ArticlePDF Available

Abstract and Figures

Energy-extraction results in significant disturbance to rangelands in Wyoming and other western US states. Although reclamation is required by law, US General Accounting Office reports from 1999 and 2005 are clear that affected government agencies have--over much of the past decade--had difficulty accomplishing mandated environmental monitoring of extraction-related disturbance. We evaluated two pipeline rights of way (ROW) using nested images (1- or 2- with 13- or 20-mm ground sample distance (GSD)) acquired during Very-Large Scale Aerial (VLSA) surveys. Aerial monitoring allowed for the collection of large numbers of geocoded samples, and for subsequent cover measurements using methods with demonstrated accuracy equal to that of conventional ground-based methods. Both pipelines had vegetative-cover deficiencies relative to their Plan of Development (POD) requirements. Using bare ground and ground-cover measurements from the higher-resolution imagery, we present a spatial representation of each pipeline ROW that allows quick identification of sections of the ROW that may need further reclamation action to meet POD standards. We also present aerial monitoring costs. We recommend VLSA pipeline surveys as a means for facilitating required environmental monitoring and for addressing the monitoring backlog that has developed with increased energy-extraction activity.
Content may be subject to copyright.
Environ Monit Assess
DOI 10.1007/s10661-008-0562-5
Dual-camera, high-resolution aerial assessment
of pipeline revegetation
D. Terrance Booth ·Samuel E. Cox
Received: 23 April 2008 / Accepted: 11 September 2008
© US Government 2008
Abstract Energy-extraction results in significant
disturbance to rangelands in Wyoming and other
western US states. Although reclamation is re-
quired by law, US General Accounting Office
reports from 1999 and 2005 are clear that affected
government agencies have—over much of the
past decade—had difficulty accomplishing man-
dated environmental monitoring of extraction-
related disturbance. We evaluated two pipeline
rights of way (ROW) using nested images (1-
or 2- with 13- or 20-mm ground sample distance
(GSD)) acquired during Very-Large Scale Aer-
ial (VLSA) surveys. Aerial monitoring allowed
for the collection of large numbers of geocoded
samples, and for subsequent cover measurements
using methods with demonstrated accuracy equal
to that of conventional ground-based methods.
Both pipelines had vegetative-cover deficiencies
relative to their Plan of Development (POD) re-
quirements. Using bare ground and ground-cover
measurements from the higher-resolution im-
agery, we present a spatial representation of each
pipeline ROW that allows quick identification of
sections of the ROW that may need further recla-
mation action to meet POD standards. We also
D. T. Booth (B)·S. E. Cox
USDA-ARS, High Plains Grasslands Research
Station, 8408 Hildreth Road, Cheyenne,
WY 82009, USA
e-mail: terry.booth@ars.usda.gov
present aerial monitoring costs. We recommend
VLSA pipeline surveys as a means for facilitating
required environmental monitoring and for ad-
dressing the monitoring backlog that has devel-
oped with increased energy-extraction activity.
Keywords Aerial photography ·Aerial
monitoring costs ·Environmental monitoring ·
Ground-cover measurements ·Point sampling ·
Spatial data
Introduction
Energy extraction is a prominent use of public
lands in Wyoming and many western U.S. states.
In 2006, oil and gas operations on US Bureau
of Land Management (BLM) administered pub-
lic lands and federal mineral estate in Wyoming
produced 28.5 million barrels of oil and 525.4
trillion cubic feet of gas, generating $1 billion
in federal oil and gas royalties (BLM 2007). Ex-
traction and associated activities result in signif-
icant disturbance to the soil and vegetation of
affected lands. Although interim reclamation is
required in some instances and monitoring of dis-
turbances and land use was mandated by the 1968
National Environmental Policy Act (NEPA) and
other laws, there is evidence that manpower and
conventional monitoring technology have been in-
sufficient to allow effective ecological monitoring
Environ Monit Assess
of disturbances (GAO 1999,2005; Bisson 2007).
In 1999, a US General Accounting Office (GAO)
report noted that the Department of the Interior
(DOI) has the challenge of “. . . striking a balance
between its two basic mandates—accommodating
the demands for greater use and more consump-
tion of resources with the demands to protect and
conserve resources for the benefit of future gen-
erations.” The report observed that DOI agencies
often cannot identify the environmental impacts
of energy extraction; instead of fact-based deci-
sions, subjective judgments are used in assess-
ing how limited budgetary resources should be
spent protecting and preserving environmental
resources.
Between 1999 and 2004, drilling permits issued
by the BLM more than tripled. A 2005 GAO
report described BLM’s ability to meet its en-
vironmental mitigation responsibilities as being
substantially lessened as a result of increases in
drilling-permit workloads. Oil and gas-related re-
vegetation monitoring has followed conventional
ground-based methods (Alberta Environment
2004; Interagency Technical Team (ITT) 1996;
Vogel 1987), but these methods have a low data-
yield-to-cost ratio relative to remote sensing me-
thods (Um and Wright 1998). Satellite imagery is
available from a variety of sources, but is often
much more costly than aerial photography and
maximum resolution at present is 60 cm ground
sample distance (GSD) (Quickbird1), a resolution
too low to monitor anything but gross changes in
vegetative cover.
The need for cost-effective, statistically-
adequate monitoring allowing unbiased mea-
surements of key environmental indicators has
motivated more than 40 years of research in
high-resolution aerial monitoring. Aldrich et al.
(1959) introduced the Hulcher 70 mm camera
for monitoring natural resources. The Hulcher
allowed shutter speeds of 1/2,000 s and was the
preferred aerial camera for large-scale (1:600 to
1Throughout this paper the mention of trade names or
commercial products is for information only and does
not imply recommendation or endorsement by the U.S.
Department of Agriculture.
1:10,000) natural-resource assessments for more
than two decades. Booth (1974) introduced inter-
mittent aerial sampling while using the Hulcher
to obtain photographs at scales near 1:1,000 for
evaluating soil erosion in Nevada rangelands.
Abel and Stocking (1987)alsousedintermittent
aerial sampling (scale = 1:2,500, 35 mm camera)
from which they estimated sediment yield from
Botswana rangelands. The equipment and
methods for obtaining high-resolution aerial
photography from fixed-wing aircraft remained
more or less static from 1959 until Booth and Cox
(2006) used a Sport airplane (FAA 2004)anda
modified Hulcher 70 mm film camera to acquire
aerial photography with minimal motion blur at
scales near 1:200, with subsequent resolutions of
1-mm GSD using a digital camera equipped with
an image-stabilized telephoto lens (Booth and
Cox 2008).
Advances in digital photography have over-
come many of film’s shortcomings and allow
high-resolution aerial imaging to be a viable, eco-
nomical monitoring tool for pipeline-reclamation
and other natural-resource monitoring. Develop-
ment of methods for acquiring and analyzing dig-
ital aerial imagery with resolutions comparable to
what can be obtained on the ground offers the
potential for substantial increases in monitoring
efficacy, the creation of a permanent record of
conditions at a specified time and place, and re-
ductions in costs (Booth et al. 2005a,b,2006a,b,c;
Luscier et al. 2006). Here we assessed reclamation
success relative to right-of-way (ROW) permitting
requirements to evaluate the utility of reclamation
monitoring using very-large scale aerial (VLSA)
imagery (1 to 20-mm ground sample distance
(GSD)).
Methods
Image acquisition
VLSA surveys are designed to systematically
sample an area of interest by acquiring numer-
ous images at regular intervals over the sur-
vey area (Booth and Cox 2006). VLSA is a
sampling tool, not a mapping tool. We used
Environ Monit Assess
an externally-mounted dual-camera equipment
module consisting of two digital SLR cameras
(1Ds 11.1 megapixel with a 100 mm, f/2.8 lens;
1DsMarkII 16.7 megapixel with an 840 mm, image
stabilized, f/5.6 lens; Canon USA, Lake Success,
NY) set for 1/4,000 s shutter speed, a Trackair
navigation system (Trackair, Oldenzaal, Nether-
lands), an LD90-3100VHS-FLP laser rangefinder
used as an altimeter (Riegl, Orlando, FL), a
401036 light meter (Extech, Waltham, MA) and
a GPS16 WAAS-enabled GPS receiver (Garmin,
Olathe, KS). During the first year of the study
(2006), three laptop computers (Dell, Round
Rock, TX) were on board the aircraft to run the
navigation system and collect images sent over
firewire from the cameras. The following year
(2007) images were all saved to 16 GB compact
flash cards (SanDisk, Milpitas, CA), so only one
laptop was required on board to handle naviga-
tion. A 10-cm LCD screen located near the pi-
lot provided a graphic display of the flightlines
and targets, as planned using ArcGIS 9.0 (ESRI,
Redlands, CA). The cameras were automatically
fired by the navigation system and images and
location data were captured concurrently. Alti-
tude above ground level (AGL) and light intensity
values were collected by a custom logger that
simultaneously displayed the information for the
pilot on the LCD monitor (Robert Berryman
Consulting, Boulder, CO). Because of its capa-
bility for safe, slow flight (65–95 kph), we used a
Moyes–Bailey Dragonfly powered by a Rotax 4-
stroke, 115-horsepower turbocharged engine. The
two-seat aircraft has a 10.5-m wing span and a
US Federal Aviation Agency “Sport Airplane”
designation (FAA 2004). Images were saved as
RAW files on both cameras, later converted to
Tagged Image File Format (TIFF) images using
Canon Digital Photo Professional 2.0. Both cam-
eras were time-synchronized with the laptop prior
to every flight, such that all images and metadata
could be linked precisely by time. We used Merge
software (Robert Berryman Consulting) to match
images with GPS coordinates, altitude AGL, light
intensity and ground speed, the latter two being
predictors of image quality. We then used ArcGIS
9.0 (ESRI) to derive slope, aspect and elevation
for each image location.
Study sites
The Lost Creek pipeline
The Lost Creek pipeline is a 61-cm natural gas
header that runs 193 km across the Red Desert
in southern Wyoming (Fig. 1), mostly on public
land administered by the BLM, Lander and
Rawlins Field Offices. The pipeline was laid and
the 25-m wide ROW reseeded with 17 kg/ha
native grass/shrub seed mix in 2000. This study
focused specifically on a 32-km ROW segment
crossing the Arapahoe Use Area in the Green Mt
Common Allotment, south of Jeffrey City, WY
(4214N, 10754W). Most of the ROW falls
inside the 25–30 cm precipitation zone dominated
by Wyoming big sagebrush (Asteraceae Artemisia
tridentata Nutt. var. wyomingensis (Beetle and
A.L. Young) S.L. Welsh), with elevation of
2,020–2,190 m and average slope of 2.3.Anaerial
survey along the ROW was completed in June
2006, by acquiring 258 pairs of images from an av-
erage AGL of 150 m using the dual-camera VLSA
Fig. 1 Location of study sites relative to the rest of
Wyoming and the US
Environ Monit Assess
system. Each image pair consisted of one 13-mm
GSD image (36 ×54-m field of view (FOV)), and
a nested 1.3-mm GSD image (4 ×6-m FOV). Dis-
tances between aerial sample locations along the
Lost Creek pipeline ranged between 54 and 917 m
with an average separation of 140 m (±76 SD).
The survey was part of a larger aerial survey of
the area, and some images from the larger survey
were used as part of this study. Ferry distance
from landing strip to pipeline was 30 km.
The Howell pipeline
The Howell pipeline is a 41-cm CO2trunkline that
runs 188 km across central Wyoming, mostly on
public land administered by the Bureau of Land
Management Casper Field Office. The pipeline
was laid in 2003 and the 25-m wide ROW re-
seeded with 15 kg/ha native grass/shrub seed mix
in late 2004. This study focused specifically on a
35-km ROW segment crossing the Salt Creek wa-
tershed, north of Casper, WY (4323N, 10626
W). Most of the ROW falls inside the 30–36 cm
precipitation zone dominated by Wyoming Big
sagebrush, with elevation of 1,450–1,700 m and
average slope of 3.2. An aerial survey along the
ROW was completed in June 2007, by acquiring
152 pairs of images from an average AGL of 240 m
using the dual-camera VLSA system. This survey
was flown at 240-m AGL, rather than the 150-m
AGL altitude of the Lost Creek pipeline survey,
to increase the FOV of the images so that more
lower-resolution images would span the full width
of the disturbed area. Each image pair consisted
of one 21-mm GSD image (57 ×86-m FOV), and
one 2.1-mm GSD image (7 ×10-m FOV). Sample
separation distance ranged from 90 to 1,500 m and
averaged 293 m (±175 SD).The survey was part of
a larger aerial survey of the area, and some images
from the larger survey were used as part of this
study. Ferry distance from landing strip to pipeline
was 50 km.
Image analysis
For both pipelines, reclamation success is defined
in submitted Plans of Development (POD) by (1)
vegetative cover of at least 75% of adjacent land,
(2) species composition to include a high percent-
age of seeded species with natural invasion, (3)
the ability to withstand established grazing, (4)
the reproduction of reseeded plants is evident,
(5) planted woody species have >50% survival,
(6) noxious weeds are controlled and (7) vege-
tative cover mitigates visual impacts (Lost Creek
Gathering Company 2000; Howell Petroleum
Corporation 2003). No time limit exists to meet
reclamation goals.
We assessed reclamation success by measur-
ing vegetation cover by life form using Sample-
Point, a manual photo-interpretation software
program that facilitates point classification of im-
agery and has a demonstrated ground-cover mea-
surement accuracy of >90% using 1-mm GSD
imagery (Booth et al. 2006b). (Comparison mea-
surements of bare ground from 1 and 2-mm
GSD imagery suggest measurements from the
latter will be within 85% of those from the
1-mm GSD imagery (unpublished data).) For
each image, 100 points arranged in a systematic
grid were classified into one of the following
life form and ground-cover categories: INgrass,
INforb, INshrub, INlitter, INsoil, INrock, INun-
known, OUTgrass, OUTforb, OUTshrub, OUT-
litter, OUTsoil, OUTrock and OUTunknown,
where IN refers to the area inside the ROW
and OUT refers to undisturbed areas adjacent to
ROW. Wide field of view images (36 ×54 m)
covered both the ROW and adjacent areas, but
were discarded if less than 25 sample points fell in
either of these areas. Vegetation cover within and
adjacent to the ROW were thus simultaneously
measured from each image. Small field of view im-
ages (4 ×6and7×10 m) typically covered either
the ROW or adjacent area, but rarely both. Thus,
the small FOV images were divided into two
groups, such that vegetation cover was measured
within and adjacent to the ROW from sepa-
rate image sets, which were also thinned where
less than 25 sample points fell within the in-
tended sampling area. Since most small FOV im-
ages taken over the ROW hit only the ROW,
additional images from the larger aerial survey
that were adjacent to, but within 300 m of, the
ROW, were used for comparison. In images taken
over the ROW that also included areas outside
the ROW, classification points falling outside the
Environ Monit Assess
ROW were not utilized in ROW cover means (i.e.,
OUT points were not counted), and vice versa.
Since every large FOV image used for analy-
sis contained both IN and OUT ROW portions,
paired T-tests were used to compare every cover
class. Since small FOV images from IN and OUT
of the ROW were not aligned, standard T-tests
were used to compare cover means for the
Howell pipeline. Welch’s T-tests were used for
the Lost Creek pipeline IN and OUT compar-
isons because F-tests showed differences in sam-
ple variance for most cover categories. Where the
F-test was not significant, use of Welch’s T-test
did not change the reported P-value, so we used
Welch’s for consistency. All statistical results were
generated using Excel 2003 (Microsoft, Redmond,
WA) and Prism 3.0 (Graph Pad Software, San
Diego, CA). Comparing vegetative cover inside
to outside the ROW is a measure of reclama-
tion success defined by indicator 1 of the PODs.
Measurements of vegetation cover also addressed
indicators 3–7 since adequate cover is a measure
of the revegetation survival—and over the long
term, of reproduction—in the face of grazing and
other environmental stressors. Each image was
examined for weedy species (indicator 6) and
if detected, the cover measured. The ease with
which the ROW can be distinguished from adja-
cent land addresses visual impacts (indicator 7).
Evaluation of the utility of VLSA monitoring
for pipeline ROW monitoring
We compared cover measurements made from
the two image resolutions used for each pipeline
to each other (within pipelines) to determine
whether the measurements for that pipeline would
give the same result. Since sample sizes were un-
equal between small and large FOV data sets for
both study sites (Table 1), Welch’s T-tests, which
do not assume equal sample variance, were used
for comparisons. Costs for the aerial assessment
were tabulated and are reported.
Results and discussion
Image analysis
Lost Creek pipeline
Measured from the small FOV imagery, shrub
and forb cover were lower inside the Lost Creek
pipeline right-of-way (ROW) relative to the un-
disturbed, adjacent land (P<0.001), while grass
and litter cover were higher (P<0.001, Table 1).
Total vegetation cover was 22% inside the ROW
compared to 40% outside, or 55% of the adjacent
undisturbed land. Bare ground was 57% within
Table 1 SamplePoint vegetation cover measurement
means±standard deviation inside and adjacent to (within
300 m) the Lost Creek pipeline ROW, from both small
FOV imagery (4 ×6 m, 1.3-mm GSD) and large FOV
imagery (36 ×54 m, 13-mm GSD)
Small FOV T-test IN Large FOV (n= 219) T-test IN T-test FOV T-test FOV
Inside ROW Outside ROW vs OUTaInside ROW Outside ROW vs OUT comparison comparison
(n=213)(n=85)INaOUTa
Grass 17.3 ±14.2 7.8 ±10.2 <0.001 12.6 ±13.9 4.9 ±10.9 <0.001 <0.001 0.03
Forb 2.4 ±4.6 5.0 ±4.9 <0.001 0.2 ±1.2 0.3 ±3.5 0.17 <0.001 <0.0001
Shrub 1.8 ±4.6 26.8 ±11.7 <0.001 2.8 ±3.8 40.0 ±14.3 <0.001 0.01 <0.0001
Litter 21.2 ±9.8 16.7 ±7.5 <0.001 24.5 ±11.2 14.4 ±6.2 <0.001 0.001 0.01
BG 56.5 ±17.4 37.9 ±13.3 <0.001 59.5 ±17.7 36.3 ±14.0 <0.001 0.08 0.4
Rock 0.9 ±1.9 2.2 ±6.3 0.07 0.7 ±1.4 4.2 ±6.2 <0.001 0.2 0.01
P-values are given for cover comparisons between areas inside and adjacent to the ROW for each image set. Paired T-tests
were used to compare cover values from the large FOV image set. P-values are also given for comparisons of cover measured
from small and large FOV images, both inside and outside the ROW
BG bare ground
aWelch’s correction was used due to unequal sample variance
Environ Monit Assess
Environ Monit Assess
the ROW—1.5 times that of the adjacent refer-
ence area (Table 1). The low vegetative cover
(7 years from seeding) raises the question of
whether the new plant community is sufficiently
resistant to grazing, drought and other environ-
mental stresses and whether it is reproducing at
sustainable levels. Clearly, visual impacts are still
present (Fig. 2). Weeds were not detected in the
images or by on-the-ground surveys (John Likens,
BLM, Lander; personal communication, October
2007) suggesting that weeds are not currently a
threat to successful reclamation although the large
amount of bare ground relative to the undisturbed
vegetation suggests the ROW is at risk for inva-
sion by weedy species.
Howell pipeline
Total vegetation cover (grass, forb and shrub),
as measured from the small FOV imagery, was
38% inside, and 51% adjacent to, the ROW (P<
0.05, Table 2), giving the Howell pipeline ROW
an overall mean vegetation cover that is 74% of
the undisturbed adjacent land after three grow-
ing seasons. A later assessment may be needed
to determine if reproduction of reseeded plants
is adequate. Visual impacts are present (Fig. 2),
and will likely remain until shrubs mature. Weeds
were not detected in the images, suggesting that
they are not currently a threat. Shrub establish-
ment on the Howell pipeline ROW, at 74% of
the undisturbed adjacent cover, has been much
more successful than on the Lost Creek ROW, at
only 7%.
Measuring reclamation success
Several sections of the Lost Creek ROW have
low bare ground (Fig. 3). We examined slope,
aspect, vegetation community and proximity to
drainages as possible factors influencing low bare
ground, but determined that none of these fac-
Fig. 2 VLSA survey images from aLost Creek pipeline
ROW and bHowell pipeline ROW. Both panels show the
full view from the wide FOV image, the outline of the
nested small FOV image (yellow) and an enlarged portion
of the small FOV image (red)
tors were common to all points in the sections.
Nevertheless, the ability to ask and answer such
questions is important and data displays like Fig. 3
are useful because they allow land managers to
readily assess land conditions at specific locations.
Single values, like the mean for vegetation cover
along the whole extent of a ROW, are an inad-
equate measure of reclamation success, and can
be misleading since much of a pipeline ROW may
fail to meet plan of development (POD) require-
ments. On a sample by sample basis, only 14%
of the sampled area in the Lost Creek pipeline
ROW, and only 28% of the sampled area in the
Howell pipeline ROW have bare ground values
equal to or less than values for the adjacent
land. Similarly, only 20% of the Lost Creek sam-
pled area, and only 40% of the Howell sam-
pled area has total vegetation cover equal to or
greater than 75% of the vegetation cover of the
adjacent land. Sample images are a permanent
record of conditions at a specific time and place.
When compared with succeeding assessments,
they allow for a robust determination of vegeta-
tion trend and on-the-ground success or the lack
thereof.
The importance of image resolution
and the utility of VLSA imagery for spatial
monitoring of ROW revegetation
Point sampling is based on the assumption that
a point is infinitely small—without area (Cook
and Stubbendieck 1986, page 59)—and we have
previously discussed the importance of image
resolution in obtaining accurate ground-cover
measurements from images (Booth et al. 2005b,
2006b). Regardless of this theoretical basis, we are
often asked if 1-mm GSD imagery is necessary
for accurate ground-cover measurements. If 1-mm
GSD imagery is not needed, then we would ex-
pect no differences between paired ground-cover
measurements from 1- and 2-mm (small FOV)
versus 13- and 20-mm GSD (large FOV) imagery.
For the Lost Creek pipeline, bare ground values
from small and large FOV images were not dif-
ferent, for both inside and outside the ROW (P>
0.05). Rock cover values from both large and small
FOV images from inside the ROW also were not
different (P= 0.2), although rock values from
Environ Monit Assess
Table 2 SamplePoint vegetation cover measurement
means±standard deviation inside and adjacent to (within
300 m) the Howell pipeline ROW, from both small FOV
imagery (7 ×10 m, 2.1-mm GSD) and large FOV imagery
(57 ×86 m, 21-mm GSD)
Small FOV (n= 67) T-test IN Large FOV (n= 131) T-test IN T-test FOV T-test FOV
Inside Outside vs OUT Inside Outside vs OUT comparison comparison
ROW ROW ROW ROW INaOUTa
Grass 17.7 ±20.6 21.0 ±18.2 0.3 43.0 ±26.1 28.7 ±17.4 <0.001 <0.001 0.005
Forb 8.04 ±8.59 13.9 ±10.3 <0.001 1.4 ±3.2 2.7 ±4.7 <0.01 <0.001 <0.001
Shrub 12.1 ±12.2 16.4 ±12.0 0.04 10.0 ±10.1 25.1 ±12.7 <0.001 0.2 <0.001
Litter 8.9 ±8.2 6.4 ±6.2 0.04 2.4 ±4.2 2.9 ±4.7 0.2 <0.001
BG 53.1 ±23.8 40.9 ±24.9 0.004 43.2 ±21.8 40.0 ±20.3 0.05 0.005 0.8
Rock 0.2 ±0.5 0.2 ±0.6 0.79 0.1 ±0.4 0.4 ±0.9 <0.001 0.2 0.06
P-values are given for cover comparisons between areas inside and adjacent to the ROW for each image set. Paired T-tests
were used to compare cover values from the large FOV image set. P-values are also given for comparisons of cover measured
from small and large FOV images, both inside and outside the ROW
BG bare ground
aWelch’s correction was used due to unequal sample variance
both image resolutions outside the ROW were
different (P= 0.01). All other cover classes—that
is all vegetation classes—were different between
small versus large FOV images, both inside and
outside the ROW (P<0.03, Table 1).
For the Howell pipeline, bare ground was not
different between large and small FOV images
outside the ROW (P>0.05, Table 2), but was
different between large and small FOV images in-
side the ROW by about 10% (P<0.01, Table 2).
Rock cover means were not different between res-
olutions, either inside or outside the ROW (P>
0.05). Shrub cover measurements were not differ-
ent between image resolutions inside the ROW
(P= 0.2), but were different by 9% outside the
ROW (P<0.01). All other cover classes were
different between small versus large FOV images,
both inside and outside the ROW (P<0.01).
Taken together, the data show that there can be
significant disagreement between measurements
made from large FOV images (13 and 21-mm
GSD) and small FOV images (1 and 2-mm GSD).
Cover differences between the two image reso-
lutions were not consistent for any cover class,
and do not seem subject to a standard offset pre-
diction. For example, grass measurements from
large FOV compared to small FOV differed by
+25,+8, 3and 5. Such variance hinders accu-
rate compensation for inherent errors in cover
measurements made from lower resolution im-
agery. Given the significant differences in mea-
surement outcomes between the two FOVs, we
reject the null hypothesis and recommend the use
of small FOV, high resolution imagery (1 and 2-
mm GSD) for measuring vegetative cover. But,
we emphasize the value of the large FOV, lower
resolution imagery (13 and 21-mm GSD) in mea-
suring such things as disturbance width and area,
occurrence of invasive species, and for providing
an overview of the area to help put small FOV
images into context with their surroundings, or
to determine whether the small FOV image falls
inside the ROW.
Satellite images, small-scale aerial imagery and
videography, while useful to some extent for mon-
itoring, cannot provide the high resolution re-
quired to identify species (including weeds) or
allow accurate ground cover measurements. How-
ever, this shortcoming must be balanced against
the inability of VLSA imagery to acquire an un-
interrupted image of the ROW, such as is avail-
able with satellite and most small-scale imagery.
Data needs will determine the appropriate moni-
toring tool. To assure valid statistical comparison
of small FOV images IN and OUT of the ROW,
we recommend aerial acquisition directly over the
ROW, and a second acquisition flight targeted for
30 m adjacent to the ROW. We found no partic-
Environ Monit Assess
Fig. 3 Cover values for bare ground, shrub, grass, forb,
litter, rock and total vegetation (shrub, forb and grass) as
a percentage of adjacent, undisturbed land cover values,
measured from small field of view (5×8 m) aerial images
and plotted to scale, provide a quick and comprehensive
view of reclamation progress along two 35 km stretches
of pipeline rights of way in central Wyoming. Cover breaks
indicate areas that meet or do not meet the vegetative
cover requirements established by the Plans of Develop-
ment. Skips in the Howell pipeline sequence are the result
of technical problems encountered during that survey
ular advantage to acquiring more images that in-
cluded the full width of the Howell pipeline ROW
and recommend the lower flight altitude for future
surveys.
Costs
Although the total in Table 3suggests a survey
cost of $67 to $73/km of pipeline, these surveys
Table 3 Costs of
monitoring each pipeline
ROW using VLSA
imagery
Task Time (h) Cost/h ($) Cost ($)
Flight planning 4 40 160
Flight time 1 150 150
Pilot ground time 2 50 100
Technician support 3 40 120
Data compilation 3 40 120
Image analysis 100 12 1,200
Data analysis 12 40 480
Total for each ROW 2,330
Environ Monit Assess
were conducted as part of larger projects. There
are additional costs associated with the larger
projects that might inflate the cost/km. Mobiliza-
tion costs include time spent preparing the air-
plane for trailering, and re-assembly at the job
site. Travel costs include time spent en route,
lodging, and per diem [14 h mobilization + 12 h
travel = 26 h at $90/h (pilot+ground support) =
$2.340; lodging and per diem = $200/day)]. A
$6,000 annual start up cost for insurance results in
a $1,000 charge to each major project during the
field season. If completed as stand-alone projects,
the total cost for each pipeline survey, including
image analysis and data summary, would have
been about $5,700. It is also important to note
that weather delays (high wind, precipitation etc.)
will increase costs. On the other hand, the pipeline
segments surveyed in this study were both 35 km
portions of pipelines totaling hundreds of kilo-
meters in length. Surveying longer portions of
pipelines would spread the mobilization and travel
costs out, reducing the cost per kilometer.
Conclusions
The demand for energy resources has put un-
precedented pressure on public land management
agencies and on the surface resources they man-
age. This pressure has highlighted the inadequacy
of conventional ground-based ecological moni-
toring for identifying revegetation insufficiencies
of pipeline ROWs. Because conventional meth-
ods have a low data-yield-to-cost ratio, they are
not practical for obtaining the high sample num-
bers and the sample distribution needed to fully
represent and describe the condition of ROW
surface resources. If problems in ROW revege-
tation are not identified, they will not be recti-
fied. The work reported here has demonstrated
that VLSA surveys provide representative knowl-
edge through systematic, sufficient, and afford-
able sampling along the length of the ROW
disturbance. Intensive VLSA sampling docu-
mented that 7 years from seeding <1/4 of the Lost
Creek pipeline ROW meets POD requirements.
This finding should raise an alarm with the respon-
sible companies and land managers. Similarly, the
Howell pipeline, 3 years from seeding, has less
than 50% of the disturbed ROW in an agreed-to
condition.
We conclude that the lower resolution imagery
acquired for the Howell pipeline (2 vs 1, and
20 vs 13 mm GSD) provided no benefit and we
recommend acquisition of the higher resolutions
as was acquired for the Lost Creek pipeline. We
also recommend a survey flight line be directly
over the ROW with a second flight line targeted
for 30 m adjacent to the ROW.
Using VLSA surveys to conduct multi-
resolution sampling along pipeline ROWs is re-
commended for obtaining the large sample
numbers and accurate ground-cover measure-
ments needed to accomplish actionable environ-
mental monitoring. The image-based data sets
created by the surveys are a permanent record,
and provide spatial evidence identifying unmi-
tigated environmental problems for each pipeline
segment, thus allowing fact-based decisions by
responsible companies and agencies for allocating
budget to protecting or rehabilitating affected
ROW surface resources.
Acknowledgements The research was funded in part by
the US Department of Interior, Wyoming State Office,
Lander Field Office and Casper Field Office of the Bureau
of Land Management. John Likins and Charlie Fifield, of
the Lander and Casper BLM Field Offices, respectively,
greatly facilitated data collection and analysis and have
provided helpful comments on the paper. The authors
thank Adele Legerski, who performed much of the image
analysis; Joe Nance of Cloud Street Flying Services, who
piloted the aircraft for data collection; Carmen Kennedy
for editorial assistance; and Mike Burns, Daniel Gredvig
and Drs. Larry Chandler, David Groeneveld, Paul
Meiman, and Mark West for reviewing the manuscript.
References
Abel, N., & Stocking, M. (1987). A rapid method for assess-
ing rates of soil erosion from rangeland: An example
from Botswana. Journal of Range Management,40,
460–466. doi:10.2307/3899612.
Alberta Environment (2004). Reclamation criteria for
wellsites, batteries and associated facilities: Draft.
Assessment protocol, (Vol. II) Retrieved from http://
www.nicholsenvironmental.com/_doclibrary/upstream_
criteria_vol_2_criteria.pdf.
Aldrich, R. C., Bailey, W. F., & Heller, R. C. (1959).
Large scale 70 mm color photography techniques
and equipment and their application to a forest
Environ Monit Assess
sampling problem. Photogrammetric Engineering,25,
747–754.
Bisson, H. (2007). Statement of Henri Bisson, Bureau of
Land Management Committee on House Natural Re-
sources Subcommittee on Energy and Mineral Re-
sources and the Subcommittee on Parks, Forests and
Public Lands Oversight hearing on Oil and Gas Im-
pacts on Public Lands. BLM.
Booth, D. T. (1974). Photographic remote sensing tech-
niques for erosion evaluations. M.S. Thesis, University
of Nevada–Reno, Reno, NV.
Booth, D. T., & Cox, S. E. (2006). Very large scale aer-
ial photography for rangeland monitoring. Geocarto,
21(3), 27–34. doi:10.1080/10106040608542390.
Booth, D. T., & Cox, S. E. (2008). Image-based mon-
itoring to measure ecological change. Frontiers in
Ecology and the Environment,6, 185–190. doi:10.
1890/070095.
Booth, D. T., Cox, S. E., & Berryman, R. D. (2006a).
Precision measurements from very large scale aer-
ial digital imagery using ImageMeasurement, Laser-
log, and Merge software applications. Environmental
Monitoring and Assessment,112, 293–307. doi:10.
1007/s10661-006-1070-0.
Booth, D. T., Cox, S. E., & Berryman, R. D. (2006b).
Point sampling digital imagery using ‘SamplePoint’.
Environmental Monitoring and Assessment,123, 97–
108. doi:10.1007/s10661-005-9164-7.
Booth, D. T., Cox, S. E., Fifield, C., Phillips, M., &
Williamson, N. (2005b). Image analysis compared
with other methods for measuring ground cover.
Arid Land Research and Management,19, 91–100.
doi:10.1080/15324980590916486.
Booth, D. T., Cox, S. E., & Johnson, D. E. (2005a).
Detection-threshold calibration and other factors
influencing digital measurements of bare ground.
Rangeland Ecology and Management,58, 598–604.
doi:10.2111/05-060R1.1.
Booth, D. T., Cox, S. E., & Simonds, G. E. (2006c). Ri-
parian monitoring using 2-cm GSD aerial photog-
raphy. Journal of Ecological Indicators,7, 636–648.
doi:10.1016/j.ecolind.2006.07.005.
Bureau of Land Management (BLM) (US Department
of the Interior) (2007). BLM oil and gas lease sale
nets nearly $14.2 million, State of Wyoming to get
half. News releases: June. Retrieved from http://www.
blm.gov/wy/st/en/info/news_room/2007/06/06ogsale.
html.
Cook, C. W., & Stubbendieck, J. (1986). Range research:
Basic problems and techniques. Denver, CO: Society
for Range Management.
FAA (Federal Aviation Administration) (2004). Airwor-
thiness certification of aircraft and related products.
Order 8130.2F.
General Accounting Office (GAO) (1999). Major man-
agement challenges and program risks: Depart-
ment of Interior. GAO/OCG-99-9. Retrieved from
http://www.gao.gov/pas/cg99009.pdf.
General Accounting Office (GAO) (2005). Oil and gas
development: Increased permitting activity has less-
ened BLM’s ability to meet its environmental pro-
tection responsibilities. GAO-05-418. Retrieved from
http://www.gao.gov/new.items/d05418.pdf.
Howell Petroleum Corporation (2003). Plan of develop-
ment. WYW148827 (p. 47). Casper, WY: Casper Field
Office, Bureau of Land Management.
Interagency Technical Team (ITT) (1996). Sampling veg-
etation attributes, interagency technical reference, re-
port no. BLM/RS/ST-96/002+1730 (p. 163). U.S.
Department of the Interior, Bureau of Land Man-
agement, National Applied Resources Science Center,
Denver, CO.
Lost Creek Gathering Company (2000). Plan of develop-
ment. WYW147148 (p. 41). Lander, WY: Lander Field
Office, Bureau of Land Management.
Luscier, J. D., Thompson, W. L., Wilson, J. M., Gorham, B.
E., & Dragut, L. D. (2006). Using digital photographs
and object-based image analysis to estimate
percent ground cover in vegetation plots. Frontiers
in Ecology and the Environment,4, 408–413.
doi:10.1890/1540-9295(2006)4[408:UDPAOI]2.0.CO;2.
Um, J. S., & Wright, R. (1998). A comparative evalu-
ation of video remote sensing and field survey for
revegetation monitoring of a pipeline route. The
Science of the Total Environment,215, 189–207.
doi:10.1016/S0048-9697(97)00340-9.
Vogel, W. G. (1987). Evaluating revegetation success. In
W. G. Vogel (Ed.), A manual for training reclamation
inspectors in the fundamentals of soils and revegetation
(chapter 7, pp. 48–63). Washington, DC: U.S. Depart-
ment of Agriculture, Forest Service and the Soil and
Water Conservation Society.
... For measuring ground cover, image resolution should be at least 1-mm resolution (Booth and Cox, 2009). Attempts to accurately measure ground-cover from 10-, 20-, and 50-mm resolution imagery were unsuccessful (Booth and Cox, 2009;Duniway et al., 2012;Weber et al., 2013). ...
... For measuring ground cover, image resolution should be at least 1-mm resolution (Booth and Cox, 2009). Attempts to accurately measure ground-cover from 10-, 20-, and 50-mm resolution imagery were unsuccessful (Booth and Cox, 2009;Duniway et al., 2012;Weber et al., 2013). Figure 11.8 shows that 1-mm resolution images are essential to capturing the detail needed for identifying species. ...
... Because image resolution is an important factor in accuracy and precision of image-based point sampling methods (Booth and Cox, 2006; Booth and Cox, 2009;Duniway et al., 2012;Weber et al., 2013), it is likely to be an important consideration for image-based frequency sampling as well. Field plots in this study were imaged at 0.48 mm GSD. ...
Article
As our understanding of ecological systems grows, natural resource management becomes ever more dependent on timely, accurate, and inexpensively-collected monitoring data to support management decisions. Vegetation cover, density, and frequency are abundance metrics used in resource management; however, frequency data can be collected more quickly than density data and with more repeatability and less sensitivity to inter- and intra-seasonal variation in plant morphology. Moreover, frequency is perhaps the best method for monitoring invasive species across extensive areas. A limitation to the use of frequency data is that plot size affects frequency. The optimal plot size is one that yields measurements suitably removed from 0 or 100% to allow detection of both upward and downward frequency trends, yet the optimum plot size cannot be known before sampling. We addressed this conundrum by developing SampleFreq software that facilitates frequency measurements from digital nadir images of any scale with up to 10 nested plot sizes within the confines of the image dimensions. We conducted accuracy and agreement tests of the software using both artificial populations and field plots. Using artificial population plots, accuracy across all users was 93.4% with a repeatability coefficient of 1.4%, indicating high precision. In a field test, SampleFreq and standard field data averaged a 3.4% difference, and were within approximately 10.5% of each other 95% of the time. From the same field test, SampleFreq repeatability coefficient was 6.7%, while the field method was 4.3%, illustrating that both methods have relatively high precision. Because SampleFreq has high potential accuracy, high agreement with field data, and high precision across a range of users, we recommend using SampleFreq with nadir digital images as a suitable alternative method for monitoring plant frequency.
... Remote sensing is a practical method for detecting surface indicators at different temporal and spatial scales within a short time frame (Sun et al., 2008). Several studies have shown that assimilating satellite remote sensing images and field data can generate surface indicators at relatively large landscape scales (Booth and Cox, 2009;Jones et al., 2018;Karl et al., 2012;Laliberte et al., 2004;Luscier et al., 2006;McCord et al., 2017). However, remote sensing techniques may have difficulty measuring all surface indicators with the required accuracy and precision (Marsett et al., 2006). ...
Article
Full-text available
Indicators of vegetation composition, vegetation structure, bare ground cover, and gap size in drylands potentially gives information about the condition of ecosystems, in part because they are strongly related to factors such as erosion, wildlife habitat characteristics, and the suitability for some land uses. Field data collection based on points does not produce spatially continuous information about surface indicators and cannot cover vast geographic areas. Remote sensing is possibly a labor- and time-saving method to estimate important biophysical indicators of vegetation and surface condition at both temporal and spatial scales impossible with field methods. Regression models based on machine learning algorithms, such as random forest (RF), can build relationships between field and remotely sensed data, while also providing error estimates. In this study, field data including over 15,000 points from the Assessment, Inventory, and Monitoring (AIM) and Landscape Monitoring Framework (LMF) programs on Bureau of Land Management (BLM) lands throughout the Western U.S., Moderate Resolution Imaging Spectroradiometer (MODIS) bidirectional reflectance distribution function (BRDF) parameters, MODIS nadir BRDF-adjusted reflectance (NBAR), and Landsat 8 Operational Land Imager (OLI) surface reflectance products with ancillary data were used as predictor variables in a k-fold cross-validation approach to RF modeling. RF regression models were built to predict fourteen indicators of vegetation cover and height, as well as bare gap parameters. The RF model estimates exhibited good correlations with independent samples, with a low bias and a low RMSE. External cross-validation showed good agreement with out-of-bag (OOB) errors produced by RF and also allowed mapping prediction uncertainty. Predicted distribution maps of the surface indicators were produced by using these relationships across the arid and semiarid Western U.S. The bias and RMSE distribution maps show that the sample insufficiency and unevenly pattern of sample strongly impact the accuracy of the RF regression and prediction. The results from this study clearly show the utility of RF as a means to estimate multiple dryland surface indicators from remotely sensed data, and the reliability of the OOB errors in assessing the accuracy of the predictions.
... In addition, RPAS have been used for the identification and mapping of weeds in agricultural (López-Granados, 2011;Peña et al., 2013;Rasmussen et al., 2013), rangeland (Hung and Sukkarieh, 2013) and aquatic landscapes (Chabot et al., 2017). Revegetation efforts have also been the subject of analysis using RPAS data in riparian areas (Dufour et al., 2013) and pipeline rights of way (Booth and Cox, 2009). High temporal resolution RPAS imagery was used for monitoring phenology of floodplain vegetation (Van Iersel et al., 2016). ...
Article
Revegetation success is a key element of mine site rehabilitation. A number of criteria related to mine site close-out are associated with revegetation. The monitoring of mine site revegetation efforts have traditionally been undertaken using field-based plot or transect methods. Often the sampling design for this monitoring is limited due to resource constraints, therefore reducing the statistical power of the data and missing information over most of the mine site. The recent advances in Remotely Piloted Aircraft Systems (RPAS) technology for remote sensing enables the collection of appropriate scale data over entire mine sites reducing the need for sampling and eliminating potential bias. This paper describes an object-based technique for extracting woody cover and estimating proportional woody cover from RPAS imagery over the rehabilitated Jabiluka mine site located in the tropical north of Australia. The technique was tested on three data sets that covered three different dates, two different sensors, and two different processing methods. Overall woody cover detection accuracies from each data set were over 95%. Proportional woody cover derived from the technique showed strong linear relationships with manually estimated cover (r2 > 0.88). This study shows that the technique is robust and works with a range of RPAS data sets and enables at scale analysis of woody cover change between dates. The technique will be an important component of ongoing monitoring of mine site revegetation in the region.
... First, there is an inherent scale mismatch between the LPI field method, which measures vegetation intercepts of a ∼1-mm- diameter point and the interpretation of images with a pixel resolution of ∼4 cm. Studies have shown that accuracy of vegetation cover estimates from image-interpretation can improve as image resolution increases (Fensham and Fairfax, 2003a; Booth and Cox, 2009), especially for coarsely-defined cover types like those used here. Second, field transects and image-interpretation transects were only approximately coincident in space, and this could result in slight differences in portions of the site that were sampled. ...
... The negative association of deer mice with proximity to pipelines may be due to the altered plant community along pipeline rights-of-way. Alterations can be quite substantial locally, given that pipe- line construction activities result in ~25-m wide corridors with decreased shrub cover and increased grass cover that persists for years following construction (Booth and Cox 2009). Within the Wyoming Basins, the extensive distribution of pipelines (Knick et al. 2011) may have a broad negative influence on deer mice. ...
Chapter
Full-text available
Understanding multi-scale floral and faunal responses to human land use is crucial for informing natural resource management and conservation planning. However, our knowledge on how land use influences sagebrush (Artemisia spp.) ecosystems is limited primarily to site-specific studies. To fill this void, studies across large regions are needed that address how species are distributed relative to type, extent, and intensity of land use. We present a study design for the Wyoming Basin Ecoregional Assessment (WBEA) to sample sagebrush-associated flora and fauna along a land cover-human land use gradient. To minimize field costs, we sampled various taxonomic groups simultaneously on transects (ungulates and lagomorphs), point counts (song birds), and area-searches of 7.29-ha survey blocks (pellet counts, burrow counts, reptile surveys, medium-sized mammals, ant mounds, rodent trapping, and vegetation sampling of native and exotic plants). We then present an exploratory approach to develop species occurrence and abundance models when a priori model building is not an option. Our study design has broad applications for large-scale evaluations of arid ecosystems.
Article
Full-text available
The recent availability of small and low-cost sensor carrying unmanned aerial systems (UAS, commonly known as drones) coupled with advances in image processing software (i.e., structure from motion photogrammetry) has made drone-collected imagery a potentially valuable tool for rangeland inventory and monitoring. Drone-imagery methods can observe larger extents to estimate indicators at landscape scales with higher confidence than traditional field sampling. They also have the potential to replace field methods in some instances and enable the development of indicators not measurable from the ground. Much research has already demonstrated that several quantitative rangeland indicators can be estimated from high-resolution imagery. Developing a suite of monitoring methods that are useful for supporting management decisions (e.g., repeatable, cost-effective, and validated against field methods) will require additional exploration to develop best practices for image acquisition and analytical workflows that can efficiently estimate multiple indicators. We embedded with a Bureau of Land Management (BLM) field monitoring crew in Northern California, USA to compare field-measured and imagery-derived indicator values and to evaluate the logistics of using small UAS within the framework of an existing monitoring program. The unified workflow we developed to measure fractional cover, canopy gaps, and vegetation height was specific for the sagebrush steppe, an ecosystem that is common in other BLM managed lands. The correspondence between imagery and field methods yielded encouraging agreement while revealing systematic differences between the methods. Workflow best practices for producing repeatable rangeland indicators is likely to vary by vegetation composition and phenology. An online space dedicated to sharing imagery-based workflows could spur collaboration among researchers and quicken the pace of integrating drone-imagery data within adaptive management of rangelands. Though drone-imagery methods are not likely to replace most field methods in large monitoring programs, they could be a valuable enhancement for pressing local management needs.
Chapter
This chapter describes the current state of the art in technology and methodology being used to develop sensors for automated weed control in cropping systems. The development of a reliable universal weed vs. crop plant sensor that works well in a wide range of crops and cropping systems is a formidable task. The discussion in this chapter highlights the significant progress that has been made in developing new, more robust, automatic sensing systems that can differentiate crop plants from weeds. Case studies documenting high levels of success in trials conducted outdoors in the natural, largely uncontrolled environment of an agricultural cropping system are presented. A discussion of the strengths and current challenges of the more successful weed and crop sensing techniques is provided. In many cases the methodology has utilized site- or condition-specific a priori knowledge to make the sensors smarter in a local context. This chapter highlights the advantages and compromises made in using these techniques. The chapter concludes with a discussion of the remaining engineering challenges to the development of a comprehensive, multifaceted fusion of several methods for sensing the differences between crops and weeds across the entire crop production cycle, and how the rapid development of advanced sensing and machine learning technologies will facilitate new plant recognition architectures and systems to achieve the level of machine recognition of weeds needed for automated weed control in cropping systems.
Chapter
Adequate assessments of vast expanses of rangeland—a primary prerequisite to effective conservation planning—requires landscape-scale evaluations that accurately represent the resources (e.g., soil, vegetation, wildlife, water), the structure and function of the resource-providing systems, and the natural range of variation in measured resource-condition indicators. We discuss why conventional rangeland survey methods are inadequate for this task and how geographic information systems (GIS), ground survey data, and high-altitude and/or satellite imagery can be used with 1- to 50-mm resolution digital imagery and new software programs to make rangeland surveys more objective, repeatable, and cost effective. We then review example applications of rangeland surveys we have conducted to answer specific landscape-scale management questions. The online exercise adds to the illustrated utility of these kinds of aerial surveys with hands-on-examples of survey planning, database queries, and how to use commonly available software in data evaluation and analyses.
Technical Report
Full-text available
The Curlew Valley sage-grouse planning area (SGPA) identifies key habitat of the greater sage-grouse (Centrocercus urophasianus) and numerous other sagebrush dependent species, and it provides for a monitoring and management focus on that habitat resource. Knowledge of habitat conditions is essential to optimize specific habitat-management activities but traditional inventory methods are inadequate for maintaining cost-effective monitoring programs that accurately represent the resource in the sense of statistically-adequate landscape sampling designs. Remote sensing technologies offer the potential of extending traditional ground sampling to produce data that is (a) more abundant, (b) more representative of inventoried landscapes, (c) more detailed for local plant communities, and (d) captures the data in permanent files that can be accessed for data validation or comparison at any subsequent date. To assess the efficacy of remote sensing methods, we compared sampling results and data obtained from aerial photography at resolutions of 1- to 30-mm ground sample distance (GSD), 0.05-mm-GSD ground photography, and field LIDAR, also known as terrestrial laser scanning / scanner (TLS). The TLS used a 532-nm green laser emitting 55,000 photons sec-1 with < 0.5-mm horizontal and vertical spacing within a 150-m radius. When multiple point clouds are surveyed from different but overlapping vantage points, shadowing and blocking are reduced and the different scans can be merged to produce 3-D data sets—including images—of the plant community. Our report is divided into three parts that describe the study site (Part I), and data acquisition methods, analysis, and results for the aerial and ground photography (Part II), and TLS (Part III). We scrutinized the use of low-altitude oblique imagery for capturing a record of understory, sage-grouse food forbs. We also examined the agreement between measurements extracted from aerial and ground imagery with traditional measurements of ground-collected data, to test the assumption that image-based plant-cover measurements from 1-mm GSD or greater resolution imagery would not yield values significantly different from traditional vegetation measurements (the null hypothesis). However, we knew of differences among ground and aerial data sets that were attributed to spatial disparity among sampling protocols; therefore, a third objective was to examine the influence of spatial disparity (i.e., data are collected in close proximity, but not from exactly the same 4 point on the ground). Fourth, the study allowed us to look at the use of aerial methods for affordable, systematic, high-density sampling across areas of interest without respect to ground accessibility or the need to make judgments in the selection of “representative” areas. This type of large-sample collecting allows important management questions to be asked of the data sets. As an example, we used a GIS overlay of lek locations to ask if lek numbers were related to habitat quality at the image locations. With respect to TLS we, 1) measured 3-D structure of sagebrush steppe plant communities, including vegetation height, density, and crown cover, 2) compared these measurements between two vegetation types (grassland and shrubland), 3) compared manual acquisition of vegetation density to automated acquisition of vegetation density using digital software, and 4) examined the potential to merge 1-mm GSD aerial imagery with TLS imagery to yield a product with greater information than either of the source images. We did not find a defining advantage for using oblique imagery to look for sage-grouse food forbs under sagebrush canopies. While there are other benefits of the oblique imagery, it may often be more valuable to use a third camera to obtain a third nadir resolution. We found that plant-cover measurements from aerial and ground imagery did not agree well, and that the traditional ground-cover measurements collected by the agencies sometimes yielded values greater that 100%, indicating some placement of Daubenmire frames in shrub understory and thus precluding a comparison of total vegetation cover with image-derived measurements of that value. Further, shrub-cover measurements did not agree well amongst methods, with the line-intercept method returning 15-25% lower shrub cover than image analysis methods. Thus, we conclude that while all methods are suitable for monitoring trend, data from these different methods cannot be compared or used as trend-data in a mixed-method data set. The accuracy of each method in this ecosystem was not determined. We found that a single ground image returns cover-measurements values that do not agree with those from 11 surrounding images, thus reinforcing spatial disparity as an explanation of differences among data collected in close proximity, but not from exactly the same point on the ground. However, we also found that over a large relatively-homogeneous area, a sufficient sample size will return the same mean values whether those measurements come from 1 image/plot or 12 images/plot. The aerial survey demonstrated the use of aerial methods for systematic, high-density sampling without respect to ground accessibility or the selection of “representative” areas; also, that as evidenced by other studies, is inexpensive relative to the costs of similar ground-based efforts. We illustrate the value of large, aerially-acquired data sets for asking management-related questions with our example use of a GIS overlay of lek locations to study lek numbers versus habitat quality at the image locations; this resulted in the unexpected finding that bare ground was greater (38 vs 25%) —and forb cover reduced (7 vs 12%) — in plots located within 3.2 km of an active lek compared to areas with no active lek within a 3.2- km radius. Although this finding is not without precedent (Sveum et al. 1998, Smith et al. 2006), it should be considered with other evidence, including the low R2 values for the reported relationships. We recommend retesting our question, perhaps by varying the area for counting leks (i.e., varying the representative area of an aerial sample) to determine if the relationship with bare ground and forb cover is constant across different sized areas and what effect—if any—the change in area has on the R2 values. Researchers working with TLS data found the software program, TreeVaW, can be used to extract plant canopy crowns for both bunch grass and shrub growth forms from the TLS point-cloud data if adjustments are made to the coefficients used to account for the differences in structure of shrubs and grasses compared to trees. TLS was demonstrated in this case, as having the capability of capturing the 3-D vegetation structure, allowing both manual and automatic measurement of vegetation parameters. The TLS work was necessarily limited to only 4 SGPA locations; however, if these findings are upheld in wider tests of TLS application to sage-grouse habitat monitoring, TLS methods might logically be considered with ground nadir imaging for replacing the labor intensive line-intercept and Daubenmire-frame methods for collecting detailed ground data to supplement aerial and other remotely sensed data.
Article
Full-text available
Ground vegetation influences habitat selection and provides critical resources for survival and reproduction of animals. Researchers often employ visual methods to estimate ground cover, but these approaches may be prone to observer bias. We therefore evaluated a method using digital photographs of vegetation to objectively quantify percent ground cover of grasses, forbs, shrubs, litter, and bare ground within 90 plots of 2m(2). We carried out object-based image analysis, using a software program called eCognition, to divide photographs into different vegetation classes (based on similarities among neighboring pixels) to estimate percent ground cover for each category. We used the Kappa index of agreement (KIA) to quantify correctly classified, randomly selected segments of all images. Our KIA values indicated strong agreement (> 80%) of all vegetation categories, with an average of 90-96% (SE = 5%) of shrub, litter, forb, and grass segments classified correctly. We also created artificial plots with known percentages of each vegetation category to evaluate the accuracy of software predictions. Observed differences between true cover and eCognition estimates for each category ranged from 1 to 4%. This technique provides a repeatable and reliable way to estimate percent ground cover that allows quantification of classification accuracy.
Article
Full-text available
Ecological assessment of ground cover by conventional on‐the‐ground point sampling is labor intensive, expensive, and biased by access. Historically, motion blur has prevented high‐resolution aerial photography from being used for ground cover measurements. To reduce motion blur we used a fixed‐wing, 225‐kg (empty weight) airplane flown at 72 km/hr ground speed 100 m above ground level with both a modified Hulcher 70 mm camera (500 mm lens, Kodak Aerocolor HS SO‐846 film, 1/4,000‐second shutter speed), and a Canon 11. 1 ‐megapixel, automatic, digital single lens reflex, color camera (420 mm focal‐length lens). The resulting very‐large scale aerial (VLSA) photography had resolutions of 5.0 and 2.1 mm GSD (Ground Sample Distance) for the scanned film and digital‐camera images, respectively. Motion blur was minimal. The cost for obtaining 450 VLSA photographs over a 70,800‐ha watershed was $0.07 ha. We found our methods well adapted for extensive aerial surveys to monitor the ecological condition of rangeland watersheds.
Article
Full-text available
Ground cover is a key indicator of rangeland health but conventional methods for measuring ground cover are labor intensive. Analysis of digital images has the potential to reduce ground-cover-measurement labor requirements. We compared cover measurements by image analyses of digital images (sensor resolution = 0.97 mm/pixel ground sample distance) with measurements derived from a laser point frame, and from two transect methods. We found there was low agreement in plot-to-plot comparisons but results were usually not different when averaged over a large number of plots or transects. We conclude that image analysis of large numbers of samples (images) produce mean values not different from conventional field methods, and, that image analysis is a superior choice for detecting relative change, since it facilitates greater data collection, reduces human bias by limiting human judgments, and provides a permanent record in images that can be retained for future scrutiny.
Article
Full-text available
New methods of image acquisition and analysis are advancing rangeland assessment techniques. Most image-analysis programs require users to adjust detection thresholds for color or object classification, a subjective process we postulated would be influenced by human error and variation. We developed a ground-cover-measurement calibration procedure, the digital grid overlay (DGO), which is similar to image point sampling (dot grid) advanced by earlier researchers. We asked 21 rangeland professionals to measure ground cover using 2 subjective visual-estimate methods (threshold adjustment process, or TAP, and external to the software visual estimate, or EVE) and the DGO on 5 different nadir-view images of rangeland. We also compared cover measurements made by DGO-calibrated software in automated batch processing against DGO manual-only measurements. We found an unacceptable range of variation among rangeland professionals using TAP. The DGO and EVE values were more closely aligned. We discovered an age-related bias in bare-ground measurements: all users over 50 years of age classified more bare ground than did all users under 50 years of age when using TAP. One explanation for this bias is age-related yellowing of the eye lens. Manual DGO measurements required up to 15 minutes per image compared to about 1 second per image for automated computer analysis after software calibration. The greatest bare-ground difference between the DGO-calibrated software and manual DGO measurements for the data sets analyzed was 5.6% and the correlations imply that reasonably accurate automated measurements can be used for bare-ground measurements from digital-image data sets. The exception is where the software cannot adequately separate litter and bare ground. The digital methods we tested need improvement. However, external calibration (DGO or EVE) of current-generation image-analysis algorithms bring economical, statistically adequate monitoring of extensive land areas within the realm of practical application.
Article
The erosion of rangeland soils is a widespread problem in Africa. Yet, there are few methods for estimating its rate. Using data from 2 catchments in Botswana, a technique for estimating erosion and sediment yield is demonstrated. It involves low level photographic sampling of vegetation cover, kriging to interpolate percentage cover from sample points, and the application of a simplified soil loss estimation procedure called SLEMSA. This modelling approach gives gross soil loss and allows the estimation of sediment yield. It is easy and cheap to apply and gave results in line with field experience.
Article
Riparian monitoring is a key aspect of sustainable resource management and is mandated by US federal law for federal land management agencies. However, it is an endeavor hampered by rising manpower costs and time-consuming travel and methods. These limitations tend to reduce sampling intensity per reach of stream and limit monitoring to the larger waterways of management units—limitations that reduce the accuracy of inferences derived from resulting data with consequential reductions in the effectiveness of landscape-level resource management. We tested the utility of low-altitude, high-resolution, intermittent aerial digital imagery for relatively inexpensive, high-intensity sampling in a watershed inhabited by the Lahonton Cutthroat trout, a species listed as threatened under provisions of the US Endangered Species Act. Measurements gleaned from the aerial imagery included late-summer open water width, number and location of late-summer dry channels, widths of riparian areas and willow coverage. All measurements were georeferenced to allow spatial data display. Riparian proper functioning condition (PFC) was assessed from the imagery by a USDI, Bureau of Land Management team. These assessments were compared to similar on-the-ground assessments made during the preceding year. PFC assessments from aerial photography were made using an average 4 staff hours per stream compared to an estimated 36 staff hours per stream for ground PFC assessments. The two assessment methods yielded roughly comparable results. We conclude that riparian-condition assessments from 2-cm GSD digital aerial imagery allowed increased sample intensity (and thus increased inference accuracy) and that it did so in our study at a cost less than half that of conventional ground-based methods. We recommend the acquisition and analysis of 2-cm GSD digital aerial imagery be further trialed for its utility and cost efficiency in ecological monitoring of riparian systems.
Article
At present, restoration monitoring of pipeline routes is based mainly on field sampling. The `integrated mapping' approach has received only limited attention from the field ecologist. The reason for this could be the largely undemonstrated value of video for the field ecologist as a monitoring tool for a corridor target, the complexity of analysing the video and the lack of research on comparative evaluation of video surveys with the field approach. An exhaustive and realistic comparison of the two techniques was conducted, based on time-lapse video data and field survey data. Both techniques have been investigated for a variety of aspects: synoptic information, ground detail and information adequacy for customer requirement. The video imagery was able to isolate the major plant communities of a narrow pipeline corridor with reliable spatial precision. Although field surveys have already gained worldwide recognition by plant ecologists as a typical method of pipeline corridor monitoring, this approach did not provide a level of information that is either scientifically reliable or economically feasible in terms of corridor site management (e.g. remedial field works). In conclusion, it is suggested that an integrated mapping approach, using aerial video in combination with supplementary ground data, can meet the requirements for monitoring revegetation of a pipeline route.
Article
High-resolution, image-based methods can increase the speed and accuracy of ecological monitoring while reducing monitoring costs. We evaluated the efficacy of systematic aerial and ground sampling protocols to detect stocking-rate differences across 130 ha of shortgrass prairie. Manual (SamplePoint) and automated (spectral) image-analysis methods were compared for both aerial and ground data. Vegetative cover changes due to grazing were detectable from 1-mm ground sample distance (GSD, a measure of resolution) digital aerial photography with as few as 30 samples yielding enough data to predict bare ground within ± 5%. We found poor agreement between automated and manual image-analysis methods, but good agreement between manual analyses of imagery from the air (100 m above ground level [AGL]) and from the ground (2 m AGL). We conclude that cover measurements made using SamplePoint from 1-mm GSD images (from 2 or 100 m AGL) can detect ecologically important changes in key indicators such as bare ground. The costs of ground and aerial methods differ markedly, and we suggest that aerial imagery is most cost effective for areas larger than 200 ha.
Article
Carbon copy of typescript. Thesis (M.S.)--University of Nevada, Reno, 1974. Includes bibliographical references.