Content uploaded by Michael Curran
Author content
All content in this area was uploaded by Michael Curran on Jul 24, 2019
Content may be subject to copyright.
PRACTICAL ARTICLE
Spatially balanced sampling and ground-level imagery
for vegetation monitoring on reclaimed well pads
Michael F. Curran1,2,3,4 , Samuel E. Cox5, Timothy J. Robinson6, Blair L. Robertson7,
Karen J. Rogers8, Zoe A. Sherman9, Todd A. Adams10, Calvin F. Strom1, Peter D. Stahl1,2,3
Land reclamation associated with natural gas development has become increasingly important to mitigate land surface
disturbance in western North America. Since well pads occur on sites with multiple land use and ownership, the progress
and outcomes of these efforts are of interest to multiple stakeholders including industry, practitioners and consultants,
regulatory agents, private landowners, and the scientic community. Reclamation success criteria often vary within, and
among, government agencies and across land ownership type. Typically, reclamation success of a well pad is judged by
comparing vegetation cover from a single transect on the pad to a single transect in an adjacent reference site and data are
collected by a large number of technicians with various eld monitoring skills. We utilized “SamplePoint” image analysis
software and a spatially balanced sampling design, called balanced acceptance sampling, to demonstrate how spatially explicit
quantitative data can be used to determine if sites are meeting various reclamation success criteria and used chi-square tests
to show how sites in vegetation percent cover differ from a statistical standpoint. This method collects eld data faster than
traditional methods. We demonstrate how quantitative and spatially explicit data can be utilized by multiple stakeholders,
how it can improve upon current reference site selection, how it can satisfy reclamation monitoring requirements for multiple
regulatory agencies, how it may help improve future seed mix selection, and discuss how it may reduce costs for operations
responsible for reclamation and how it may reduce observer bias.
Key words: balanced acceptance sampling, chi-square, multiple land use, reclamation success, reference site, SamplePoint,
stakeholders
Implications for Practice
•Reclamation monitoring associated with natural resource
development involves many stakeholders and providing
standard monitoring protocol to satisfy a variety of needs,
such as delivering useful information to regulatory agen-
cies and data to improve future reclamation practices,
should be strived for.
•Many technicians are necessary to conduct vegetation
monitoring of reclaimed sites associated with natural
resource development, providing techniques which are
rapid and which may reduce observer bias which is critical
to obtain meaningful data.
•As GPS and camera technologies continue to advance
and become more accessible, incorporating them into
monitoring programs are likely to improve data reliability
and reduce costs for operators responsible for acquiring
eld data.
Introduction
Natural gas development in western North America has resulted
in various forms of environmental degradation (Finn & Knick
2011). Land reclamation ensures land surface disturbances do
not have long-term negative impacts (Wortley et al. 2013).
Aside from assisting environmental recovery, reclamation may
have legal implications, and denitions of successful reclama-
tion may vary among a wide range of stakeholders (Zedler
2007). In Wyoming, reclaimed well pads are compared to
adjacent reference areas and subject to various regulatory
Author contributions: MFC, SEC, BLR, TAA, CFS, PDS conceived and designed
research; MFC, ZAS collected eld data; MFC, TJR analyzed data and statistics;
MFC, KJR analyzed data and policies; MFC wrote rst draft; all authors edited
subsequent drafts of manuscript.
1Wyoming Reclamation and Restoration Center, University of Wyoming, 1000 E.
University Avenue, Department 3354, Laramie, WY 82071, U.S.A.
2Department of Ecosystem Science and Management, University of Wyoming, 1000
E. University Avenue, Department 3354, Laramie, WY 82071, U.S.A.
3Program in Ecology, University of Wyoming, 1000 E. University Avenue, Berry
Center 231, Laramie, WY 82071, U.S.A.
4Address correspondence to M. F. Curran, email mcurran2@uwyo.edu
5Bureau of Land Management, Wyoming State Ofce, 5353 Yellowstone Road,
Cheyenne, WY 82009, U.S.A.
6Department of Statistics, University of Wyoming, 1000 E. University Avenue, Depart-
ment 3332, Laramie, WY 82071, U.S.A.
7School of Mathematics and Statistics, University of Canterbury, Private Bag 4800,
Christchurch, 8140, New Zealand
8Wyoming Game and Fish Department, 5400 Bishop Boulevard, Cheyenne, WY
82006, U.S.A.
9School of Energy Resources, University of Wyoming, 1000 E. University Avenue,
Department 3012, Laramie, WY 82071, U.S.A.
10Environmental Field Technician Services, 85 Meadow Plains Road, Laramie, WY
82070, U.S.A.
© 2019 Society for Ecological Restoration
doi: 10.1111/rec.12956
Supporting information at:
http://onlinelibrary.wiley.com/doi/10.1111/rec.12956/suppinfo
Restoration Ecology 1
Vegetation monitoring on reclaimed well pads
criteria from Bureau of Land Management (BLM) local eld
ofce, inter-agency ofces or the Wyoming Department of
Environmental Quality’s (WYDEQ) Storm Water Pollution
Prevention Plan (SWPPP), and by the State of Wyoming’s
Sage-grouse Executive Order (SGEO) habitat suitability crite-
ria to evaluate management efforts for a species of concern such
as the Greater Sage-grouse (Centrocercus urophasianus). Moni-
toring plans are therefore often narrowly tailored to comply with
specic local or agency policies, and lack species-specic infor-
mation (Curran & Stahl 2015). This focus may limit advance-
ment of scientic knowledge, understanding of environmental
issues, and may result in operators investing money in unproven
management practices such as seeding reclamation areas with
poor seed mixes (Lovett et al. 2007). Developing an integrated
monitoring program, having distinct monitoring objectives and
improved decision-making capabilities for multiple stakehold-
ers, will likely increase the value of collected data (e.g. Toevs
et al. 2011).
Monitoring metrics often vary based on regulatory criteria.
Monitoring timing and methods can vary among eld tech-
nicians and are often inconsistent over time (Curran & Stahl
2015). Well pads undergoing reclamation have been monitored
using Daubenmire frames (Daubenmire 1959), ocular estimates,
and line point intercept (LPI) methods (Coulloudon et al. 1999)
to compare a reclaimed site area to an adjacent reference site.
Due to the large amount of natural resource development in
western United States, many technicians with various capa-
bilities are conducting vegetation monitoring. Both Dauben-
mire frame methods and ocular estimates are vulnerable to high
observer bias (Kennedy & Addison 1987; Bergstedt et al. 2009;
Morrison 2016). A typical quantitative method to measure veg-
etation cover on a reclaimed well pad versus a reference site is
to place a single transect in each area and conduct quantitative
analyses using the LPI technique (Curran & Stahl 2015). Several
pitfalls exist with a single transect comparison. First, transect
data spatial information may be suspect, even when transects are
permanently marked, because there is no location-verication
mechanism to ensure the data recorded on a page actually
came from the intended transect. We have seen many examples
of user-error resulting in mislabeled transect data, a mistake
impossible to rectify since there exists no independent spatial
information. Where temporal transect data change dramatically
and abruptly, a mislabeled transect is often suspected, leav-
ing managers wondering how to proceed. Second, LPI methods
are time and labor intensive, making them expensive for large
projects (West 1999). Due to LPI time requirements, pheno-
logical changes in some species may result in misidentica-
tion or underestimation (Cagney et al. 2011). Third, LPI may
overestimate standing cover in rangeland ecosystems (Cook &
Stubbendieck 1986). Fourth, since most reclaimed well pads are
drill seeded in rows, a linear primary sampling frame with sys-
tematic secondary samples (e.g. a point read every 50 cm along
a 100 m transect) inadequately estimates population variance
(Thompson 2002). Fifth, despite multiple observations along a
transect, it is still a single primary sampling unit from which
population variance cannot be estimated, thus limiting statis-
tical inference. Sixth, a single reference transect is unlikely
to capture variability of the reference ecosystem (Ruiz-Jaen &
Aide 2005) or the variability of the reclaimed area itself if spatial
heterogeneity exists (Stevens & Olsen 2004). Seventh, dramatic
temporal changes in vegetation may be due to user bias or vari-
ation in plant identication skill rather than actual changes in
vegetation, yet these biases cannot be detected or conrmed.
Finally, traditional techniques often rely on technician opinion
for transect placement, adding an additional source of observer
bias and eliminating a crucial stochastic component of sampling
design.
Alternative methods, such as image-based sampling, have
been shown to reduce monitoring costs and time while increas-
ing statistical power and reducing observer bias (Cagney et al.
2011; Duniway et al. 2011). Additionally, since nearby units
associated with environmental surveys tend to be spatially
auto-correlated, a spatially balanced sample is likely to improve
monitoring design (Stevens & Olsen 2004). In addition to reduc-
ing costs compared to traditional LPI methods, Cagney et al.
(2011) highlighted that images provide a permanent record,
which can be reevaluated if data are questionable, if software
improves, or if management objectives change. Here, we add
a spatially balanced sampling design, called balanced accep-
tance sampling (BAS) (Robertson et al. 2013, 2017), along with
geo-tagged imagery, to sample reclaimed natural gas well pads
and reference sites. BAS is a representative sample of the pop-
ulation from which it is drawn (Robertson et al. 2013). In this
study, BAS was used to generate points to take spatially bal-
anced images across the study populations, reclaimed well pads
and reference sites. We demonstrate this method’s ability to pro-
vide valid statistical comparisons, to satisfy various regulatory
criteria, to evaluate species on reclaimed sites compared to their
seed mixes, and discuss its improved time efciency, and reduc-
tion of observer bias.
Methods
Study Areas
Jonah Infill. Four well pads undergoing interim reclamation
and their adjacent reference areas were monitored in the Jonah
Inll natural gas eld in Sublette County, WY, U.S.A. (Fig. 1).
This eld is comprised almost entirely of federal land reg-
ulated by the Jonah Interagency Ofce (JIO— consisting of
Pinedale BLM Field Ofce, Wyoming Department of Agricul-
ture, WYDEQ, and Wyoming Game and Fish Department). All
well pads were located in the Stud Horse Butte (SHB) section
of this gas eld. One well pad, SHB 32-10, was seeded with
“mix B” in fall 2008, while three other sites, SHB 68-11, SHB
80-11, and SHB 118-11, were seeded with “mix B1” in fall 2013
(Tables S1 and S2, Supporting Information). Well pads aver-
aged 5.3 acres in size.
Douglas. Two well pads undergoing interim reclamation
and adjacent reference areas were monitored on private land
near Douglas, WY, U.S.A. (Fig. 1). These sites are located on a
private ranch in an oil exploration area. One well pad, Trinity
3569-16-14, was seeded with “grass mix” in fall 2016 and
2Restoration Ecology
Vegetation monitoring on reclaimed well pads
Figure 1. A location map showing where the Jonah Inll and Douglas
natural gas elds lie within Wyoming, USA.
one well pad, Trinity 3468-17-44, was seeded with “grass and
alfalfa mix” in fall 2016 (Tables S3-S4). Well pads averaged
7.2 acres in size.
Reclamation Policies
Existing reclamation policies for all study well pads were iden-
tied. All sites are required to meet WYDEQ SWPPP, which
requires plant density or cover on the reclaimed area to be
greater than 70% of an adjacent reference site with no erosion
features present on the reclaimed area (WYDEQ 2016). Since
future development surrounding the Jonah Inll may fall within
Greater Sage-grouse core habitat, we examined how those sites
compared to the SGEO reclamation criteria, which requires a
reclaimed site to have two native forb species, two native grass
species (at least one bunchgrass), and a reference site no further
than 60 m away consisting of 5% or greater sagebrush cover.
Jonah Inll sites are also required to meet the following JIO
criteria (JIO 2007):
1. The site must be in stable condition as indicated by BLM’s
Erosion Control Classication System (BLM Tech Note 346),
the percentage of bare ground must be equal to or less than the
reference site;
2. The average density or frequency of forbs on the reclaimed
site must be a minimum of 75% of the reference site, with forb
richness on the reclaimed site being equal or greater than the
reference site;
3. The average density or frequency of shrubs on the
reclaimed site must be a minimum of 50% of the reference site,
with shrub richness on the reclaimed site being equal or greater
than the reference site;
4. The reclaimed site must have a minimum of three native
perennial grass species present, with at least two being native
bunchgrasses;
5. Sites must be free from all species listed on WY or federal
noxious weed lists; and
6. Plants must be resilient as evidenced by well-developed
root systems, owers, and seed heads.
Qualitative analysis on site was recorded for erosion criteria
to determine if any rills, gullies, or other major erosion fea-
tures were present. Simple division was used to determine if
reclaimed sites passed various criteria which compare percent
cover against the reference area for WYDEQ and JIO require-
ments. Vegetation presence counts were used to determine if
species richness criteria were met. Flowers (forbs) and seed
heads (grasses) were observed in imagery, and if both existed it
was assumed that root systems were developing well, although
root development is not typically measured using traditional
techniques or on reclamation sites since redisturbance would
be required. Sage-grouse Executive Order criteria were eval-
uated rst by determining the amount of sagebrush cover in
the reference area (if >5%, the reclaimed area was evaluated)
and then using forb and grass richness information from the
reclaimed site.
Vegetation Sampling and Analysis
A two-stage sampling design was utilized in both study areas
(i.e. reclaimed and reference sites) where images served as the
primary sampling units, and pixels within images served as sec-
ondary sampling units to estimate vegetation cover. This data
structure is considered a product multinomial design, and dif-
ferences between reclaimed and reference areas were deter-
mined using chi-square tests. If the pixels used to classify
vegetation cover were equally distributed among functional
groups on reclaimed versus reference sites, sites would not be
considered statistically different, whereas if sites were signi-
cantly different the chi-square analyses show which categories
are responsible for the difference. Chi-square analyses were
conducted to compare reclamation and reference areas using
the following groups: bare ground, herbaceous litter, woody lit-
ter, native forb, native shrub, bunchgrass, rhizomatous grass,
and weeds.
Prior to eld work, aerial imagery (ESRI World Imagery,
Digital Globe) in ArcMap v10.3 (ESRI, Redlands, CA, U.S.A.)
was used to draw polygons around interim reclamation portions
of each well pad, and a buffer polygon was placed around the
outer perimeter of the well pad (Fig. 2). In Jonah Inll, a 30 m
wide undisturbed zone around each reclaimed well pad was used
as a reference area. Primary sampling locations were gener-
ated quasi-randomly by ArcMap so that the minimum distance
between two locations was greater than 10 m. Sample sizes of
40 were used on both reclaimed well pads and reference areas.
Reclaimed well pads in the Douglas area were surrounded by
60 m wide undisturbed zones (reference areas) and 30 primary
sampling locations (BAS points) were drawn from both the
reclaimed well pads and reference areas using package “SDraw”
(McDonald 2016) and transferred from Program R to ArcMap
using package “rgdal” (R Core Team 2016; Bivand et al. 2017).
All primary sampling locations (either ArcMap generated points
or BAS points) were transferred from ArcMap into ArcPad v.
10.2 and loaded into a Juno 5 handheld GPS unit (Trimble, Sun-
nyvale, CA, U.S.A.).
Study areas were sampled on the ground in the Jonah Inll on
16 July, 2016 and in Douglas on 1 August, 2017. At each site, the
Restoration Ecology 3
Vegetation monitoring on reclaimed well pads
Figure 2. Aerial view of well pad trinity 3468-17-44. A 60 m buffer was
placed around the well pad and 30 BAS points were selected on the interim
reclamation area and reference area. The long-term disturbance area of the
well pad and a pipeline leaving the well pad were not part of the sampling
frame.
Juno 5 GSP was used to locate preselected BAS points. At each
point, a ground-level image was acquired with a 12-megapixel
Ricoh G700SE camera with a GPS unit attached (Ricoh, Tokyo,
Japan). The camera was held over the point in a nadir position
1.3 m above ground level to acquire a 0.40 mm ground sam-
ple distance (GSD) image with a 0.5 m2footprint. This reso-
lution is similar to other studies using SamplePoint (e.g. Booth
et al. 2010). Images were acquired in sunny conditions during
daytime hours and shadowing was mitigated by the photog-
rapher adjusting body position. Image collection times were
recorded with a stopwatch.
Upon returning from the eld, images were associated with
their corresponding BAS point using GeoJot+Core v. 2.4.11, a
computer software package designed to extract metadata from
geo-tagged images (Geospatial Experts, Thornton, CO, U.S.A.).
The images were then evaluated in SamplePoint, a free software
that facilitates manual, pixel-based digital image classication
(Booth et al. 2006). The software selects a user-dened num-
ber of sample pixels from an image, either systematically or
randomly, and displays a crosshair around that pixel on-screen.
Using the surrounding pixels for context, an observer is able to
classify each sample pixel by clicking on 1 of 30 user-dened
buttons to certify the individual pixel classication. The soft-
ware records observer classications to a database with each
click. We used SamplePoint v 1.59 to classify 50 pixels in each
image as either bare ground, herbaceous litter, woody litter, or
plant (to species level; see tutorial on www.SamplePoint.org).
Classied pixels were randomly chosen by the software with a
200-pixel exclusion buffer around each to avoid randomly clus-
tered sample pixels. Upon completion of image monitoring for
each area of interest automated summary reports are created
within seconds as comma separated value (.csv) les (Booth
et al. 2006). Time taken to analyze images was recorded with
a stopwatch. Upon completion of image analysis, species were
manually categorized into functional groups (see above) when
necessary for regulatory analysis and for chi-square tests, but
left as species-specic when analyzing species richness. Species
richness was used to determine seed mix establishment success,
as well as the presence of nonmix species.
Results
Statistical Comparison Between Reclaimed Well Pads
and Reference Areas
Vegetation communities on reclaimed sites were different than
reference sites (p<0.0001), but categories driving differences
varied (Table 1). In Jonah Inll, all reclaimed sites had higher
bunchgrass and herbaceous litter percent cover than reference
areas, whereas all reference areas had higher shrub and woody
litter cover (Table 1). All reclaimed sites in Jonah Inll except
SHB118-11 had less bare ground than reference sites (Table 1).
Although all reclaimed sites in Jonah Inll had greater forb rich-
ness than reference sites, only SHB68-11 had higher forb cover.
Weed cover was no different between Jonah Inll reclaimed
and reference sites In Douglas, Trinity 3468-17-44 had less
bare ground and rhizomatous grass cover than its reference, and
higher herbaceous litter, shrub, and forb cover (Table 1). Trin-
ity 3569-16-14 had higher bare ground, rhizomatous grass, and
weed cover than its reference, while having less herbaceous lit-
ter and bunchgrass cover (Table 1).
Reclamation Policies
All sites, in Douglas and Jonah Inll, passed WYDEQ SWPPP
criteria (Table 2). In Jonah Inll, all four sites passed site
stability criteria, forb richness criteria, and forb density criteria,
but no site passed all JIO Regulatory Criteria (Table 3). In Jonah
Inll, three of four sites passed the SGEO Sage-grouse habitat
suitability criteria (Table 4).
Seed Mixes in Monitoring Data
In Jonah Inll, SHB 32-10 contained 8 of 15 species from
the seed mix, with 2 species present which were not included
in the seed mix (Table S1). All other sites had 12 of 21 species
from the seed mix present, 11 of which were common to all three
sites and 7 of which were absent from all three sites (Table S2).
All three sites contained three species which were not present
in the seed mix (Table S2). In Douglas, Trinity 3648-17-44
contained all seven species from the “grass and alfalfa mix” and
two species which were not in the seed mix, whereas Trinity
3569-16-14 contained four of ve species from the “grass mix”
and three species not in the seed mix (Tables S3 and S4). All
sites had at least one species on the site which was neither in
the intended seed mix nor in the reference area (Tables S1 –S4).
We did not have the ability to conduct tests to determine if these
species were a result of a contaminated seed mix, encroachment
from the references area, or were preexisting in the replaced
stockpiled topsoil.
Time
Image collection on reclaimed and reference sites in the Jonah
Inll averaged 16.2 and 17.7 minutes, respectively (n=40).
4Restoration Ecology
Vegetation monitoring on reclaimed well pads
Tabl e 1 . A table depicting which categories drove signicant differences between reclaimed sites and reference sites in chi-square tests. All sites were
signicantly different (p<0.0001), individual categories with a * are those which were responsible for signicance. Numbers in parentheses indicate the
proportion of the chi-square statistics accounted for by the given category.
Site Bare Ground Herbaceous Litter Woody Litter Shrub Forb Bunchgrass Rhizomatous Grass Weed
SHB32-10 *(28.6) *(27.4) *(8.7) *(20.1) (0.2) *(13.3) (1.4) (0.3)
SHB68-11 *(32) *(27.1) *(11.3) *(11) *(3.9) *(10.6) (2.6) (1.6)
SHB80-11 *(28) *(29) *(7.8) *(13.3) (0.6) *(6.6) *(12.6) (2)
SHB118-11 *(29.4) *(29.3) *(10.6) *(13) (2.3) *(9.25) (3.3) (2.9)
Trinity3468-17-44 *(38.9) *(40.7) (0.1) (3.6) *(5) (2.3) (6.4) (3)
Trinity3569-16-14 *(48.8) *(26.3) (0.1) (0) 2 *(10) *(8.7) *(4)
Tabl e 2 . Reclaimed well pads and from both Douglas and Jonah Inll com-
pared against Wyoming Department of Environmental Quality Stormwater
Pollution Prevention Plan. Passing criteria were considered binary (Y =yes,
N=no), and percent ground cover on the reclaimed area compared to ref-
erence area is shown in parentheses. *Denotes a site which has passed all
WYDEQ SWPPP criteria.
Site Name
Pass Erosion Control
Criteria (No Erosion
Control Features
Present)?
Pass 70% Ground
Cover Compared to
Reference Criteria?
SHB32-10* Y Y (110.3%)
SHB68-11* Y Y (107.2%)
SHB80-11* Y Y (101.5%)
SHB118-11* Y Y (93.7%)
Trinity3648-17-44* Y Y (128.9%)
Trinity3569-16-14* Y Y (91.2%)
Reclaimed and reference sites near Douglas averaged 24 and
25.1 minutes, respectively (n=30). Image analysis averaged
1.3 minutes/image, with quantitative reports completed for
areas (i.e. reclaimed sites or reference sites) taking an average
of 51.2 minutes for Jonah Inll and 38.4 minutes in Douglas.
We attribute the increased time taken to record images in
Douglas to larger well pad size.
Discussion
Since transportation is the most expensive cost of large-scale
vegetation sampling, the ability to rapidly collect data while
in the eld is critical to monitoring projects (Stohlgren et al.
1998). Traditional LPI methods have been timed to take a
two-person team 99 minutes/site (Cagney et al. 2011). The
method proposed in this study allows one individual to cap-
ture equivalent eld data in roughly one-fth that time, sug-
gesting a 10-fold increase in eld data collection efciency.
Even factoring in image analysis time, the photo-monitoring
method took less time than required for LPI eld data collec-
tion in Cagney et al.’s (2011) study. Since the proposed method
samples individual sites faster than LPI methods, fewer trips to
the study area may be required to achieve the desired sample
size, therefore reducing costs. Additionally, rapid methods to
capture data are critical in areas where phenologically sensi-
tive forb species are being increasingly recognized in conser-
vation efforts for the Greater Sage-grouse (Curran et al. 2015).
Providing a simple monitoring method which allows for data
to be collected once and utilized many times (e.g. to report to
varying regulatory agencies and to provide seed establishment
information) is likely to reduce operator costs.
Quantitative data obtained using this method are able to sat-
isfy multiple stakeholder groups and determine if reclamation
sites are meeting various regulatory criteria, while chi-square
analyses can assist in determining where signicant differences
between reclaimed sites and reference areas exist, regardless
of regulatory criteria. While the JIO has the most stringent
regulatory criteria associated with reclamation of oil and gas
disturbances in WY, data gathered from this technique can sat-
isfy other BLM eld ofce criteria, which are based on percent
ground cover and lack of noxious weeds (Curran et al. 2013).
Two metrics SamplePoint cannot currently capture, which may
have importance to wildlife species, include vegetation height
and canopy gap. Height can easily be measured with a meter
stick. Canopy gap has been measured from digital photos using
various simple techniques (Booth & Cox 2011; Karl et al. 2012).
Both height and canopy gap may be metrics which need only be
monitored when the reclaimed area is nearing bond release. It
has also been reported that image analysis may result in dif-
culties capturing vegetation under dense canopy cover (Hulvey
et al. 2018). Our study areas were not prone to this issue and
it is unlikely for reclaimed sites in semiarid sagebrush-steppe
ecosystems to have dense canopy cover during the interim recla-
mation phase.
Along with providing a rapid means of data collection, the
spatially balanced nature of this method coupled with more
primary and secondary sampling units increases the statistical
power of this method relative to the single primary sampling unit
associated with traditional LPI techniques. Quasi-random selec-
tion of sample points also reduces observer bias. Image-based
vegetation analysis has also shown to reduce inter-observer
bias compared to traditional techniques in previous studies and
allows for multiple individuals to reexamine the same image at
different times if necessary, and for images to be analyzed by
one expert regardless of how many individuals collected images
(Cagney et al. 2011), making it less time and cost intensive than
verifying observer bias along a transect. The images in this study
contain both temporal and geospatial metadata, allowing them
to be classied long after they are captured, providing bene-
t to regulatory agents who may be busy with permitting new
well pads while monitoring is being conducted and the ability
Restoration Ecology 5
Vegetation monitoring on reclaimed well pads
Tabl e 3 . Reclaimed well pads from Jonah Inll compared to Jonah Interagency Ofce Regulatory Criteria. Each category was binary (Y=yes, N =no) with
values in parentheses denoting percent (%) or richness (reclaimed site, reference site).
Site
Site
Stable
Percent Bare
Ground Cover
(≤100% of
Reference)
Desirable Forb
Richness
(≥Reference)
Percent Forb
Cover (≥75% of
Reference)
Shrub Richness
(≥Reference)
Percent Shrub
Cover (≥50%
of Reference)
Desirable Grass
Richness
(3 or >Tot a l
Grass Species,
2or>Bunch)
Free o f All
Noxious
Weeds
Plants Show
Resilience
Based on
Flowers and
Seed Heads
SHB32-10 Y Y (110.3%) Y (1, 0) Y N (1, 3) N Y Y Y
SHB68-11 Y Y (107.2%) Y (4, 1) Y Y (2, 2) N Y N Y
SHB80-11 Y Y (101.5%) Y (3, 0) Y N (2, 3) N Y N Y
SHB118-11 Y N (93.7%) Y (4, 0) Y Y (2, 2) N Y N Y
Tabl e 4 . Reclaimed well pads from Jonah Inll compared to Sage-grouse
Executive Order criteria for reclaimed sites. *Denotes a site which has
passed all SGEO criteria for sage-grouse habitat suitability.
Site
Reference
Sagebrush
Cover (≥5%)
Reclaimed
Site Forb
Richness (≥2)
Reclaimed Site
Grass Richness
(≥2, At Least 1
Bunchgrass)
SHB32-10 Y (24%) N Y
SHB68-11* Y (17%) Y Y
SHB80-11* Y (21.1%) Y Y
SHB118-11* Y (19.6%) Y Y
to ground-truth if data are questionable. The spatial precision
of the GPS units used in this study is within 3 m, though GPS
advances will result in better precision. While spatial precision
may be a drawback of this study, it has been documented that
permanent transect locations can be unreliable for repeatabil-
ity in some instances (Vittoz & Guisan 2007). Additionally, by
increasing the number of sample units, it is likely the accuracy
of our study design is greatly improved compared to a single
transect (Kowalewski et al. 2015). Species-level classication
allows for seed mixes to be evaluated on reclaimed sites. For
example, the 11 species which were present on all three Jonah
Inll sites which received seed mix “B1” are likely to be recom-
mended for future seeding, whereas the 7 species which were
absent from all sites may be dropped for future projects in the
area. Spatially explicit data can also inform decisions on weed
management and may help identify contaminated seed mixes.
This method is best suited for ecosystems where vegetation
is short enough to be captured by a hand-held camera. Using a
camera with a separate navigation unit does not optimize ef-
ciency or cost. We used two because neither unit excels at both
navigation and image acquisition, although it can be assumed a
single unit will arise. More studies should be conducted to deter-
mine the adequate or optimal sample size, which may change
over the lifecycle of a reclaimed well pad (e.g. a 1-year-old site
may be more homogeneous than a 5-year-old site and may need
less intensive sampling). Using a nonparametric bootstrap of our
representative BAS sample, we found that 40 photos resulted
in an ability to estimate the proportion of bare ground within
5% of its estimated value on the sites we studied. Additionally,
determining the optimal amount of pixels to sample within an
image may need further consideration. Here, we used 50 pix-
els, though recent studies suggest 20–30 pixels per image is
adequate to estimate vegetation at a landscape scale when multi-
ple images are acquired (Ancin-Murguzur et al. 2018). Finally,
although using BAS points in a buffer around a well pad as a
reference site improves spatial coverage and statistical analy-
ses compared to a single LPI transect, reclaimed areas in this
study are limited to being compared against their own unique
reference site, which may limit efciency.
Acknowledgments
The study was funded in majority by Wyoming Reclamation
and Restoration Center. Funding for the GPS unit was provided
by Warren Resources and Escalera Resources. David Legg pro-
vided suggestions to improve study design. We appreciate Jonah
Energy LLC and Anadarko Petroleum Company for access
to eld sites. Taylor Crow, Robert Kirkwood, and Nicholas Graf
provided support with gure creation. The authors declare no
conict of interest.
LITERATURE CITED
Ancin-Murguzur FJ, Munoz L, Monz C, Fauchald P, Hausner V (2018) Efcient
sampling for ecosystem service supply assessment at a landscape scale.
Ecosystems and People 15:33– 41
Bergstedt J, Westerberg L, Milberg P (2009) In the eye of the beholder: bias and
stochastic variation in cover estimates. Plant Ecology 204:271– 283
Bivand R, Keitt T, Rowlingson B (2017) rgdal: bindings for the geospatial data
abstraction library. R package version 1.2-7. https://CRAN.R-project.org/
package=rgdal
Booth DT, Cox SE (2011) Art to science: tools for greater objectivity in resource
monitoring. Rangelands 33:27– 34
Booth DT, Cox SE, Berryman RD (2006) Point sampling digital imagery with
‘SamplePoint’. Environmental Monitoring and Assessment 123:97– 108
Booth DT, Cox SE, Teel D (2010) Aerial assessment of leafy spurge. Native
Plants Journal 11:327– 339
Cagney J, Cox SE, Booth DT (2011) Comparison of point intercept and image
analysis for monitoring rangeland transects. Rangeland Ecology & Man-
agement 64:309– 315
Cook CW, Stubbendieck J (1986) Range research: basic problems and tech-
niques. Society for Range Management, Denver, Colorado
Coulloudon B, Eshelman K, Gianola J, et al. (1999) Sampling vegetation
attributes. Denver, Colorado: Interagency Technical Reference 1734–4.
Curran MF, Stahl PD (2015) Database management for large scale restoration
projects in Wyoming: developing better data acquisition, monitoring and
6Restoration Ecology
Vegetation monitoring on reclaimed well pads
models for application to future projects. Journal of Environmental Solu-
tions for Oil, Gas, and Mining 1:31– 43
Curran MF, Wolff BJ, Stahl PD (2013) Approaching oil and gas pad reclamation
with data management: a framework for the future. Journal American
Society of Mining and Reclamation 2:195– 204
Curran MF, Crow TM, Hufford KM, Stahl PD (2015) Forbs and greater
sage-grouse habitat restoration: suggestions for improving commercial
seed availability and restoration practices. Rangelands 37:211– 216
Daubenmire RF (1959) A canopy-coverage method of vegetational analysis.
Northwest Science 33:43– 64
Duniway MC, Karl JW, Schrader S, Baquera N, Herrick JE (2011) Rangeland
and pasture monitoring: an approach to interpretation of high-resolution
imagery focused on observer calibration for repeatability. Environmental
Monitoring and Assessment 184:3789– 3804
Finn SP, Knick ST (2011) Changes in the Wyoming Basins landscape from
oil and natural gas development. Pages 46– 68. In: Sagebrush ecosystem
conservation and management: ecoregional assessment tools and models
of the Wyoming Basins. Allen Press, Lawrence, Kansas
Hulvey KB, Thomas K, Thacker E (2018) A comparison of two herbaceous cover
sampling methods to assess ecosystem services in high-shrub rangelands:
photography-based grid point intercept (GPI) versus quadrat sampling.
Rangelands 40:152– 159
Jonah Interagency Ofce (2007) Reclamation criteria: Jonah interagency ofce
recommendations. https://www.wy.blm.gov/jio-papo/jio/reclamation/
reclamation_criteria_020807.pdf (last accessed 12 Jan 2018)
Karl JW, Duniway MC, Schrader TS (2012) A technique for estimating range-
land canopy-gap size distributions from high-resolution digital imagery.
Rangeland Ecology & Management 65:196– 207
Kennedy KA, Addison PA (1987) Some considerations for the use of visual
estimates of plant cover in biomonitoring. Journal of Ecology 75:151– 157
Kowalewski LK, Chizinski CJ, Powell LA, Pope KL, Pegg MA (2015) Accuracy
or precision: implications of sample design and methodology on abundance
estimation. Ecological Modeling 316:185– 190
Lovett GM, Burns DA, Driscoll CT, Jenkins JC, Mitchell MJ, Rustad L, Shanley
JB, Likens GE, Haeuber R (2007) Who needs environmental monitoring?
Frontiers in Ecology and the Environment 5:253– 260
McDonald T (2016) SDraw: spatially balanced sample draws for spatial objects.
R package version 2.1.3. https://CRAN.R-project.org/package=SDraw
Morrison LW (2016) Observer error in vegetation surveys: a review. Journal of
Plant Ecology 9:367– 379
R Core Team (2016) R: a language and environment for statistical computing.
R Foundation for Statistical Computing, Vienna, Austria. https://www.R-
project.org/
Robertson BL, Brown JA, McDonald T, Jaksons P (2013) BAS: balanced
acceptance sampling of natural resources. Biometrics 69:776–784
Robertson BL, McDonald T, Price CJ, Brown JA (2017) A modication
of balanced acceptance sampling. Statistics & Probability Letters 109:
107– 112
Ruiz-Jaen MC, Aide TM (2005) Restoration success: how is it being measured?
Restoration Ecology 13:569– 577
Stevens DL, Olsen AR (2004) Spatially balanced sampling of natural resources.
Journal of the American Statistical Association 99:262– 278
Stohlgren TJ, Bull KA, Otsuki Y (1998) Comparison of rangeland vegetation
sampling techniques in the central grasslands. Journal of Range Manage-
ment 51:164– 172
Thompson SK (2002) Sampling: Wiley series in probability and statistics. John
Wiley and Sons, London, United Kingdom
Toevs GR, Karl JW, Taylor JJ, Spurrier CS, Karl M, Bobo MR, Herrick JE (2011)
Consistent indicators and methods and a scalable sample design to meet
assessment, inventory, and monitoring needs across scales. Rangelands
33:14– 20
Vittoz P, Guisan A (2007) How reliable is the monitoring of permanent vegeta-
tion plots? A test with multiple observers. Journal of Vegetation Science
18:413– 422
West NE (1999) Accounting for rangeland resources over entire landscapes.
Pages 726– 736. In: Eldridge D, Freudenberger D (eds). Proceedings of the
VI International Rangeland Congress, July 1999, Aitkenvale, Queensland,
Australia
Wortley L, Hero JM, Howes M (2013) Evaluating ecological restoration success:
a review of the literature. Restoration Ecology 21:537– 543
Wyoming Department of Environmental Quality (2016) Storm Water Per-
mitting. https://deq.wyoming.gov/wqd/storm-water-permitting/resources/
construction-general-permits/
Zedler JB (2007) Success: an unclear, subjective description of restoration
outcomes. Ecological Restoration 25:162– 168
Supporting Information
The following information may be found in the online version of this article:
Table S1. List of species from the seed mix and presence (P) or absence (A) as well as
list of species not in the seed mix which were present on site Stud Horse Butte 32-10.
Table S2. List of species from the seed mix and presence (P) or absence (A) as well as
list of species not in the seed mix which were present on sites Stud Horse Butte 68-11,
80-11, and 118-11.
Table S3. List of species from the seed mix and presence (P) or absence (A) as well
as list of species not in the seed mix which were present on site Trinity 3468-17-44.
Table S4. List of species from the seed mix and presence (P) or absence (A) as well
as list of species not in the seed mix which were present on site Trinity 3569-16-14.
Coordinating Editor: Purushothaman Abhilash Received: 25 March, 2018; Firstdecision: 25 June, 2018; Revised: 8 April, 2019;
Accepted: 9 April, 2019
Restoration Ecology 7