delta Difference".
This was calculated by the taking the sum of the absolute values of the differences in inches
between interpolated depths along the respective ground-truth and LiDAR-derived transects,
divided by the linear distance measured and the number of years in the series of measurements.
This yields a normalized value that expresses the inches per year per linear foot of measured
difference per measurement method for each stream transect location. Table 7 and Figure 11
show the ground-truth vs. LiDAR-Derived Sum Inter-Annual Delta Differences measured.
Appendix Four shows a sequence of images that display the ground-truth stream transect
measurements and LiDAR-derived stream transect elevation values.
Sample Location Ground-Truth LiDAR-Derived
LSSB101 Al XI 0.4047 2.8104
LSLS104A1X1 0.4148 1.3688
LSLS104A1X2 0.5269 1.6167
LSLS104A2X1 0.4579 2.6024
LSLS104A2X2 0.3445 1.7940
LSLS104A3X1 0.9265 2.2474
LSLS104A3X2 0.7639 1.7919
LSLS104A3X3 0.9083 2.3634
LSLS104A3X4 0.8778 3.3982
LSLS109A1X1 0.1250 1.3024
LSLS109A1X2 0.2492 0.8203
LSLS109A2X1 0.2785 3.3259
LSLS109A2X2 0.6027 3.5540
LSLS109A3X1 0.1793 1.4810
Table 7: Ground-Truth vs. LiDAR-Derived Sum Inter-Annual Delta Differences
(inches/foot/year).
16
-------
4.00
3.50
3.00 -
9 2.00
DC.
j 1.50
1.00
0.50
0.00
y=1.3192x + 1.
R2 = 0.1743
n = 14
0.00 0.25 0.50 0.75 1.00
Ground-Truth
Ground-Truth vs LiDAR-Derived Sum Inter-Annual Delta Differences
Figure 11: Stream Transects: Ground Truth vs. LiDAR-Derived Transect Elevations.
The LiDAR-derived Sum Inter-Annual Delta Differences measured were from 2 to 12
times larger than the ground-truth-derived differences. Additionally, the year-to-year difference
seen between some LiDAR transects was extreme. The relationship between the two variables
was noisy (R2 = 0.17) and the ability of repeat LiDAR to reveal stream channel morphology
changes at a 1-foot spatial scale appears to be quite poor.
17
-------
Part 2: LiDAR Precision Assessment Using Multi-Year LiDAR Coverages Where
No Ground Truth Measurements Exist.
Large-Scale Elevation Changes Shown in LiDAR
Appendix Five shows a sequence of images that display the year-to-year LiDAR-derived
elevation differences between 3-foot DEMs for the 500-foot buffered areas around the Sopers
Branch, Tributary 104 (T104), and Tributary 109 (T109) watersheds in the CSPA. Year-to-year
differences in 3-foot LiDAR-derived DEMs are shown for sequential years 2002-2004, 2004-
2006, 2006-2007, and 2007-2008. All elevation differences have been scaled to the same metric,
with the difference between the later year minus the earlier year shown; increasing red (negative)
values indicate that the elevation was greater in the prior year (elevation has decreased over
time) while increasing blue (positive) values indicate that the elevation was less in the prior year
(elevation has increased over time). Sopers Branch is a control watershed, not currently
undergoing urban development. As expected, year-to-year differences are largely minor (within
± 1 foot) or largely confined to the outer edges of the DEMs where edge effects are present and
some development activities have occurred in the 500-foot buffer outside the watershed
boundary. T104 and T109 are areas undergoing development and large-scale cut-and-fill
operations associated with urban development can be detected using the sequential LiDAR
coverages. For comparisons of LiDAR precision (repeatability of LiDAR-derived elevation),
only areas with no elevation change ± 2 feet over time were used as comparison locations.
LiDAR-Derived DEM Differences by LULC
To look at the precision of LiDAR-derived 3-foot DEM elevations by LULC class, areas
of no elevation change (± 2 feet) over time and no LULC change over time were compared for
the 2002-2004, 2004-2006, 2006-2007, and 2007-2008 LiDAR derived elevation differences for
the Sopers Branch, T104, and T109 watersheds.. A random sample of 60 points was selected
from the coverages for the Agricultural, Forest, Impervious Surfaces, and Urban Grasses
Cultivated LULC classes. A mean was calculated for the absolute values of the differences for
the four coverage pairs for each watershed for each of the four LULC classes considered. Each
comparison yielded significantly different means for each group with Forest showing the greatest
mean difference and the Agricultural LULC class showing the least. Due to the very large
sample sizes, these significant differences between LULC classes must be taken with a grain of
salt. Smaller within-LULC samples (n = 30) showed a high degree of internal variation of the
mean. Figures 11-13 and Tables 8-10 show the mean differences found.
18
-------
e,
t
Q
c
-------
u.ou -
0.45 '-_
0.40 '-_
~ 0.35 ;
"*•"*'
£ 0.30 -
Q
n 0.25 '-_
0.15 |
0.10 ;
0.05 :
n nn
\j.\j\j
Error Bars are ANOVA F = 17.819
±95% C.I. df=3
n = 2880
I D < 0.001
T
]
!
Agricultural Forest Impervious Surface Urban Grasses
Cultivated
Land Cover Class
T104: Mean Abs_Difference in LiDAR-Derived Elevation
Between Land Cover Classes
Figure 13: LiDAR-Derived DEM Differences by LULC Class for T104.
Land Cover
Class
(T104)
Agricultural
Forest
Impervious
Surface
Urban
Grasses
Cultivated
Sample
All
All
All
All
Mean
Abs_Diff(ft)
0.275
0.389
0.326
0.312
n
720
720
720
720
Std Dev
0.177
0.382
0.348
0.261
± 95 % C.I.
0.013
0.028
0.025
0.019
Table 9: LiDAR-Derived DEM Differences by LULC Class for T104.
20
-------
u.ou -
0.45 j
0.40 ;
0.35 ;
§ :
IE 0.30 ;
Q
J2 0.25 i
a 0.20 i
i :
0.15 ;
0.10 ~:
0.05 '-_
0.00 -
Error Bars are ANOVA F = 6.342
±95% C.I. df=3
n = 2880
p < 0.001
T
T
T
T
Agricultural Forest Impervious Surface Urban Grasses
Cultivated
Land Cover Class
T109: Mean Abs_Difference in LiDAR-Derived Elevation
Between Land Cover Classes
Figure 14: LiDAR-Derived DEM Differences by LULC Class for T109.
Land Cover
Class
(T109)
Agricultural
Forest
Impervious
Surface
Urban
Grasses
Cultivated
Sample
All
All
All
All
Abs_Diff(ft)
0.247
0.302
0.257
0.264
n
720
720
720
720
Std Dev
0.222
0.286
0.269
0.237
± 95 % C.I.
0.016
0.021
0.020
0.017
Table 10: LiDAR-Derived DEM Differences by LULC Class for T109.
21
-------
LiDAR-Derived DEM Precision by Slope Gradient by LULC
To investigate the effect of slope on LiDAR-derived elevations by LULC class, a percent
slope gradient was calculated for the Forest and Impervious Surfaces LULC classes for Sopers
Branch, T104, and T109 using the 3-foot DEM derived from the 2007 LiDAR coverage. To
avoid edge effects, an interior 3-foot buffer was used and only no-LULCC-change and no-
elevation-change areas were used. The slope percentages at 3-ft DEM resolution were derived
by year and rounded to whole integer percentages; the few cells of gradient > 30% were rolled
into the 30% category; the five years' class and cell count distributions were combined across
years and the mean average value (MAV) of the absolute differences between subsequent years
calculated for each slope percentage. Quintiles were formed by using the closest equal-number
technique from the slope percentage classes and the MAVs averaged for each quintile. Figures
15-17 and Tables 11-13 display the all-years values computed for the Forest LULC class and
Figures 18-20 and Tables 14-16 display the all-years values computed for the Impervious
Surfaces LULC class. Individual between-year differences are shown in Appendix Six for both
LULC classes.
There was little change in between-year elevation differences computed until the largest
slope-percent quintile was considered. In the Forest LULC class, Sopers Branch showed the
least difference in elevation between slope-percent quintiles while T104 and T109 showed an
increased difference in elevation with the largest slope-percent quintile. For the Impervious
Surfaces LULC class, the largest slope-percent quintile showed the largest difference in
elevation in all three watersheds. See the Discussion section for speculation as to why this may
have occurred.
22
-------
u.o
0.7
0.6
o 0.5
01
^ 0.4
01
o
§ 03
i
Q
S °'2
Ol
0.1
n n
Error Bars are the
Standard Deviation
of the Quintile Mean
I
I
I
I
I
0-6
7-10 11-14
Slope Gradient Quintile
15-19
20 - 30+
All Years Forest No-Change Land Cover, Sopers Branch:
Mean Difference in Elevation by Slope Gradient Quintile
Figure 15: No-Change Land Cover, Mean Differences in Elevation by Slope Gradient Quintiles
Sopers Branch, Forest LULC Class
Quintile Slope
Groups
0-6
7-10
11-14
15-19
20 - 30+
Quintile n
2902604
3482788
3533244
3244784
3427552
Mean Diff All Years
0.303
0.293
0.294
0.303
0.340
Std Dev
0.079
0.064
0.061
0.062
0.076
Table 11: No-Change Land Cover, Mean Differences in Elevation by Slope Gradient Quintiles
Sopers Branch, Forest LULC Class.
23
-------
u.o
0.7
0.6
g
o 0.5
rence in Elevat
O O
io «.
•2
O
c 0.2
8
0.1
n n
Error Bars are the
Standard Deviation
of the Quintile Mean
I
I
T
I
0-6
7-10 11-14
Slope Gradient Quintile
15-19
20 - 30+
All Years Forest No-Change Land Cover, T104:
Mean Difference in Elevation by Slope Gradient Quintile
Figure 16: No-Change Land Cover, Mean Differences in Elevation by Slope Gradient Quintiles
T104, Forest LULC Class.
Quintile Slope
Groups
0-6
7-10
11-14
15-19
20 - 30+
Quintile n
282464
273672
286812
287704
398008
Mean Diff All Years
0.343
0.328
0.317
0.332
0.389
Std Dev
0.085
0.079
0.075
0.077
0.101
Table 12: No-Change Land Cover, Mean Differences in Elevation by Slope Gradient Quintiles
T104, Forest LULC Class.
24
-------
0.8
0.7
0.6
o 0.5
1
_
m
01
o
0.4
§ 0.3
£
O
c 0.2
0.1
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-4
5-7 8-10
Slope Gradient Quintile
11 -14
15-30+
All Years Forest No-Change Land Cover, T109:
Mean Difference in Elevation by Slope Gradient Quintile
Figure 17: No-Change Land Cover, Mean Differences in Elevation by Slope Gradient Quintiles
T109, Forest LULC Class.
Quintile Slope
Groups
0-4
5-7
8-10
11-14
15 - 30+
Quintile n
320104
425816
395800
349564
355772
Mean Diff All Years
0.268
0.265
0.270
0.283
0.374
Std Dev
0.021
0.017
0.021
0.029
0.081
Table 13: No-Change Land Cover, Mean Differences in Elevation by Slope Gradient Quintiles
T109, Forest LULC Class.
25
-------
0.8
0.7
0.6
o 0.5
1
_
m
01
o
ro
01
0.4
a. 0.3
O
c 0.2
0.1
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-2
3-4 5-6 7-9
Slope Gradient Quintile
10-30+
All Years Impervious Surface No-Change Land Cover, Sopers Branch:
Mean Difference in Elevation by Slope Gradient Quintile
Figure 18: No-Change Land Cover, Mean Differences in Elevation by Slope Gradient Quintiles
Sopers Branch, Impervious Surfaces LULC class.
Quintile Slope
Groups
0-2
3-4
5-6
7-9
10 - 30+
Quintile n
144972
216864
181252
144952
104676
Mean Diff All Years
0.318
0.293
0.276
0.277
0.471
Std Dev
0.083
0.081
0.069
0.066
0.173
Table 14: No-Change Land Cover, Mean Differences in Elevation by Slope Gradient Quintiles
Sopers Branch, Impervious Surfaces LULC class.
26
-------
ro
01
0)
o
£
ro
01
0.7
0.6
0.5
0.4
0.2
0.1
Error Bars are the
Standard Deviation
of the Quintile Mean
_L I JL
f
0-3
4-5 6-7
Slope Gradient Quintile
8-11
12-30+
All Years Impervious Surface No-Change Land Cover, T104:
Mean Difference in Elevation by Slope Gradient Quintile
Figure 19: No-Change Land Cover, Mean Differences in Elevation by Slope Gradient Quintiles
T104, Impervious Surfaces LULC class.
Quintile Slope
Groups
0-3
4-5
6-7
8-11
12 - 30+
Quintile n
68188
68924
62696
75984
91480
Mean Diff All Years
0.301
0.293
0.308
0.335
0.459
Std Dev
0.068
0.072
0.080
0.078
0.117
Table 15: No-Change Land Cover, Mean Differences in Elevation by Slope Gradient Quintiles
T104, Impervious Surfaces LULC class.
27
-------
c
o
4=
ro
LU
C
c
ra
0.8
0.7
0.6 -
0.5
0.4 -
0.3 -
0.2
0.1 -
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-2
3-4 5-6 7-9
Slope Gradient Quintile
10-30+
All Years Impervious Surface No-Change Land Cover, T109:
Mean Difference in Elevation by Slope Gradient Quintile
Figure 20: No-Change Land Cover, Mean Differences in Elevation by Slope Gradient Quintiles
T109, Impervious Surfaces LULC class.
Quintile Slope
Groups
0-2
3-4
5-6
7-9
10 - 30+
Quintile n
41196
62208
55788
48028
50648
Mean Diff All Years
0.260
0.244
0.249
0.287
0.457
Std Dev
0.071
0.061
0.061
0.069
0.141
Table 16: No-Change Land Cover, Mean Differences in Elevation by Slope Gradient Quintiles
T109, Impervious Surfaces LULC class.
28
-------
Stream Channels
One of the primary goals of this research was to assess the ability of repeat LiDAR
coverages to accurately assess changes in stream channels over time in areas undergoing urban
development. LiDAR-derived 3-foot DEMs were used to calculate flow accumulation and
delineate stream channels in ArcGIS. Figure 21 shows the three study watershed and the stream
channels derived from the 2002, 2004, 2006, 2007, and 2008 LiDAR coverages.
Figure 21: ArcGIS-Derived Stream Channels from LiDAR-Derived 3-foot DEMs from the 2002,
2004, 2006, 2007, and 2008 LiDAR coverages. The stream channels for Sopers Branch, T104,
and T109 are shown overlaid on a shaded-relief image derived from the 2007 LiDAR-derived
DEM and the three watersheds are overlaid on 2002 IKONOS imagery of the CSPA.
29
-------
Areas where multiple LiDAR-derived stream channels closely overlaid each other were
classified as 'low-variance' channels and areas where the channels did not overlay closely were
classified as 'high-variance' channels. Three areas were identified for each of the channel-
variance groups for each of the three watersheds (Figures 22-24).
Figure 22: Fligh-Variance and Low-Variance LiDAR-Derived Stream Channel Areas derived
from the 2002, 2004, 2006, 2007, and 2008 LiDAR coverages at a 3-foot DEM spatial scale for
the Sopers Branch watershed. Channels are overlaid on a shaded-relief image derived from the
2007 3-foot LiDAR-derived DEM.
30
-------
Figure 23: High-Variance and Low-Variance LiDAR-Derived Stream Channel Areas derived
from the 2002, 2004, 2006, 2007, and 2008 LiDAR coverages at a 3-foot DEM spatial scale for
the T104 watershed. Channels are overlaid on a shaded-relief image derived from the 2007 3-
foot LiD AR-derived DEM.
31
-------
Figure 24: High-Variance and Low-Variance LiDAR-Derived Stream Channel Areas derived
from the 2002, 2004, 2006, 2007, and 2008 LiDAR coverages at a 3-foot DEM spatial scale for
the T109 watershed. Channels are overlaid on a shaded-relief image derived from the 2007 3-
foot LiD AR-derived DEM.
32
-------
Review of the LiDAR-derived DEMs for the High vs. Low-Variance Stream Channel
Areas shows that in High-Variance Areas, the stream channel was not readily visible in the
shaded relief imagery (see Figures 25-28 for High vs. Low-Variance Stream Channel Areas in
T104).
T104
Low-Variance
Stream Channels
over Shaded-Relief
Imagery
Figure 25: T104 Low-Variance Stream Channel Area Example: LiDAR-Derived Stream
Channels Overlaid on 2007 Shaded Relief Imagery.
33
-------
Figure 26: T104 Low-Variance Stream Channel Area Example: 2007 Shaded Relief Imagery
the stream channel is relatively well delineated.
34
-------
T104
High-Variance
Stream Channels
over Shaded-Relief
Imagery
Figure 27: T104 High-Variance Stream Channel Area Example: LiDAR-Derived Stream
Channels Overlaid on 2007 Shaded Relief Imagery.
35
-------
T104
High-Variance
Stream Channels Area:
Shaded-Relief
Imagery
Figure 28: T104 High-Variance Stream Channel Area Example: 2007 Shaded Relief Imagery -
the stream channel is relatively not apparent.
36
-------
The reason for the difference between High vs. Low-Variance Stream Channel Areas is
not readily apparent. I tested a number of variables to compare metrics on High vs. Low-
Variance Stream Channel Areas (LiDAR Ground Point Density, ratio of LiDAR All Points to
Ground Point Density, LiDAR-Derived Elevation Differences, etc.). No general metric provided
a significant explanation for the difference between High vs. Low-Variance Stream Channel
Areas. LiDAR Ground Point Density appeared to explain the most variance but not to a
significant level (Figure 29). Appendix Seven displays some of the metrics calculated.
0)
a
o
a.
o
5
0.25
0.20
0.15
0.10
0.05
0.00
Pooled Variance t = 1.679
df=16
P = 0.113
High Channel
Variability
Low Channel
Variability
Stream Channel Variability Group
2007 Mean LiDAR Ground-Point Density vs. Stream Channel Variability Group
Figure 29: Mean LiDAR Ground Point Density for High vs. Low-Variance Stream Channel
Areas.
37
-------
Stream Sinuosity
Stream sinuosity is a measurement of the actual path length of a stream compared to the
shortest path length along a given distance. A higher degree of sinuosity measured means a
greater degree of meandering along a given distance of stream length. Changes in sinuosity over
time can be indicative of changes due to urbanization and using LiDAR to calculate changes in
sinuosity could be a useful tool to track streams over time. Rosgen sinuosity measurements
(Rosgen, 1994, 1996) were applied to the LiDAR-derived DEM-based flow accumulation stream
channels derived from ArcHydro (Maidment, 2002) for the 2002, 2004, 2006, 2007, and 2008
LiDAR coverages. The stream channels were masked to focus on reaches without Barren or
Impervious Surfaces within the buffers and stratified by Strahler stream order (tributary
networks, Strahler, 1957). The weighted mean average Rosgen sinuosity was measured for
Strahler stream orders 1 and 2 for the LiDAR coverages. No clear trend was apparent when
comparing T104 and T109 over time (where development had occurred) to Sopers Branch
(where no development had occurred). Figure 30 and Table 17 show the Sinuosity over Time.
1.50
t I-*
o
3
C
«
O)
< 1.30 H
*J
o>
O)
V)
5 1.20 H
* Sopers Branch 81
• T104S1
AT109S1
* Sopers Branch S2
AT104S2
1.10
t
A
t
A
2002
2003
2004
2005
Year
2006
2007
2008
Sinuosity over Time
Figure 30: Weighted Mean Average Rosgen Sinuosity over Time for Sopers Branch, T104, and
T109 for 2002, 2004, 2006, 2007, and 2008.
38
-------
Watershed Strahler Order Year Wt. Avg. Rosgen Sinuosity
Sopers Branch
Sopers Branch
Sopers Branch
Sopers Branch
Sopers Branch
T104
T104
T104
T104
T109
T109
T109
T109
T109
T109
Sopers Branch
Sopers Branch
Sopers Branch
Sopers Branch
Sopers Branch
T104
T104
T104
T104
T104
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
2002
2004
2006
2007
2008
2002
2004
2006
2007
2008
2002
2004
2006
2007
2008
2002
2004
2006
2007
2008
2002
2004
2006
2007
2008
1.33
1.34
1.34
1.37
1.32
1.27
1.31
1.29
1.31
1.26
1.28
1.30
1.31
1.39
1.29
1.34
1.34
1.37
1.36
1.29
1.34
1.31
1.33
1.34
1.28
Table 17: Weighted Mean Average Rosgen Sinuosity over Time for Sopers Branch, T104, and
T109 for 2002, 2004, 2006, 2007, and 2008.
Figures 31 and 32 display the lack of general relationship between the LiDAR-derived
sinuosity measurements and increasing flow magnitude (calculated via flow accumulation
values) and percent slope gradient.
39
-------
1.50
1.25
HI
3 1.00 H
in
c
5? °76
o
a)
>
0.50
0.25
0.00
Strahler Order 1 y = 0.0006x -M .3079
R2 = 0.0013
StrahlerOrder2
.3898
R2 = 0.3756
10
12
14
16
18
Wt. Ave. LiDAR-Dei ivecl Slope Gradient
Rosgen Sinuosity vs. LiDAR-Derived Slope Gradient
Figure 31: LiDAR-Derived Rosgen Sinuosity vs. LiDAR-Derived Slope Gradient. For 2nd order
stream segments, we see a weak declining sinuosity with increasing slope, which would be
expected. This relationship did not appear with 1st order stream segments.
1.5
V)
o
C
CO
1.25
1
0.75
0.5
0.25
*FlowAccml 1
*FlowAccml2
*FlowAccml3
6 8 10 12
Wt. Ave. Slope Gradient
14
16
18
Rosgen Sinuosity vs. Slope Gradient by
LiDAR-Derived Flow Accumulation Gradient
Figure 32: LiDAR-Derived Rosgen Sinuosity vs. LiDAR-Derived Flow Accumulation Gradient.
40
-------
LiDAR-Derived DEM Watershed Boundaries Precision
We used the repeat LiDAR-derived 3-foot OEMs for 2002, 2004, 2006, 2007, and 2008
to calculate flow accumulation and place stream channels within the topographic landscape.
Using the stream channel pixel closest to the stream gauge location determined via GPS, we
calculated the contributing watershed at the gauge pour point. As might be expected from the
year-to-year variations in LiDAR-derived stream channels, the pour points used to calculate the
catchments in ArcHydro varied somewhat from year-to-year as well. Figures 33-35 display the
variations seen in pour point placement for the five years of LiDAR data.
41
-------
0 2.5
10 Feet
Sopers Branch:
' 2002-2008
LiDAR-Derived DEM
Stream Channels and
Watershed Pour Points
Figure 33: LiDAR-Derived 3-foot DEM Stream Channels and Watershed Pour Points for 2002,
2004, 2006, 2007, and 2008 for Sopers Branch. The gray shaded pixels reveal the number of
years a particular pixel was included in the watershed defined by the DEM pour point: lighter
gray = more included.
42
-------
Figure 34: LiDAR-Derived 3-foot DEM Stream Channels and Watershed Pour Points for 2002,
2004, 2006, 2007, and 2008 for T104. The gray shaded pixels reveal the number of years a
particular pixel was included in the watershed defined by the DEM pour point: lighter gray =
more included.
43
-------
Figure 35: LiDAR-Derived 3-foot DEM Stream Channels and Watershed Pour Points for 2002,
2004, 2006, 2007, and 2008 for T109. The gray shaded pixels reveal the number of years a
particular pixel was included in the watershed defined by the DEM pour point: lighter gray =
more included.
44
-------
The differences in the LiDAR-derived DEM pour points combine with the differences in
LiDAR derived elevations across the landscape to produce LiDAR-derived watersheds that vary
from year-to-year. Figures 36-38 display the watersheds derived from the LiDAR coverages.
Figure 36: 2002, 2004, 2006, 2007, and 2008 LiDAR-Derived DEM Watershed Boundaries for
Sopers Branch. The gray shaded pixels reveal the number of years a particular pixel was
included in the watershed defined by the DEM pour point: lighter gray = more included.
45
-------
Figure 37: 2002, 2004, 2006, 2007, and 2008 LiDAR-Derived DEM Watershed Boundaries for
T104. The gray shaded pixels reveal the number of years a particular pixel was included in the
watershed defined by the DEM pour point: lighter gray = more included.
46
-------
Figure 38: 2002, 2004, 2006, 2007, and 2008 LiDAR-Derived DEM Watershed Boundaries for
T109. The gray shaded pixels reveal the number of years a particular pixel was included in the
watershed defined by the DEM pour point: lighter gray = more included.
47
-------
The LiDAR-derived elevation differences are largely restricted to edge-effects in the
year-to-year watershed boundaries. The large section in the NW of Sopers Branch is an anomaly
produced in part by the lack of sufficient LiDAR coverage in the 2002 and 2004 LiDAR
overflights on the west side of the watershed and the 2006 LiDAR coverage had some quality
control issues (dropouts) in the NW corner that also affected the area included in the watershed
or not. In both T104 and T109, urban development also affected the area included in the
watershed or not. Table 17 shows the absolute percentage difference between subsequent years
in LiDAR coverages. Figure 39 shows the 2007-2008 watershed difference in T109, with the
section of the watershed in red being present in the contributing DEM-shed in 2007 but not in
2008. This occurred as a result of construction in the upper watershed and an east-west road
(and elevated grade associated with the road) that cuts off that portion of the watershed from the
pour point. In reality, culverts and the anthropogenic drainage patterns associated with the road
construction may still hydraulically link the upper portion of the watershed to the remainder but
this cannot be determined from the LiDAR coverages.
Watershed Years Compared Relative Percent Change Area
T104 2002-2004 1.35
T104 2004-2006 2.41
T104 2006-2007 0.74
T104 2007-2008 2.37
T109 2002-2004 0.09
T109 2004-2006 1.05
T109 2006-2007 1.43
T109 2007-2008 12.06
Table 18: Watershed Area Percent Differences based on LiDAR-Derived OEMs for 2002, 2004,
2006, 2007, and 2008 in T104 and T109.
Appendix Eight shows the year-to-year differences in watersheds calculated from
subsequent years of LiDAR coverages.
48
-------
T109
2007-2008
>
Year-to-Year
Differences in
Watershed Area
,^ - Determined from
Subsequent
LiDAR-Dervived
3-foot DEMs
Figure 39: 2007-2008 watershed difference in T109, with the section of the watershed in red
being present in the contributing DEM-shed in 2007 but not in 2008. Gray = Area Present in
both coverages; Red = Area Present in Year 1 but not Year 2; Blue = Area Present in Year 2 but
not Year 1.
49
-------
Discussion
Previous studies of LiDAR accuracy (Hodgson and Bresnahan, 2004; Hodgson et al.,
2005; Gardina, 2008) have shown that there are several factors that affect the accuracy of
LiDAR-derived elevations: internal errors within the LiDAR system, processing errors within the
LiDAR algorithms (determining which points in the All-Data cloud are Ground and which are
not), interpolation errors between LiDAR points, error in placement of ground control points
(ground-truth), LiDAR point variation, slope of the surface being measured, and vegetation (land
cover). Fowler (2000, 2001) stressed that there are physical limits on the ability of airborne
LiDAR pulses to capture data based upon aircraft flying height, LiDAR impulse wavelength,
GPS limitations, etc. There is therefore a physical limit beyond which airborne LiDAR cannot
resolve. Fowler determined that the best you could achieve with airborne LiDAR under optimal
conditions was about ± 0.5 foot. Additional errors due to system processing error compounded
by vegetative conditions, slope, land cover, etc. all would cause actual accuracy to be less than
optimal.
Internal errors within the LiDAR system and data processing errors are beyond the ability
of the typical LiDAR user to know or control but knowledge and quantification of the ground-
based sources of error in LiDAR-derived elevations could help the typical LiDAR user
understand the abilities and limitations of the data. Gardina (2008) looked at the 2002 and 2006
LiDAR coverages in the CSPA and used the same ground truth data as used in this project. He
determined that up to 75% of the overall error in LiDAR-derived elevations was due to internal
system error in the LiDAR data.
In our study, the results obtained in Part 1 (LiDAR accuracy and precision assessment
using ground truth measurements) confirmed the previously reported sources of LiDAR error.
Both increasing slope of the surface being imaged and increasing density of surface and
overhanging vegetation had a negative effect on the accuracy of the LiDAR data. The year-to-
year precision of LiDAR-derived stream transects was surprisingly poor considering the overall
measured accuracy of individual LiDAR points. The user of LiDAR needs to remember that
comparing multi-year LiDAR coverages in streambeds combines the worst of all error
conditions. The near-stream and in-stream conditions of high slope, narrow streambeds (which
increase the difficulty in resolving topography with a limited LiDAR point density), and
increased near-stream and in-stream vegetation all combine to create an uncertain LiDAR-
derived elevation surface used to delineate stream channels. Comparing multiple years of these
coverages increases the error when computing changes. Given the limitations of LiDAR-derived
DEM-based stream channels revealed in this study, the LiDAR user is advised to use LiDAR-
indicated areas of stream channel change with a grain of salt and a dash of skepticism. LiDAR
did show a weak association with the stream transects ground truth (Figure 11 and Table 7).
Airborne LiDAR may be useful to identify areas where ground surveys should be conducted to
verify the LiDAR-derived results. Hand-held, ground-based LiDAR may be a good method of
measuring stream channel change and the use of raw LiDAR data (rather than the LiDAR-
derived OEMs) may also improve year-to-year comparative accuracy and precision.
The results in Part 2 (LiDAR precision assessment using multi-year LiDAR coverages
where no ground truth measurements exist) were also less clear than I had hoped. LiDAR is
50
-------
clearly useful in mapping large-scale changes in topography that result along with urban
development. LiDAR-derived elevations in forested areas appear to have less accuracy/precision
that those in more open areas studied (agricultural, impervious surfaces, and urban cultivated
grasses).
When looking at LiDAR-derived DEM precision by slope gradient by LULC, the results
were also less than clear. The largest slope-percent quintile typically displayed increased error
but the first few quintiles typically did not. The Forest LULC class, particularly in Sopers
Branch, showed the least difference in elevation between slope-percent quintiles. Sopers Branch
is the largest watershed and the most forested and would have the greatest range of forested slope
gradients compared to T104 and T109 where the forested areas were typically found in the
riparian areas. I believe that within-group vegetation variability within the Forested land cover
class is so high that it overwhelms the slope gradient/LiDAR accuracy relationship. Both of the
"OK" and "Bad" vegetation classes delineated in the ground-truth portion of the study (Appendix
Nine) would be included in the general land cover class of "Forest". I believe that the near-
ground impact of vegetation condition is larger than the general LULC impact on predicting
LiDAR accuracy/precision. This limits the use of LULC as a filter for predicting LiDAR
accuracy/precision to a first-order estimator. For the LULC classes representing more open areas
(Agricultural, Impervious Surfaces, and Cultivated Urban Grasses), using LULC as a filter to
predict relative LiDAR accuracy/precision appears more justified.
The variability in stream channel placement using LiDAR-derived OEMs to predict
stream channels using ArcHydro flow accumulation algorithms led to highly variable stream
channel locations and these impacted the watershed boundaries and areas computed to a much
greater extent that I would have predicted at the start of this study. Particularly disturbing to
people attempting to model watershed hydrology was the variability in watershed area derived
from year-to-year LiDAR coverages. The results of this study should be taken as a warning to
watershed modelers and managers that watershed boundaries and areas may not be nearly so
certain as one might otherwise assume. The final conclusion of this study is that a user of
LiDAR who attempts to delineate changes at or near to the spatial level of resolution of the base
data (about 1 meter or 3 feet in our study) cannot rely on LiDAR alone to assess those changes.
Ground truth is needed to verify changes predicted by repeat LiDAR measurements. The loose
relationship seen by the LiDAR transects and ground-truth transects (Figure 11) should act as a
warning to the LiDAR user: Use LiDAR to identify where change appears to have occurred and
then use ground -truth to verify.
Further research would be useful to determine if the use of raw data LiDAR-derived
surfaces to look for stream channel change would improve the accuracy and utility of year-to-
year geomorphologic assessments of stream channel change compare to the DEM-based
comparisons studied in this report.
51
-------
References
Caraco, D., R. Claytor, P. Hinkel, H. Y. Kwon, T. Schueler, C. Swann, S. Vysotsky, and J.
Zielinske. 1998. Rapid Watershed Planning Handbook. Center for Watershed Protection,
Ellicott City, Maryland.
Fowler, R. A. 2000. The Lowdown on Lidar. Earth Observation Magazine (EOM) 9(3): 27-30.
Fowler, R. A. 2001. The thorny problem of LIDAR Specifications. Earth Observation
Magazine (EOM) 10(4): 25-28.
Gardina, V. J. 2008. Analysis Of Lidar Data For Fluvial Geomorphic Change Detection At A
Small Maryland Stream. Master of Science Thesis, University of Maryland, College Park MD.
Hodgson, M. E. and P. Bresnahan. 2004. Accuracy of Airborne Lidar-Derived Elevation:
Empirical Assessment and Error Budget. Photogrammetric Engineering & Remote Sensing
70(3): 331-339.
Hodgson, M. E., J. Jensen, G. Raber, J. Tullis, B. A. Davis, G. Thompson, and K. Schuckman.
2005. An evaluation of Lidar-derived elevation and terrain slope in leaf-off conditions.
Photogrammetric Engineering & Remote Sensing 71(7): 817-824.
Jarnagin, S. T. 2007. Historical analysis of the relationship of streamflow flashiness with
population density, imperviousness, and percent urban land cover in the Mid-Atlantic region.
Internal Report APM 408. United States Environmental Protection Agency, Office of Research
and Development, Environmental Photographic Interpretation Center (EPIC), Reston VA,
September 2007, EPA/600/X-07/021, lOOp.
Jarnagin, S. T. 2004. Regional and Global Patterns of Population, Land Use, and Land Cover
Change: An Overview of Stressors and Impacts. GIScience & Remote Sensing 41(3): 207-227.
Jennings, D. B. and S. T. Jarnagin. 2002. Changes in anthropogenic impervious surfaces,
precipitation and daily streamflow discharge: a historical perspective in a mid-Atlantic
subwatershed. Landscape Ecology 17(5): 471-489.
Maidment, D. R. (Editor) 2002. Arc Hydro: GIS for Water Resources. ESRI Press, Redlands
CA.
Maune, D. F. (Editor) 2001. Digital Elevation Model Technologies and Applications: The Dem
Users Manual. ASPRS Publications, Bethesda, MD. ISBN: 1570830649 / 978-1-5708306-4-8 /
1-57083-064-9.
Rosgen, D. L. 1994. A Stream Classification System. Catena, 22 169199. Elsevier Science,
Amsterdam, Netherlands.
52
-------
Rosgen, D. L. 1996. Applied River Morphology. 2nd ed. Wildland Hydrology, Fort Collins,
CO. ISBN: 978-0-9653289-0-6.
Schueler T. R. 1994. The Importance of Imperviousness. Watershed Protection Techniques
1(3): 100-111.
Strahler, A. N. 1957. Quantitative analysis of watershed geomorphology. Transactions of the
American Geophysical Union 8(6): 913-920.
USEPA. 1994. The Quality of Our Nation's Water: 1992. United States Environmental
Protection Agency, EPA-841-S-94-002, Washington, D.C., USEPA, Office of Water.
53
-------
Appendices
Table of Appendices 54
Appendix One: Year-by-year mean absolute accuracy by slope quintile. I
Appendix Two: Year-by-year mean absolute difference by slope quintile. VI
Appendix Three: Sequence of modified Anderson Level One LULC classes mapped
from 1-foot (or better) digital orthoimages. X
Appendix Four: Sequence of figures that display the ground-truth stream transect
measurements and LiDAR-derived stream transect elevation values. XXVI
Appendix Five: Large-Scale Elevation Changes Shown in LiDAR. Sequence of
images that display the year-to-year LiDAR-derived elevation differences
between 3-foot DEMs. XL
Appendix Six: LiDAR-Derived DEM Precision by Slope Gradient by LULC. LIII
Appendix Seven: Metrics calculated for stream buffer High vs. Low-Variance
Stream Channel Areas. LXXVII
Appendix Eight: Year-to-year differences in watersheds calculated from
subsequent years of LiDAR coverages. Gray = Area Present in both coverages;
Red = Area Present in Year 1 but not Year 2;
Blue = Area Present in Year 2 but not Year 1. LXXXV
Appendix Nine: Examples of Vegetation Classes used in the LiDAR Accuracy
Assessment Ground Truth Survey Area. XCVII
54
-------
Appendix One: Year-by-year mean absolute accuracy by slope quintile for ground
control survey points (n = 604). These data are noisier than the sum of years but show the same
general trend: increasing accuracy with lower slope.
1.5
1.0
o
u
(0
<
-------
1.5 7
1.0
u
u
<
ro 0.5
0.0 4
ANOVA
F-ratio = 17.189
n = 604
df=4
P < 0.001
Error Bars = ± 95% C.I.
0-3.6 3.6-8.5 8.5-19.3 19.3-31.7
Slope Quintiles (Percent Change)
31.7-77.9
2004 Mean Absolute Accuracy by Slope Quintile
Slope Quintiles
0-3.6
3.6 - 8.5
8.5 - 19.3
19.3 - 31.7
31.7 - 77.9
Mean 2004
Accuracy
0.24
0.45
0.67
0.89
0.90
n
116
123
121
122
122
stdev
0.29
0.44
0.69
0.72
0.84
± 95% C.I.
0.05
0.08
0.12
0.13
0.15
II
-------
1.5 7
S KH
u
0
u
« 0.5
0.0
AN OVA
F-ratio = 29.797
n = 604
df=4
P < 0.001
Error Bars = ± 95% C.I.
0-3.6 3.6-8.5 8.5-19.3 19.3-31.7
Slope Quintiles (Percent Change)
31.7-77.9
2006 Mean Absolute Accuracy by Slope Quintile
Slope Quintiles
0-3.6
3.6 - 8.5
8.5 - 19.3
19.3 - 31.7
31.7 - 77.9
Mean 2006
Accuracy
0.21
0.21
0.52
0.52
0.66
n
116
123
121
122
122
stdev
0.13
0.22
0.49
0.61
0.66
± 95% C.I.
0.02
0.04
0.09
0.11
0.12
III
-------
1.5 7
u
o
V)
£1
<5 0.5 -
o.o
ANOVA
F-ratio = 25.936
n = 604
df=4
P < 0.001
Error Bars = ± 95% C.I.
0-3.6 3.6-8.5 8.5-19.3 19.3-31.7
Slope Quintiles (Percent Change)
31.7-77.9
2007 Mean Absolute Accuracy by Slope Quintile
Slope Quintiles
0-3.6
3.6 - 8.5
8.5 - 19.3
19.3 - 31.7
31.7 - 77.9
Mean 2007
Accuracy
0.19
0.25
0.43
0.56
0.61
n
116
123
121
122
122
stdev
0.15
0.24
0.45
0.49
0.58
± 95% C.I.
0.03
0.04
0.08
0.09
0.10
IV
-------
1.5 7
g
u
o
1.0
0.5
AN OVA
F-ratio = 21.468
n = 604
df=4
P < 0.001
Error Bars = ± 95% C.I.
0-3.6 3.6-8.5 8.5-19.3 19.3-31.7
Slope Quintiles (Percent Change)
31.7 - 77.9
2008 Mean Absolute Accuracy by Slope Quintile
Slope Quintiles
0-3.6
3.6 - 8.5
8.5 - 19.3
19.3 - 31.7
31.7 - 77.9
Mean 2008
Accuracy
0.17
0.42
0.69
0.94
1.18
n
116
123
121
122
122
stdev
0.21
0.35
0.71
0.89
1.07
± 95% C.I.
0.04
0.06
0.13
0.16
0.19
V
-------
Appendix Two: Year-by-year mean absolute difference by slope quintile for ground
control survey points (n = 604). These data are noisier than the sum of years but show the same
general trend: increasing precision with lower slope.
1.5
CD
o
1.0
0.5
ANOVA
F-ratio = 21.983
n = 604
df = 4
P < 0.001
0-3.6 3.6-8.5 8.5-19.3 19.3-31.7 31.7-77.9
Error Bars = ± 95% C.I. sl°Pe Quintiles (Percent Change)
2002-2004: Mean Absolute Difference by Slope Quintile
Slope Quintiles
0-3.6
3.6 - 8.5
8.5 - 19.3
19.3 - 31.7
31.7 - 77.9
MeanDiff2002-
04
0.17
0.27
0.34
0.42
0.53
n
116
123
121
122
122
stdev
0.14
0.25
0.29
0.39
0.45
± 95% C.I.
0.03
0.04
0.05
0.07
0.08
VI
-------
1.5
0
o
c
0
W
1.0
0.5
0.0
ANOVA
F-ratio = 22.355
n = 604
df = 4
P < 0.001
0-3.6 3.6-8.5 8.5-19.3 19.3-31.7 31.7-77.9
Error Bars = ± 95% C.I. sl°Pe Quintiles (Percent Change)
2004-2006: Mean Absolute Difference by Slope Quintile
Slope Quintiles
0-3.6
3.6 - 8.5
8.5 - 19.3
19.3 - 31.7
31.7 - 77.9
MeanDiff2004-
06
0.36
0.46
0.64
0.86
0.92
n
116
123
121
122
122
stdev
0.24
0.40
0.64
0.67
0.70
± 95% C.I.
0.04
0.07
0.11
0.12
0.12
VII
-------
1.5
1.0
o
c
£
I
Q
JS
<
0.5
0.0
ANOVA
F-ratio = 27.587
n = 604
df = 4
P < 0.001
0-3.6 3.6-8.5 8.5-19.3 19.3-31.7 31.7-77.9
Error Bars = ± 95% C.I. sl°Pe Quintiles (Percent Change)
2006-2007: Mean Absolute Difference by Slope Quintile
Slope Quintiles
0-3.6
3.6 - 8.5
8.5 - 19.3
19.3 - 31.7
31.7 - 77.9
MeanDiff2006-
07
0.12
0.19
0.32
0.43
0.47
n
116
123
121
122
122
stdev
0.14
0.17
0.34
0.41
0.38
± 95% C.I.
0.03
0.03
0.06
0.07
0.07
VIII
-------
1.5
1.0
01
o
£
I
Q
ra 0.5
ID
0.0
ANOVA
F-ratio = 18.489
n = 604
df = 4
P < 0.001
JL
0-3.6 3.6-8.5 8.5-19.3 19.3-31.7 31.7-77.9
Error Bars = ± 95% C.I. sl°Pe Quintiles (Percent Change)
2007-2008: Mean Absolute Difference by Slope Quintile
Slope Quintiles
0-3.6
3.6 - 8.5
8.5 - 19.3
19.3 - 31.7
31.7 - 77.9
MeanDiff2007-
08
0.23
0.36
0.51
0.57
0.91
n
116
123
121
122
122
stdev
0.34
0.57
0.72
0.60
0.84
± 95% C.I.
0.06
0.10
0.13
0.11
0.15
IX
-------
Appendix Three: Sequence of modified Anderson Level One LULC classes mapped from
1-foot (or better) digital orthoimages (used as the ground-truth for LULC mapping) from the
1998, 2002, 2004, 2006, and 2008 aerial overflights by Montgomery County in the Clarksburg
Special Protection Area. 1998 was used as the base year. LULC coverages were done for the
Sopers Branch, Tributary 104 (T104), and Tributary 109 (T109) watershed areas, with the LULC
mapped to a 500-foot buffer and the County-mapped stream channels overlaid.
x
-------
Level 1 Land Use/
Land Cover Types
Sopers_lulc_1998_Level_1
LULC_Class, Value
0, Unclassified
1, Agricultural
2, Barren
3, Forest
4, Impervious Surface
5, Natural Clearing
6, Urban Grasses Cultivated
7, Urban Grasses Fallow
8, Water
9, Wetland
Sopers Branch - 1998 - Level 1 LULC
-------
Level 1 Land Use/
Land Cover Types
Sopers_lulc_2002_Level_1
LULC_Class, Value
| | 0, Unclassified
| 1, Agricultural
| 2, Barren
| 3, Forest
| 4, Impervious Surface
| | 5, Natural Clearing
| 6, Urban Grasses Cultivated
7, Urban Grasses Fallow
^H 8, Water
9, Wetland
^S^l^i
D 600 1,200
I i i i I i
2,400 Feet
l
Sopers Branch - 2002 - Level 1 LULC
XII
-------
Level 1 Land Use/
Land Cover Types
Sopers_lulc_ 2004_Level_1
LULC_Class, Value
0, Unclassified
1, Agricultural
2, Barren
3, Forest
4, Impervious Surface
5, Natural Clearing
6, Urban Grasses Cultivated
7, Urban Grasses Fallow
8, Water
9, Wetland
^»/v\
Sopers Branch - 2004 - Level 1 LULC
XIII
-------
Level 1 Land Use/
Land Cover Types
Sopers_luIc_2006_Level_1
LULC_Class, Value
0, Unclassified
1, Agricultural
2, Barren
3, Forest
4, Impervious Surface
5, Natural Clearing
6, Urban Grasses Cultivated
7. Urban Grasses Fallow
8, Water
9, Wetland
v^/*\
Sopers Branch - 2006 - Level 1 LULC
XIV
-------
Level 1 Land Use/
Land Cover Types
Sopers_lulc_2008_Level_1
LULC_Class, Value
0, Unclassified
1, Agricultural
2, Barren
3, Forest
4, Impervious Surface
5, Natural Clearing
6, Urban Grasses Cultivated
7, Urban Grasses Fallow
8, Water
9, Wetland
X-^v\
Sopers Branch - 2008 - Level 1 LULC
-------
Level 1 Land Use/
Land Cover Types
T104_lulc_1998_Level_1
LULC_Class, Value
0, Unclassified
1, Agricultural
2, Barren
3, Forest
4, Impervious Surface
5, Natural Clearing
6, Urban Grasses Cultivated
7, Urban Grasses Fallow
j^H 8, Water
^ 9, Wetland
T104- 1998-Level 1 LULC
XVI
-------
Level 1 Land Use/
Land Cover Types
T104_l u lc_ 2002 _Level_1
LULC_Class, Value
0, Unclassified
1, Agricultural
2, Barren
3, Forest
4, Impervious Surface
5, Natural Clearing
6, Urban Grasses Cultivated
7, Urban Grasses Fallow
8, Water
9, Wetland
T104 - 2002 - Level 1 LULC
XVII
-------
Level 1 Land Use/
Land Cover Types
T104_lulc_2004_Level_1
LULC_Class, Value
0, Unclassified
1, Agricultural
2, Barren
3, Forest
4, Impervious Surface
5, Natural Clearing
6. Urban Grasses Cultivated
7, Urban Grasses Fallow
8, Water
9, Wetland
475 950 1,900 Feet
I i i i l i i i I
T104 - 2004 - Level 1 LULC
XVIII
-------
Level 1 Land Use/
Land Cover Types
T104_lulc_ 2006_Level_1
LULC_Class, Value
0, Unclassified
1, Agricultural
2, Barren
3, Forest
4, Impervious Surface
5, Natural Clearing
6, Urban Grasses Cultivated
7, Urban Grasses Fallow
8, Water
9, Wetland
J 475 950 1,900 Feet
I i i i I i i i
T104 - 2006 - Level 1 LULC
XIX
-------
Level 1 Land Use/
Land Cover Types
T104_lulc_2008_Level_1
LULC_Class, Value
0, Unclassified
1, Agricultural
2, Barren
3, Forest
4, Impervious Surface
5, Natural Clearing
6. Urban Grasses Cultivated
7, Urban Grasses Fallow
8, Water
9, Wetland
475 950 1,900 Feet
I i i i l i i i I
T104 - 2008 - Level 1 LULC
xx
-------
D
)
l i
390
780
i l
1,560 Feel
i l
_evel 1 Land Use/Land Cover Types
T109_lulc_1998_Level_1
LULC_Class, Value
| 0, Unclassified
| 1, Agricultural
| 2, Barren
^B 3
-------
D
390
780
I
1,560 Feel
i l
.evel 1 Land Use/Land Cover Types
T109_l u I c_ 2002 _Level_1
LULC_Class, Value
| 0, Unclassified
| 1, Agricultural
[ | 2, Barren
| 3, Forest
| 4, Impervious Surface
| 5, Natural Clearing
| | 6, Urban Grasses Cultivated
| 7, Urban Grasses Fallow
j^H 8, Water
9, Wetland
T109 - 2002 - Level 1 LULC
XXII
-------
D
390
780
I
1,560 Feet
i I _
T109-2004-Level 1
_evel 1 Land Use/Land Cover Types
T109_lulc_2004_Level_1
LULC_Class, Value
| 0, Unclassified
| 1, Agricultural
| 2, Barren
| 3, Forest
| 4, Impervious Surface
^^ 5, Natural Clearing
[ 6, Urban Grasses Cultivated
[ 7, Urban Grasses Fallow
j^H 8. Water
| | 9, Wetland
LULC
XXIII
-------
D
3 390
I i i
780 1,560 Feet
i I i i i I
T109-2006-Level 1
_evel 1 Land Use/Land Cover Types
T109_l u lc_2006 _Level_1
LULC_Class, Value
| 0, Unclassified
| | 1, Agricultural
| | 2, Barren
| 3, Forest
| 4, Impervious Surface
| 5, Natural Clearing
| | 6, Urban Grasses Cultivated
| | 7, Urban Grasses Fallow
^B °' Water
~~| 9, Wetland
LULC
XXIV
-------
D
)
l i
390
780
i l
1,560 Feel
i l
_evel 1 Land Use/Land Cover Types
T109_lulc_2008_Level_1
LULC_Class, Value
| 0, Unclassified
| 1, Agricultural
| 2, Barren
^B 3
-------
Appendix Four: Sequence of figures that display the ground-truth stream transect
measurements and LiDAR-derived stream transect elevation values.
10
15
Distance (ft)
20
25
30
35
40
1.0
£
Q.
O)
Q
"33
-------
10
Distance (ft)
15 20
25
30
35
-1
g
f *
Ground-Truth
Sum Inter-Annual
Delta Difference =
0.41
-m- 2003 Transect (Ground Truth)
2004 Transect (Ground Truth)
2005 Transect (Ground Truth)
-*- 2006 Transect (Ground Truth)
-«- 2007 Transect (Ground Truth)
2002 Transect (Ground Truth)
-»- LiDAR-Derived Depth - 2002
-m- LiDAR-Derived Depth -2004
-t- LiDAR-Derived Depth -2006
-*- LiDAR-Derived Depth -2007
-m- LiDAR-Derived Depth - 2008
LiDAR-Derived
Sum Inter-Annual
Delta Difference =
1.37
LSLS104A1 X1
Interpolated Transect Depths 2002-2007
and LiDAR-Derived Depths 2002-2008
XXVII
-------
s.
-------
10
Distance (ft)
15
20
25
30
-0.5 -
-1 -
-1.5 -
-2 -
-2.5 -
-3
Ground-Truth
Sum Inter-Annual
Delta Difference =
0.46
2004 Transect (Ground Truth)
2005 Transect (Ground Truth)
- 2006 Transect (Ground Truth)
- 2007 Transect (Ground Truth)
-2003 Transect (Ground Truth)
-LiDAR-Derived Depth -2002
-LiDAR-Derived Depth -2004
-LiDAR-Derived Depth -2006
-LiDAR-Derived Depth -2007
-LiDAR-Derived Depth -2008
-2002 Transect (Ground Truth)
LiDAR-Derived
Sum Inter-Annual
Delta Difference =
2.60
LSLS104A2X1
Interpolated Transect Depths 2002-2007
and LiDAR-Derived Depths 2002-2008
XXIX
-------
10
20
30
Distance (ft)
40 50
60
70
80
-1
-2
-------
Distance (ft)
10
15
20
25
£
&
Q
0
-0.5
-1 -
-1.5
-2
-2.5 -
Ground-Truth
Sum Inter-Annual
Delta Difference =
0.93
-*- 2004 Transect (Ground Truth)
2005 Transect (Ground Truth)
—*- 2006 Transect (Ground Truth)
-•-2007 Transect (Ground Truth)
-•—2003 Transect (Ground Truth)
-»-LiDAR-Derived Depth -2002
-•-LiDAR-Derived Depth -2004
-*-LiDAR-Derived Depth -2006
-»- LiDAR-Derived Depth -2007
-•-LiDAR-Derived Depth -2008
2002 Transect (Ground Truth)
LiDAR-Derived
Sum Inter-Annual
Delta Difference =
2.25
LSLS104A3X1
Interpolated Transect Depths 2002-2007
and LiDAR-Derived Depths 2002-2008
XXXI
-------
10
Distance (ft)
15 20
25
30
35
£
Q.
-------
10
Distance (ft)
15 20 25
30
35
40
Q.
<1>
Q
0
-0.5
-1
-1.5 -
-2
-2.5
-3 -
Ground-Truth
Sum Inter-Annual
Delta Difference =
0.91
-*- 2004 Transect (Ground Truth)
2005 Transect (Ground Truth)
—*- 2006 Transect (Ground Truth)
-•- 2007 Transect (Ground Truth)
2003 Transect (Ground Truth)
-»-LiDAR-Derived Depth -2002
-m- LiDAR-Derived Depth - 2004
-*- LiDAR-Derived Depth - 2006
-»-LiDAR-Derived Depth -2007
-•-LiDAR-Derived Depth -2008
LiDAR-Derived
Sum Inter-Annual
Delta Difference =
2.36
LSLS104A3X3
Interpolated Transect Depths 2003-2007
and LiDAR-Derived Depths 2002-2008
XXXIII
-------
10
Distance (ft)
15
20
25
30
-0.5
-1 -
-1.5 -
-2
-2.5
-3
Ground-Truth
Sum Inter-Annual
Delta Difference =
0.88
-*- 2004 Transect (Ground Truth)
2005 Transect (Ground Truth)
-*- 2006 Transect (Ground Truth)
-«- 2007 Transect (Ground Truth)
2003 Transect (Ground Truth)
-»-LiDAR-Derived Depth -2002
-•-LiDAR-Derived Depth -2004
-*-LiDAR-Derived Depth -2006
-»- LiDAR-Derived Depth -2007
-•-LiDAR-Derived Depth -2008
LiDAR-Derived
Sum Inter-Annual
Delta Difference =
3.40
LSLS104A3X4
Interpolated Transect Depths 2003-2007
and LiDAR-Derived Depths 2002-2008
XXXIV
-------
10
Distance (ft)
15 20 25
30
35
40
0
-0.5
-1
-1.5
-2
-2.5
•3
Ground-Truth
Sum Inter-Annual
Delta Difference =
0.13
- 2006 Transect (Ground Truth)
- 2007 Transect (Ground Truth)
-2005 Transect (Ground Truth)
LiDAR-Derived Depth -2004
-LiDAR-Derived Depth -2006
-LiDAR-Derived Depth -2007
LiDAR-Derived
Sum Inter-Annual
Delta Difference =
1.30
LSLS109A1 X1
Interpolated Transect Depths 2005-2007
and LiDAR-Derived Depths 2004-2007
XXXV
-------
10
Distance (ft)
15 20
25
30
35
0.00
-1.00 -
-2.00
£ -3.00 -
Q.
W
Q
4.00
Ground-Truth
Sum Inter-Annual
Delta Difference =
0.25
-*- 2006 Transect (Ground Truth)
-*- 2007 Transect (Ground Truth)
2005 Transect (Ground Truth)
-•-LiDAR-Derived Depth -2004
-*- LiDAR-Derived Depth - 2006
-»- LiDAR-Derived Depth -2007
LiDAR-Derived
Sum Inter-Annual
Delta Difference =
1.30
LSLS109A1 X2
Interpolated Transect Depths 2005-2007
and LiDAR-Derived Depths 2004-2007
XXXVI
-------
Distance (ft)
10 15
-1
£
Q.
-3
Ground-Truth
Sum Inter-Annual
Delta Difference =
0.28
-+- 2006 Transect (Ground Truth)
—«— 2007 Transect (Ground Truth)
2005 Transect (Ground Truth)
-•-LiDAR-Derived Depth -2004
-*-LiDAR-Derived Depth -2006
-»- LiDAR-Derived Depth -2007
20
25
LiDAR-Derived
Sum Inter-Annual
Delta Difference =
3.33
LSLS109A2X1
Interpolated Transect Depths 2005-2007
and LiDAR-Derived Depths 2004-2007
XXXVII
-------
Distance (ft)
8 10
12
14
16
18
-1 -
-2
-3
-4
-5
Ground-Truth
Sum Inter-Annual
Delta Difference =
0.60
2006 Transect (Ground Truth)
2007 Transect (Ground Truth)
-2005 Transect (Ground Truth)
-LiDAR-Derived Depth -2004
-LiDAR-Derived Depth -2006
-LiDAR-Derived Depth -2007
LiDAR-Derived
Sum Inter-Annual
Delta Difference =
3.55
LSLS109A2X2
Interpolated Transect Depths 2005-2007
and LiDAR-Derived Depths 2004-2007
XXXVIII
-------
Distance (ft)
10 15
20
25
0.00
-0.50
-1.00
-1.50
s -2.oo ^
Q.
Q -2.50
-3.00
Ground-Truth
Sum Inter-Annual
Delta Difference =
0.18
2006 Transect (Ground Truth)
- 2007 Transect (Ground Truth)
•2005 Transect (Ground Truth)
-LiDAR-Derived Depth -2004
-LiDAR-Derived Depth -2006
-LiDAR-Derived Depth -2007
LiDAR-Derived
Sum Inter-Annual
Delta Difference =
1.48
LSLS109A3X1
Interpolated Transect Depths 2005-2007
and LiDAR-Derived Depths 2004-2007
XXXIX
-------
Appendix Five: Large-Scale Elevation Changes Shown in LiDAR. Year-to-year
differences in 3-foot LiDAR-derived DEMs are shown for sequential years 2002-2004, 2004-
2006, 2006-2007, and 2007-2008 for the Sopers Branch, T104, and T109 watersheds within the
CSPA. All elevation differences have been scaled to the same metric, with the difference
between the later year minus the earlier year shown; increasing red (negative) values indicate
that the elevation was greater in the prior year (elevation has decreased over time) while
increasing blue (positive) values indicate that the elevation was less in the prior year (elevation
has increased over time).
• -40.29556274 - -35.23375323
• -35.23375327 - -24.09777246
• -24.09777245 - -21.06068673
• -21.06068677 - -19.03596299
• -19.03596298 - -17.01123921
• -17.0112392 - -13,97415353
• -13.97415352 - -10.93706785
• -10.93706784 - -7,899982168
• -7.899982167 - -4,862896489
CH-4.862896488--2
fZI-1-999999999--1
n -0.999999999-0
no-1
n 1.000000001 - 2
02.000000001 - 3.742179601
LZI3.742179602 - 7.285446227
07.285446228 - 10.82871285
0 10.82871286 - 13,86579853
0 13.86579854 - 16,90288421
• 16.90288422 - 19.93996989
• 19.9399699 - 22.47087462
• 22.47087463 - 24,49559841
• 24.49559842 - 26.52032219
• 26.5203222 - 28.54504598
• 28.54504599 - 30,56976977
• 30.56976978 - 32.59449355
• 32.59449356 - 37,65630302
• 37.65630303 - 47,273741
• 47.27374101 - 58,40972182
• 58.40972183 - 73,59515022
• 73.59515023 - 88.78057861
Legend for Large-Scale Elevation Changes
XL
-------
Sopers Branch
Elevation
Differences (ft)
Between
2002 - 2004
-40.29556274 - -35.23375328
-35.23375327 - -24.09777246
-24.09777245 - -21.06068678
-21.06068677 - -19.03596299
-19.03596298 - -17.01123921
-17.0112392 - -13.97415353
-13.97415352 - -10.93706785
-10.93706784 - -7.899982168
CD-7.899982167 - -4.862896489
O-4.8628964BB--2
d-1.999999999 --1
d -0.999999999-0
OO-l
i.ooooooooi- 2
LH2.000000D01 - 3.742179601
£33.742179602 - 7.285446227
D 7.285446228 - 10.82871285
• 10.82871286 - 13.86579853
• 13.86579854 - 16.90288421
• 16.90288422 - 19.93996989
• 19.9399699 - 22.47087462
• 22.47087463 - 24.49559841
• 24.49559842 - 26.52032219
• 26.5203222 - 28.54504598
• 28.54504599 - 30,56976977
• 30.56976978 - 32.59449355
• 32.59449356 - 37.6563D302
• 37,65630303 - 47.273741
• 47.27374101 - 58.40972182
58.40972183 - 73,59515022
73.59515023 - 88.78057861
Sopers Branch 2002-2004
XLI
-------
I Nl
A
Sopers Branch
850
1,700
I
3,400 Feet
J
Elevation
Differences (ft)
Between
2004 - 2006
• -40.29556274 - -35.23375328
• -35.23375327 - -24.09777246
• -24.09777245 - -21.06068679
• -21.06068677 - -19.03596299
• -19.0359629B - -17.01123921
• -17.0112392 - -13.97415353
• -13.97415352 - -10.93706785
•-10.93706784 - -7.B999S216B
d -7.899982167 - -4.862896489
d-4.862896488--2
d-1.999999999 --1
n-0.999999999-0
CJO-1
n 1.000000001 - 2
IZ12.000000001 - 3.742179601
03.742179602 - 7.285446227
IH 7.285446228 - 10.82871285
• 10.82871286 - 13.86579853
• 13.86579854 - 16.90288421
• 16.90288422 - 19.93996989
• 19.9399699 - 22.47087462
• 22.47087463 - 24.49559841
• 24.49559842 - 26.52032219
• 26.5203222 - 28.54504598
• 28.54504599 - 30.56976977
• 30.56976978 - 32.59449355
• 32.59449356 - 37.65630302
• 37.65630303 - 47.273741
• 47.27374101 - 58.40972182
• 58.409721B3 - 73.59515022
• 73.59515023 - BB.7B057B61
Sopers Branch 2004-2006
XLII
-------
Sopers Branch
Elevation
Differences (ft)
Between
2006 - 2007
-10.29556274 - -35.23375328
-35.23375327 - -24.09777246
-24.09777245 - -21,06066678
-21.06068677 - -19.03596299
-19.03596298 - -17.01123921
-17.0112392 - -13.97415353
-13.97415352 - -10.93706785
-10.93706784 - -7.899982168
-7.899982167 - -4.862896489
-4.862896488 - -2
r>l.999999999 - -1
Zl -0.999999999-0
PO-I
1.000000001 - 2
!Z!2.000000001 - 3.742179601
CH3.742179602 - 7.285446227
H7.285446228 - 10.82871285
• 10.82871286 - 13,86579853
• 13.86579854 - 16,90288421
• 16.90288422 - 19.93996989
• 19.9399699 - 22.47087462
• 22.47087463 - 24,49559841
• 24.49559842 - 26.52032219
• 26.5203222 - 28.54504598
• 28.54504599 - 30.56976977
• 30.56976978 - 32,59449355
• 32.59449356 - 37.65630302
• 37.65630303 - 47.273741
• 47.27374101 - 58.40972182
• 58.40972183 - 73,59515022
Sopers Branch 2006-2007
XLIII
-------
Sopers Branch
Elevation
Differences (ft)
Between
2007 - 2008
-10.29556274 - -35.23375328
-35.23375327 - -24.09777246
-24.09777245 - -21,06066678
-21.06068677 - -19.03596299
-19.03596298 - -17.01123921
-17.0112392 - -13.97415353
-13.97415352 - -10.93706785
-10.93706784 - -7.899982168
-7.899982167 - -4.862896489
-4.862896488 - -2
r>l.999999999 - -1
Zl -0.999999999-0
PO-I
1.000000001 - 2
!Z!2.000000001 - 3.742179601
CH3.742179602 - 7.285446227
H7.285446228 - 10.82871285
• 10.82871286 - 13,86579853
• 13.86579854 - 16,90288421
• 16.90288422 - 19.93996989
• 19.9399699 - 22.47087462
• 22.47087463 - 24,49559841
• 24.49559842 - 26.52032219
• 26.5203222 - 28.54504598
• 28.54504599 - 30.56976977
• 30.56976978 - 32,59449355
• 32.59449356 - 37.65630302
• 37.65630303 - 47.273741
• 47.27374101 - 58.40972182
• 58.40972183 - 73,59515022
Sopers Branch 2007-2008
XLIV
-------
Elevation
Differences (ft)
Between
2002 - 2004
-40,29556274
-35.23375327
-24.09777245
-21.06068677
-19.03596296
-17.0112392 -
-13.97415352
-10.937067B4
-7.899982167
O4.8628964BB
d-1. 999999999
d -0.999999999
- -35.23375328
- -24.09777246
- -21.0606S678
- -19.03596299
- -17.01123921
-13.97415353
- -10.93706785
- -7.89998216B
- -4.862896489
--2
- -1
-0
1.000000DD1
n2.00000D001
• 3.742179602
7.285446228
• 10.82871286
• 13.86579854
• 16.90288422
• 19.9399699 -
• 22.47087463
• 24.49559842
• 26.5203222 -
• 28.54504599
• 30.56976978
• 32.59449356
• 37.65630303
• 47.27374101
• 58.40972183
73.59515023
- 3.742179601
- 7.285446227
- 10,82871285
- 13.86579853
- 16.90288421
- 19.93996989
22.47087462
- 24.49559841
- 26.52032219
28.54504598
- 30.56976977
- 32.59449355
- 37.65630302
- 47.273741
- 58.40972182
- 73.59515022
- 88.78057861
T104 2002-2004
XLV
-------
Elevation
Differences (ft)
Between
2004 - 2006
-40.29556274 - -35.23375328
-35.23375327 - -24.09777246
-24.09777245 - -21,06068678
-21.06068677 - -19.03596299
-19.03596299 - -17,01123921
-17.0112392 - -13.97415353
-13.97415352 - -10.93706785
-10.93706784 - -7.8999B2168
-7,899982167 - -4.862896489
-4.862896438 - -2
-1,999999999 --1
-0,999999999-0
na-i
d 1,000000001 - 2
D2,000000001 - 3.742179601
• 3.742179602 - 7.285446227
CH 7.285446228 - 1D.82871285
• 10.82871286 - 13.86579853
• 13.86579854 - 16.90288421
• 16.90288422 - 19.93996989
• 19.9399699 - 22.47087462
• 22.47087463 - 24,49559841
• 24.49559842 - 26.52032219
• 26.5203222 - 28.54504598
• 28.54504599 - 30,56976977
• 30.5697697B - 32.59449355
• 32.59449356 - 37.65630302
• 37.65630303 - 47.273741
• 47.27374101 - 58.40972182
• 58.40972183 - 73.59515022
.59515023 - 88.78D57861
T104 2004-2006
XL VI
-------
Elevation
Differences (ft)
Between
2006 - 2007
-40.29556274
-35.23375327
-24.09777245
-21.06068677
-19.03596298
-17.0112392 -
-13.97415352
-10.93706784
-7.899902167
-4.862896488
-1. 999999999
Zl -0.999999999
- -35,23375328
- -24,09777246
- -21.06068678
- -19,03596299
- -17.01123921
-13.97415353
- -10.93706785
- -7.899982168
- -4.862896489
- -2
- -1
-0
Zl 1.00000DD01
n2. 000000001
• 3.742179602
07. 285446228
• 10.82871286
0 13.86579854
• 16.90288422
• 19.9399699 -
• 22.47087463
• 24.49559842
• 26.5203222 -
• 28.54504599
• 30.56976978
• 32.59449356
• 37.65630303
• 47.27374101
• 58.40972183
3.59515023
- 2
- 3.742179601
- 7.285446227
- 10,82871285
- 13.86579853
- 16,90288421
- 19,93996989
22.47087462
- 24,49559841
- 26,52032219
28.54504598
- 30.56976977
- 32,59449355
- 37.65630302
- 47.273741
- 58.40972182
- 73.59515022
- 88,78057861
J
1
625
i i
1,250
i l i
2,500 Feel
i i 1
T104 2006-2007
XL VII
-------
Elevation
Differences (ft)
Between
2007 - 2008
-40,29556274 - -35.233'5?,2E:
-35.23375327 - -24.09777246
-24.09777245 - -21.06068678
-21.06068677 - -19.03596299
_ -19.03596298 - -17.01123921
• -17.0112392 - -13.97415353
• -13.97415352 - -10.93706785
• -10.93706784 - -7.899982168
-7. 899982167 - -4.862896489
-4. 862896488 --2
-1.999999999 --1
n -0.999999999-0
1,000000001
02,000000001
3.742179602
07.285446228
10,82871286
• 13.86579854
• 16,90288422
• 19,9399699 -
• 22,47087463
• 24,49559842
• 26,5203222 -
• 28,54504599
• 30,56976978
• 32.59449356
• 37,65630303
• 47,27374101
- 3.742179601
- 7.285446227
- 10.82871285
- 13.86579853
- 16.90288421
- 19.93996989
22.47087462
- 24.49559841
- 26.52032219
28.54504598
- 30.56976977
- 32.59449355
- 37.65630302
- 47.273741
- 58.40972182
625 1,250 2,500 Feel
i i i I i i i I
T104 2007-2008
XLVIII
-------
Elevation
Differences (ft)
2002 - 2004
-40.29556274
-35.23375327
-24.09777245
-21.06068677
-19.03596298
-17.0112392 -
-13.97415352
-10.93706784
-7.899902167
-4.862896488
-1. 999999999
Zl -0.999999999
- -35,23375328
- -24,09777246
- -21.06068678
- -19,03596299
- -17.01123921
-13.97415353
- -10.93706785
- -7.899982168
- -4.862896489
- -2
- -1
-0
Zl 1.00000DD01
n2. 000000001
• 3.742179602
07. 285446228
• 10.82871286
0 13.86579854
• 16.90288422
• 19.9399699 -
• 22.47087463
• 24.49559842
• 26.5203222 -
• 28.54504599
• 30.56976978
• 32.59449356
• 37.65630303
• 47.27374101
• 58.40972183
73.59515023
- 2
- 3.742179601
- 7.285446227
- 10,82871285
- 13.86579853
- 16,90288421
- 19,93996989
22.47087462
- 24,49559841
- 26,52032219
28.54504598
- 30.56976977
- 32,59449355
- 37.65630302
- 47.273741
- 58.40972182
- 73.59515022
- 88,78057861
5 500 1,000 2,000 Feet
I i i i I i i i I
T109 2002-2004
XLIX
-------
Elevation
Differences (ft)
Between
2004 - 2006
-40,29556274
-35.23375327
-24.09777245
-21.06068677
-19.03596296
-17.0112392 -
-13.97415352
-10.937067B4
-7.899982167
O4.8628964BB
d-1. 999999999
d -0.999999999
- -35.23375328
- -24.09777246
- -21.0606S678
- -19.03596299
- -17.01123921
-13.97415353
- -10.93706785
- -7.89998216B
- -4.862896489
--2
- -1
-0
1.000000DD1
2.000000001
• 3.742179602
07.285446228
• 10.82871286
• 13.86579854
• 16.90288422
• 19.9399699 -
• 22.47087463
• 24.49559842
• 26.5203222 -
• 28.54504599
• 30.56976978
• 32.59449356
• 37.65630303
• 47.27374101
- 3.742179601
- 7.285446227
- 10,82871285
- 13.86579853
- 16.90288421
- 19.93996989
22.47087462
- 24.49559841
- 26.52032219
28.54504598
- 30.56976977
- 32.59449355
- 37.65630302
- 47.273741
- 58,40972182
T109 2004-2006
-------
Elevation
Differences (ft)
2006 - 2007
-40.29556274
-35.23375327
-24.09777245
-21.06068677
-19.03596298
-17.0112392 -
-13.97415352
-10.93706784
-7.899902167
-4.862896488
-1. 999999999
Zl -0.999999999
- -35,23375328
- -24,09777246
- -21.06068678
- -19,03596299
- -17.01123921
-13.97415353
- -10.93706785
- -7.899982168
- -4.862896489
- -2
- -1
-0
Zl 1.00000DD01
n2. 000000001
• 3.742179602
07. 285446228
• 10.82871286
0 13.86579854
• 16.90288422
• 19.9399699 -
• 22.47087463
• 24.49559842
• 26.5203222 -
• 28.54504599
• 30.56976978
• 32.59449356
• 37.65630303
• 47.27374101
• 58.40972183
73.59515023
- 2
- 3.742179601
- 7.285446227
- 10,82871285
- 13.86579853
- 16,90288421
- 19,93996989
22.47087462
- 24,49559841
- 26,52032219
28.54504598
- 30.56976977
- 32,59449355
- 37.65630302
- 47.273741
- 58.40972182
- 73.59515022
- 88,78057861
ar
3
|
500
i i
1,000
1 1 1
2,000 Feet
i i l
T109 2006-2007
LI
-------
Elevation
Differences (ft)
-40.29556274
-35.23375327
-24.09777245
-21.06068677
-19.03596298
-17.0112392 -
-13.97415352
-10.93706784
-7.899902167
-4.862896488
-1. 999999999
Zl -0.999999999
- -35,23375328
- -24,09777246
- -21.06068678
- -19,03596299
- -17.01123921
-13.97415353
- -10.93706785
- -7.899982168
- -4.862896489
- -2
- -1
-0
Zl 1.00000DD01
n2. 000000001
• 3.742179602
07. 285446228
• 10.82871286
0 13.86579854
• 16.90288422
• 19.9399699 -
• 22.47087463
• 24.49559842
• 26.5203222 -
• 28.54504599
• 30.56976978
• 32.59449356
• 37.65630303
• 47.27374101
• 58.40972183
73.59515023
- 2
- 3.742179601
- 7.285446227
- 10,82871285
- 13.86579853
- 16,90288421
- 19,93996989
22.47087462
- 24,49559841
- 26,52032219
28.54504598
- 30.56976977
- 32,59449355
- 37.65630302
- 47.273741
- 58.40972182
- 73.59515022
- 88,78057861
3
|
500
i i
1,000
1 1 1
2,000 Feet
i i l
T109 2007-2008
LII
-------
Appendix Six: Individual Between-Year Differences in LiDAR-Derived DEM Precision
by Slope Gradient by LULC
U.I
0.4
0.3
g
73
s
uj 02
01
o
IB 0.2
Q
Ol
0.1
n n
Error Bars are the
Standard Deviation
of the Quintile Mean
T
I
0-6
7-10 11-14
Slope Gradient Quintile
15-19
20 - 30+
2004-2002 Forest No-Change Land Cover, Sopers Branch:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-6
7-10
11-14
15-19
20 - 30+
Quintile n
725651
870697
883311
811196
856888
MeanDiff04-02
0.288
0.278
0.278
0.291
0.338
Std Dev
0.003
0.002
0.001
0.007
0.038
LI 11
-------
0.8
0.7
0.6
g
o 0.5
|
5 0.4
s
IB 0.3
ro
01
0.2
0.1
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-6
7-10 11-14
Slope Gradient Quintile
15-19
20 - 30+
2006-2004 Forest No-Change Land Cover, Sopers Branch:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-6
7-10
11-14
15-19
20 - 30+
Quintile n
725651
870697
883311
811196
856888
Mean Diff 06 - 04
0.433
0.398
0.394
0.404
0.448
Std Dev
0.014
0.004
0.001
0.006
0.041
LIV
-------
0.8
0.7
0.6
g
o 0.5
ro
I
m
0.4
01
o
S 0.3
O
c 0.2
CO
01
0.1
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-6
7-10 11-14
Slope Gradient Quintile
15-19
20 - 30+
2007-2006 Forest No-Change Land Cover, Sopers Branch:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-6
7-10
11-14
15-19
20 - 30+
Quintile n
725651
870697
883311
811196
856888
MeanDiff07-06
0.248
0.249
0.247
0.249
0.272
Std Dev
0.001
0.000
0.001
0.002
0.020
LV
-------
ro
01
£
ro
01
0.8
0.7
0.6
0.4
0.3
0.2
0.1
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-6
7-10 11-14
Slope Gradient Quintile
15-19
20 - 30+
2008-2007 Forest No-Change Land Cover, Sopers Branch:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-6
7-10
11-14
15-19
20 - 30+
Quintile n
725651
870697
883311
811196
856888
Mean Diff 08-07
0.243
0.247
0.255
0.268
0.303
Std Dev
0.002
0.002
0.004
0.004
0.045
LVI
-------
u.t
0.4
0.3
g
o 0.3
I
- 0.2
01
o
|ii 0.2
O
| °'1
0.1
n n
Error Bars are the
Standard Deviation
of the Quintile Mean
T
I
0-6
7-10 11-14
Slope Gradient Quintile
15-19
20 - 30+
2004-2002 Forest No-Change Land Cover, T104:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-6
7-10
11-14
15-19
20 - 30+
Quintile n
70616
68418
71703
71926
99502
MeanDiff04-02
0.262
0.256
0.251
0.268
0.320
Std Dev
0.004
0.004
0.004
0.010
0.037
LVII
-------
0.8
0.7
0.6
o 0.5
IS
E 0.4
01
o
§ 0.3
ro
01
0.2
0.1
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-6
7-10 11-14
Slope Gradient Quintile
15-19
20 - 30+
2006-2004 Forest No-Change Land Cover, T104:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-6
7-10
11-14
15-19
20 - 30+
Quintile n
70616
68418
71703
71926
99502
Mean Diff 06 - 04
0.472
0.449
0.433
0.454
0.535
Std Dev
0.005
0.010
0.003
0.012
0.074
LVIII
-------
0.8
0.7
0.6
o 0.5
IS
E 0.4
01
o
§ 0.3
ro
01
0.2
0.1
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-6
7-10 11-14
Slope Gradient Quintile
15-19
20 - 30+
2007-2006 Forest No-Change Land Cover, T104:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-6
7-10
11-14
15-19
20 - 30+
Quintile n
70616
68418
71703
71926
99502
MeanDiff07-06
0.360
0.335
0.326
0.331
0.368
Std Dev
0.008
0.006
0.002
0.004
0.048
LIX
-------
0.8
0.7
0.6
o 0.5
IS
E 0.4
01
o
§ 0.3
ro
01
0.2
0.1
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-6
7-10 11-14
Slope Gradient Quintile
15-19
20 - 30+
2008-2007 Forest No-Change Land Cover, T104:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-6
7-10
11-14
15-19
20 - 30+
Quintile n
70616
68418
71703
71926
99502
Mean Diff 08-07
0.277
0.272
0.260
0.275
0.334
Std Dev
0.003
0.005
0.001
0.009
0.042
LX
-------
o
15
0.5
0.4
0.4
«
0.3
c
2!
•2 0.2
0.1
0.1
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-4
5-7 8-10
Slope Gradient Quintile
11 -14
15-30+
2004-2002 Forest No-Change Land Cover, T109:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-4
5-7
8-10
11-14
15 - 30+
Quintile n
80026
106454
98950
87391
88943
MeanDiff04-02
0.242
0.247
0.261
0.279
0.348
Std Dev
0.004
0.004
0.005
0.006
0.051
LXI
-------
0.8
0.7
0.6
o 0.5
IS
E 0.4
01
o
§ 0.3
ro
01
0.2
0.1
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-4
5-7 8-10
Slope Gradient Quintile
11 -14
15-30+
2006-2004 Forest No-Change Land Cover, T109:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-4
5-7
8-10
11-14
15 - 30+
Quintile n
80026
106454
98950
87391
88943
Mean Diff 06 - 04
0.298
0.285
0.286
0.301
0.438
Std Dev
0.006
0.001
0.002
0.010
0.100
LXII
-------
0.8
0.7
0.6
g
o 0.5
ro
I
m
01
o
0.4
a> 0.3
§ °-2
Ol
0.1
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-4
5-7 8-10 11-14
Slope Gradient Quintile
15-30+
2007-2006 Forest No-Change Land Cover, T109:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-4
5-7
8-10
11-14
15 - 30+
Quintile n
80026
106454
98950
87391
88943
MeanDiff07-06
0.271
0.253
0.242
0.240
0.322
Std Dev
0.006
0.005
0.004
0.004
0.068
LXIII
-------
0.8
0.7
0.6
g
o 0.5
ro
I
m
01
o
0.4
a> 0.3
§ °-2
Ol
0.1
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-4
5-7 8-10 11-14
Slope Gradient Quintile
15-30+
2008-2007 Forest No-Change Land Cover, T109:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-4
5-7
8-10
11-14
15 - 30+
Quintile n
80026
106454
98950
87391
88943
Mean Diff 08-07
0.259
0.277
0.290
0.311
0.389
Std Dev
0.004
0.006
0.003
0.009
0.048
LXIV
-------
0.7
0.6
£• °-5
o
I 0.4
8 0.3
01
ro
01
0.2
0.1
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-2
3-4 5-6 7-9
Slope Gradient Quintile
10-30+
2004-2002 Impervious Surface No-Change Land Cover, Sopers Branch:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-2
3-4
5-6
7-9
10 - 30+
Quintile n
36243
54216
45313
36238
26169
MeanDiff04-02
0.330
0.257
0.254
0.257
0.440
Std Dev
0.065
0.001
0.002
0.008
0.149
LXV
-------
0.8
0.7
0.6
o 0.5
IS
E 0.4
01
o
§ 0.3
ro
01
0.2
0.1
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-2
3-4 5-6 7-9
Slope Gradient Quintile
10-30+
2006-2004 Impervious Surface No-Change Land Cover, Sopers Branch:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-2
3-4
5-6
7-9
10 - 30+
Quintile n
36243
54216
45313
36238
26169
Mean Diff 06 - 04
0.425
0.416
0.370
0.364
0.569
Std Dev
0.026
0.020
0.013
0.009
0.184
LXVI
-------
u.o
0.7
0.6
c
ce in Elevation
O O
*>. Ul
| 0.3
1 °'2
0.1
n n
Error Bars are the
Standard Deviation
of the Quintile Mean
T
0-2
3-4 5-6 7-9
Slope Gradient Quintile
10-30+
2007-2006 Impervious Surface No-Change Land Cover, Sopers Branch:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-2
3-4
5-6
7-9
10 - 30+
Quintile n
36243
54216
45313
36238
26169
MeanDiff07-06
0.220
0.215
0.193
0.190
0.389
Std Dev
0.011
0.011
0.003
0.003
0.169
LXVII
-------
co
s
0.8
0.7
0.6
0.4
01
o
S 0.3
Q
c 0.2
CO
01
0.1
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-2
3-4 5-6 7-9
Slope Gradient Quintile
10-30+
2008-2007 Impervious Surface No-Change Land Cover, Sopers Branch:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-2
3-4
5-6
7-9
10 - 30+
Quintile n
36243
54216
45313
36238
26169
Mean Diff 08-07
0.296
0.284
0.288
0.300
0.485
Std Dev
0.012
0.003
0.002
0.008
0.146
LXVIII
-------
0.6
0.5
~ 0.4
o
1
01
Ol
o
c
Ol
ro
01
0.3
0.2
0.1
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-3
4-5 6-7
Slope Gradient Quintile
8-11
12-30+
2004-2002 Impervious Surface No-Change Land Cover, T104:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-3
4-5
6-7
8-11
12 - 30+
Quintile n
17047
17231
15674
18996
22870
MeanDiff04-02
0.243
0.237
0.245
0.275
0.449
Std Dev
0.015
0.005
0.004
0.021
0.096
LXIX
-------
0.8
0.7
0.6
o 0.5
1
_aj
0.4
0.3
Q
I
0.1
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-3
4-5 6-7
Slope Gradient Quintile
8-11
12-30+
2006-2004 Impervious Surface No-Change Land Cover, T104:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-3
4-5
6-7
8-11
12 - 30+
Quintile n
17047
17231
15674
18996
22870
Mean Diff 06 - 04
0.410
0.407
0.435
0.462
0.538
Std Dev
0.014
0.009
0.011
0.009
0.094
LXX
-------
0.8
0.7
0.6
o 0.5
1
8
0.4
0.3
0.1
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-3
4-5 6-7
Slope Gradient Quintile
8-11
12-30+
2007-2006 Impervious Surface No-Change Land Cover, T104:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-3
4-5
6-7
8-11
12 - 30+
Quintile n
17047
17231
15674
18996
22870
MeanDiff07-06
0.263
0.267
0.283
0.309
0.363
Std Dev
0.002
0.006
0.005
0.010
0.084
LXXI
-------
o
1
01
Ol
o
c
Ol
ro
01
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-3
4-5 6-7
Slope Gradient Quintile
8-11
12-30+
2008-2007 Impervious Surface No-Change Land Cover, T104:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-3
4-5
6-7
8-11
12 - 30+
Quintile n
17047
17231
15674
18996
22870
Mean Diff 08-07
0.290
0.260
0.269
0.293
0.488
Std Dev
0.023
0.001
0.003
0.018
0.120
LXXII
-------
c
o
to
>
.2!
LU
o
c
a)
i
a
c
0.8
0.7 -
0.6 -
0.5 -
0.4
0.3
0.2
0.1
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-2
3-4 5-6 7-9
Slope Gradient Quintile
10-30+
2004-2002 Impervious Surface No-Change Land Cover, T109:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-2
3-4
5-6
7-9
10 - 30+
Quintile n
10299
15552
13947
12007
12662
MeanDiff04-02
0.213
0.205
0.208
0.250
0.501
Std Dev
0.004
0.004
0.001
0.012
0.166
LXXIII
-------
0.8
0.7
0.6
.2 0.5 H
LU
C
o
C
c
m
a)
0.4
0.3
0.2
0.1
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-2
3-4 5-6 7-9
Slope Gradient Quintile
10-30+
2006-2004 Impervious Surface No-Change Land Cover, T109:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-2
3-4
5-6
7-9
10 - 30+
Quintile n
10299
15552
13947
12007
12662
Mean Diff 06 - 04
0.356
0.340
0.344
0.393
0.498
Std Dev
0.016
0.004
0.008
0.021
0.112
LXXIV
-------
c
o
is
HI
c
o
c
c
(0
0.8
0.7 -
0.6 -
0.5 -
0.4 -
0.3 -
0.2 -
0.1
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-2
3-4 5-6 7-9
Slope Gradient Quintile
10-30+
2007-2006 Impervious Surface No-Change Land Cover, T109:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-2
3-4
5-6
7-9
10 - 30+
Quintile n
10299
15552
13947
12007
12662
MeanDiff07-06
0.188
0.200
0.204
0.225
0.345
Std Dev
0.009
0.001
0.001
0.014
0.104
LXXV
-------
c
o
15
o
c
fc
b
c
co
o
0.8
0.7 -
0.6
0.5 -
0.4 -
0.3
0.2
0.1 -
0.0
Error Bars are the
Standard Deviation
of the Quintile Mean
0-2
3-4 5-6 7-9
Slope Gradient Quintile
10- 30+
2008-2007 Impervious Surface No-Change Land Cover, T109:
Mean Difference in Elevation by Slope Gradient Quintile
Quintile Slope
Groups
0-2
3-4
5-6
7-9
10 - 30+
Quintile n
10299
15552
13947
12007
12662
Mean Diff 08-07
0.285
0.231
0.238
0.279
0.485
Std Dev
0.039
0.004
0.006
0.013
0.118
LXXVI
-------
Appendix Seven: Metrics calculated for stream buffer High vs. Low-Variance Stream
Channel Areas.
LXXVII
-------
LXXVIII
-------
LXXIX
-------
LXXX
-------
LXXXI
-------
LXXXII
-------
LXXXIII
-------
LXXXIV
-------
Appendix Eight: Year-to-year differences in watersheds calculated from subsequent years of
LiDAR coverages. Gray = Area Present in both coverages; Red = Area Present in Year 1 but not
Year 2; Blue = Area Present in Year 2 but not Year 1.
Year-to-Year
Differences in
s v. Watershed Area
Determined from
Subsequent
\ LiDAR-Dervived
V 3-foot OEMs
\
uiciiiuii
2002-2004
'.
Gray = Area Present in both coverages; Red = Area Present in Year 1 but not Year 2; Blue
Area Present in Year 2 but not Year 1.
LXXXV
-------
Year-to-Year
.
\ Differences in
Watershed Area
Determined from
Subsequent
\ LiDAR-Dervived
3-foot DEMs
\
-\
^x
>
/
\
Sopers Branch
2004-2006
'
N
s
s
x
\
Gray = Area Present in both coverages; Red = Area Present in Year 1 but not Year 2; Blue
Area Present in Year 2 but not Year 1.
LXXXVI
-------
V
Year-to-Year
Differences in
Watershed Area
Determined from
Subsequent
LiDAR-Dervived
3-foot OEMs
Sopers Branch
2006-2007
\
-
-
^_
/
Gray = Area Present in both coverages; Red = Area Present in Year 1 but not Year 2; Blue
Area Present in Year 2 but not Year 1.
LXXXVII
-------
Year-to-Year
Differences in
> Watershed Area
Determined from
Subsequent
LiDAR-Dervived
3-foot OEMs
y
\
Sopers Branch
2007-2008 \
n
\
\
-
Gray = Area Present in both coverages; Red = Area Present in Year 1 but not Year 2; Blue
Area Present in Year 2 but not Year 1.
LXXXVIII
-------
T104
2002-2004
\
\
1
Year-to-Year
Differences in
Watershed Area
Determined from
Subsequent
LiDAR-Dervived
3-foot OEMs
\
\
•»
Gray = Area Present in both coverages; Red = Area Present in Year I but not Year 2; Blue
Area Present in Year 2 but not Year 1.
LXXXIX
-------
,—A.
T104
2004-2006
X
Year-to-Year
Differences in
Watershed Area
Determined from
Subsequent
LiDAR-Dervived
3-foot OEMs
Gray = Area Present in both coverages; Red = Area Present in Year 1 but not Year 2; Blue
Area Present in Year 2 but not Year 1.
-------
T104
2006-2007
*
\
Year-to-Year \ *
f >
\ •*
Differences in )
Watershed Area
Determined from V >
Subsequent
LiDAR-Dervived ^^-
3-foot DEMs
Gray = Area Present in both coverages; Red = Area Present in Year I but not Year 2; Blue
Area Present in Year 2 but not Year 1.
xci
-------
yx T104
2007-2008
>
>
\
/ 5
Year-to-Year
Differences in
Watershed Area \
Determined from (
Subsequent '"^
LiDAR-Dervived ^
3-foot OEMs \^
Gray = Area Present in both coverages; Red = Area Present in Year 1 but not Year 2; Blue
Area Present in Year 2 but not Year 1.
xcn
-------
T109
2002-2004
,
f
_ .*
r
f
•**
>
J
•
i
/
\ Year-to-Year
Differences in
Watershed Area
Determined from
Subsequent
LiDAR-Dervived
3-foot DEMs
Gray = Area Present in both coverages; Red = Area Present in Year 1 but not Year 2; Blue
Area Present in Year 2 but not Year 1.
XCIII
-------
T109
2004-2006
s
-
/
f
_»«-/
f
Year-to-Year
i
/' Differences in
Watershed Area
I
\ Determined from
^.^,,~ Subsequent
LiDAR-Dervived
3-foot DEMs
Gray = Area Present in both coverages; Red = Area Present in Year 1 but not Year 2; Blue
Area Present in Year 2 but not Year 1.
XCIV
-------
T109
2006-2007
,'•
•
-------
T109
2007-2008
Year-to-Year
Differences in
Watershed Area
Determined from
Subsequent
LiDAR-Dervived
3-foot DEMs
Gray = Area Present in both coverages; Red = Area Present in Year 1 but not Year 2; Blue
Area Present in Year 2 but not Year 1.
XCVI
-------
Appendix Nine: Examples of Vegetation Classes used in the LiDAR Accuracy Assessment
Ground Truth Survey Area.
.--
Goo
3"
-stream ground conditions have open,
level, and firm to hard surfaces:
Good" near-stream vegetative conditions have
few or no overhanging branches:
X
photo by K. Brubaker - UMD • Mar-2006
XCVII
-------
XCVIII
-------
OK" near-stream vegetative conditions have
relatively little ground vegetation:
I photo by K. Brubaker - UMD - Mar-2006
OK near-stream vegetative conditions have
relatively more overhanging branches:
i *t . ~m
4 A
Jamagin 2009 EPA LEB
XCIX
-------
OK" near-stream vegetative conditions have more
irregular and softer ground conditions:
photo by K. Bmbaker - UMD - Mar
Bad" near-stream vegetative conditions have
dense underbrush:
-------
Bad near-stream vegetative conditions have a
relatively dense overhanging branches:
photo by K. Brubaker - UMD • Mar-2006
CI
-------
-------
&EPA
United States
Environmental Protection
Agency
Office of Research
and Development (8101R)
Washington, DC 20460
Official Business
Penalty for Private Use
$300
EPA/600/R-10/120
October 2010
www.epa.gov
Please make all necessary changes on the below label,
detach or copy, and return to the address in the upper
left-hand corner.
If you do not wish to receive these reports CHECK HERE D;
detach, or copy this cover, and return to the address in the
upper left-hand corner.
PRESORTED STANDARD
POSTAGE & FEES PAID
EPA PERMIT No. G-35
Recycled/Recyclable
Printed with vegetable-based ink on
paper that contains a minimum of
50% post-consumer fiber content
processed chlorine free
------- |