Content uploaded by Katelin Goebel
Author content
All content in this area was uploaded by Katelin Goebel on Jul 31, 2023
Content may be subject to copyright.
Insecticide drift and impacts on arthropod prey resources of birds in public grasslands in
Minnesota
A THESIS
SUBMITTED TO THE FACULTY OF THE
UNIVERSITY OF MINNESOTA
BY
Katelin Mary Goebel
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF
MASTER OF SCIENCE
David E. Andersen
February 2021
© Katelin Mary Goebel 2021
i
Acknowledgments
Funding for this project was provided by the Minnesota Environment and Natural
Resources Trust Fund as recommended by the Legislative-Citizen Commission on
Minnesota Resources and the Minnesota Department of Natural Resources (MNDNR)
Section of Wildlife through the Federal Aid in Wildlife Restoration (Pittman-Robertson)
Act. This research was conducted through the Minnesota Cooperative Fish and Wildlife
Research Unit (U.S. Geological Survey) housed at the University of Minnesota in
partnership with MNDNR. I am deeply grateful to David Andersen and Nicole Davros
for their guidance, advice, and support through the duration of this project. Your insight
and teaching has helped me grow as a scientist and critical thinker. I thank Pamela Rice
for her support in developing data collection methods and feedback that has been
instrumental to my research. This project would not have been possible without
cooperation from landowners, renters, agronomists, cooperative representatives, and
pilots – I am grateful for their willingness to be in contact with me in the midst of busy
growing seasons. Tonya Klinkner, Hattie Saloka, and Katie Steffl provided logistical
support and I am thankful for their help. I am grateful to Sophie Crosby, Chuck Fortier,
Greg Gehring, John Letlebo, Lindsey Welch, and Haley Witt for their dedication in
collecting data in the field in all conditions. Theresa Johnston and Jessica Petersen
contributed to the design of this study. Véronique St-Louis provided valuable feedback
on statistical analyses. Lindsey Christianson and Jessica Miller spent many hours in the
lab identifying arthropod samples. Joel Alvarado, Baylee Bessingpas, Madison Fox, Kade
Friedrichs, Shalesa Johnson, Elizabeth Oechsle, Bishal Parajuli, Lydia Spann, Fadzai
Teramayi, Marae Vipond, and Jake Westfield provided their time and effort to sort and
ii
measure arthropod samples, enter data, or contribute to other essential tasks. I am grateful
to Judy Markl and the staff at Talcot Lake Wildlife Management Area who helped make
the fieldhouse feel like home. I thank the MNDNR land managers that provided insight
into choosing study sites, especially Wendy Krueger, Cory Netland, and Curt Vacek. I
am so thankful to Kaly Adkins, Marissa Cent, Annie Hawkinson, Nina Hill, Ami
Thompson, and David Wolfson for their support and friendship.
iii
Dedication
This work is dedicated to my family. Mom and Dad – for your endless support and
encouragement of my love of nature, thank you. Caro – I’m so thankful to have grown up
and spent so many hours playing outside with you. I love you.
iv
Table of Contents
Acknowledgments................................................................................................................ i
Dedication .......................................................................................................................... iii
Table of Contents ............................................................................................................... iv
List of Tables .................................................................................................................... vii
List of Figures ................................................................................................................... xii
List of Appendices ............................................................................................................ xv
Chapter 1
Grassland wildlife exposure to insecticide spray drift on public lands in Minnesota’s
farmland landscape
Overview ......................................................................................................................... 1
Introduction ..................................................................................................................... 3
Methods........................................................................................................................... 7
Study Area .................................................................................................................. 7
Landowner Contact ..................................................................................................... 9
Experimental Design ................................................................................................... 9
Measuring Insecticide Deposition with Passive Sampling Devices ......................... 10
Measuring Insecticide Deposition on Arthropods .................................................... 11
Laboratory Analysis .................................................................................................. 13
Weather Measurements ............................................................................................. 14
Vegetation Measurements ......................................................................................... 14
Researcher Safety...................................................................................................... 15
v
Data Analysis ............................................................................................................ 16
Results ........................................................................................................................... 21
Landowner Surveys .................................................................................................. 21
Insecticide Spraying .................................................................................................. 22
Insecticide Deposition on Passive Sampling Devices .............................................. 22
Insecticide Deposition on Arthropod Samples ......................................................... 25
Model Selection: Insecticide Deposition on Passive Sampling Devices .................. 27
Model Selection: Insecticide Deposition on Arthropod Samples ............................. 27
Discussion ..................................................................................................................... 28
Management Implications ......................................................................................... 34
Chapter 2
Impacts of insecticide spray drift on arthropod prey resources of birds in grasslands in
Minnesota
Overview ....................................................................................................................... 66
Introduction ................................................................................................................... 68
Methods......................................................................................................................... 71
Study Area ................................................................................................................ 71
Landowner Contact ................................................................................................... 72
Experimental Design ................................................................................................. 73
Arthropod Sampling.................................................................................................. 74
vi
Vegetation Measurements ......................................................................................... 76
Data Analysis ............................................................................................................ 77
Results ........................................................................................................................... 80
Insecticide Spraying .................................................................................................. 80
Arthropod Sampling.................................................................................................. 80
Arthropod Total Abundance Models ........................................................................ 81
Arthropod Total Biomass Models ............................................................................. 82
Bird Prey Abundance Models ................................................................................... 82
Bird Prey Biomass Models ....................................................................................... 83
Araneae Family Richness Models ............................................................................ 83
Coleoptera Family Richness Models ........................................................................ 83
Hemiptera Family Richness Models ......................................................................... 84
Orthoptera Family Richness Models ........................................................................ 84
Discussion ..................................................................................................................... 84
Management Implications ......................................................................................... 88
Literature Cited ............................................................................................................... 111
vii
List of Tables
Chapter 1
Table 1. LD50 values for chlorpyrifos for various species of birds (Solomon et al. 2001)
and pollinators. The acute contact LD50 for honey bees (Apis mellifera) is reported as 70
ng/bee in Tomlin (2000) and was converted to ng/cm2 by dividing this value by a honey
bee’s apparent exposure surface area (1.05 cm2; Poquet et al. 2014). The acute LD50 is a
common measure of acute toxicity and represents the lethal dose that causes death in 50%
of animals from a single brief exposure. Exposure to chlorpyrifos in contaminated pollen
and nectar of adult honey bees is also representative of non-Apis bee species (e.g.,
bumblebees; Cutler et al. 2014, U.S. EPA et al. 2014). .................................................... 36
Table 2. Weather, vegetation, and primary factors of interest and their hypothesized
relationships to chemical deposition. Each set of covariates constitute steps in the
hierarchical modeling process used to assess insecticide deposition on passive sampling
devices (PSDs) and arthropods in Minnesota’s farmland region. Height was included in
models of insecticide deposition on PSDs but not in models of deposition on arthropods.
........................................................................................................................................... 37
Table 3. Locations and sampling dates of study sites during the summers of 2017 and
2018 in Minnesota's farmland region. Filter paper and arthropod samples were collected
to assess insecticide deposition from adjacent soybean fields. Regions of Minnesota
sampled in this study include the southwest (SW), west central (WC), and central (C)
regions. Treatment sites were grasslands adjacent to soybean fields that were sprayed
with insecticides to control aphids; control sites were grasslands adjacent to corn fields
that were not sprayed for aphids. ...................................................................................... 38
Table 4. Spraying methods and other application information for soybean aphid spraying
events by cooperating landowners. Insecticide applications occurred on soybean fields
adjacent to treatment sites that were sampled during the summers of 2017 and 2018 in
Minnesota's farmland region. Formulated products are commercially available that
contain active ingredients and inert ingredients, and their labels provide recommended
application rates and percentages of active ingredients. Sprayer application rate refers to
the rate of tank mix applied to the field; a tank mix could include combinations of
insecticides, other pesticides, adjuvants, and solvents. Some landowners declined to
provide some information. ................................................................................................ 39
Table 5. Means and coefficients of variation (x
[CV]) of target chemicals (i.e.,
chlorpyrifos, lambda-cyhalothrin, and bifenthrin) detected on passive sampling devices
(PSDs) by distance from soybean field edge to grassland interior. Samples were collected
in Minnesota’s farmland region in 2017 and 2018. Treatment sites consisted of grasslands
adjacent to soybean fields that were treated with insecticides; control sites were
grasslands adjacent to unsprayed corn fields. PSD height refers to samplers placed at
mid-canopy height (0.5 m above ground) or ground level (0 m above ground) and these
viii
calculations included samples from treatment sites with both spray methods. Spray
method refers to whether the bordering soybean field was sprayed using an airplane or
ground boom and these calculations included samples from both mid-canopy and ground
height PSDs. Mean values are reported in ng/cm2. ........................................................... 40
Table 6. Means and coefficients of variation (x
[CV]) of chlorpyrifos residues detected
on passive sampling devices (PSDs) by distance from soybean field edge to grassland
interior. Samples were collected in Minnesota’s farmland region in 2017 and 2018.
Treatment sites consisted of grasslands adjacent to soybean fields that were treated with
insecticides; control sites were grasslands adjacent to unsprayed corn fields. PSD height
refers to samplers placed at mid-canopy height (0.5 m above ground) or ground level (0
m above ground) and these calculations included samples from treatment sites with both
spray methods. Spray method refers to whether the bordering soybean field was sprayed
using an airplane or ground boom and these calculations included samples from both
mid-canopy and ground height PSDs. Mean values are reported in ng/cm2. ................... 41
Table 7. Means and coefficients of variation (x
[CV]) of chlorpyrifos residues detected
on arthropod samples by distance from soybean field edge to grassland interior in
Minnesota’s farmland region in the summers of 2017 and 2018. Treatment sites consisted
of grasslands adjacent to soybean fields that were treated with insecticides either by
airplanes or ground sprayers; control sites were grasslands adjacent to unsprayed corn
fields. Means are reported in mg/kg. Residues were collected from arthropod samples
with unknown masses, and calculations converting parts per billion to mg/kg included the
maximum estimates of target chemical residues per sample. ........................................... 42
Table 8. Minimums, maximums, and coefficients of variation (CV) of the masses of
chlorpyrifos that would be in a birds’ daily food intake if consuming items with arthropod
sample residue amounts estimated in this study, and masses of arthropods needed to reach
the LD50 values for various bird species. Chlorpyrifos residues were collected from
arthropod samples with unknown masses and calculations included the maximum
estimates of chlorpyrifos residues per sample. Chlorpyrifos residues (mg/kg) were
multiplied by the masses of food that various bird species typically consume in one day
(kg; Solomon et al. 2001) to calculate masses of chlorpyrifos in daily food intake. LD50
values for birds (mg/bird; Solomon et al. 2001) were divided by the chlorpyrifos residues
on arthropod samples (mg/kg) to calculate masses of arthropods needed to reach LD50
values. Treatment sites consisted of grasslands adjacent to soybean fields that were
treated with insecticides either by airplanes or ground sprayers; control sites were
grasslands adjacent to unsprayed corn fields. Arthropod samples were collected during
the summers of 2017 and 2018 in Minnesota's farmland region. ..................................... 43
Table 9. Minimums, maximums, and coefficients of variation (CV) of the masses of
chlorpyrifos that would be in a birds’ daily food intake if consuming items with arthropod
sample residue amounts estimated in this study, and masses of arthropods needed to reach
the LD50 values for various bird species. Chlorpyrifos residues were collected from
arthropod samples with unknown masses, and calculations included the minimum
ix
estimates of chlorpyrifos residues per sample. Chlorpyrifos residues (mg/kg) were
multiplied by the masses of food that various bird species typically consume in one day
(kg; Solomon et al. 2001) to calculate masses of chlorpyrifos in daily food intake. LD50
values for birds (mg/bird; Solomon et al. 2001) were divided by the chlorpyrifos residues
on arthropod samples (mg/kg) to calculate masses of arthropods needed to reach LD50
values. Treatment sites consisted of grasslands adjacent to soybean fields that were
treated with insecticides either by airplanes or ground sprayers; control sites were
grasslands adjacent to unsprayed corn fields. Arthropod samples were collected during
the summers of 2017 and 2018 in Minnesota's farmland region. ..................................... 44
Table 10. Number of parameters (K), Akaike's Information Criterion corrected for
sample size (AICc; n = 206), conditional R2 value (R2; variation explained by the entire
model including random effects), deviance (d), and model weight (ω) for models of target
chemical deposition (ng/cm2) onto passive sampling devices (PSDs) at treatment study
sites in the farmland region of Minnesota during the summers of 2017 and 2018. PSDs
were used to assess direct exposure of wildlife to drift from target insecticides (i.e.,
chlorpyrifos, lambda-cyhalothrin, and bifenthrin) sprayed to control soybean aphids. A
hierarchical model selection approach was used in which the first set of models assessed
weather conditions during the spraying event: whether the study site was downwind of
the sprayed field (WDIR), ambient air temperature (TEMP), and wind speed (WSP). The
best-supported weather model was then used as a base model to assess vegetation
covariates in step 2: percentage of the canopy consisting of live vegetation (CCLIVE),
maximum height of live vegetation (MHL), and the vertical density (visual obstruction
reading) from the direction of the sprayed field (VOR). The best-supported weather +
vegetation model was then used in step 3 to assess primary factors of interest: distance of
the PSD from the grassland/soybean edge (DIST), whether the PSD was placed at mid-
canopy or ground level (HT), and whether insecticides were applied via airplane or
ground sprayer (SPRAY). The column ∆AICc compares models within each step of
model development. Models were linear mixed models, included site as a random effect,
and were fitted using the maximum likelihood method.................................................... 45
Table 11. Number of parameters (K), Akaike's Information Criterion corrected for
sample size (AICc; n = 45), conditional R2 value (R2; variation explained by the entire
model including random effects), deviance (d), and model weight (ω) for models of target
chemical deposition (ng/g) on arthropod samples collected from treatment study sites in
the farmland region of Minnesota during the summers of 2017 and 2018. Arthropods
were used to assess potential for indirect exposure of wildlife to drift from target
insecticides (i.e., chlorpyrifos, lambda-cyhalothrin, and bifenthrin) sprayed to control
soybean aphids. A hierarchical model selection approach was used in which the first set
of models assessed weather conditions during the spraying event: whether the study site
was downwind of the sprayed field (WDIR), ambient air temperature (TEMP), and wind
speed (WSP). The best-supported weather model was then used as a base model in step 2
to assess vegetation covariates: percentage of the canopy consisting of live vegetation
(CCLIVE), maximum height of live vegetation (MHL), and vertical density (visual
obstruction reading) from the direction of the sprayed field (VOR). The best-supported
x
weather + vegetation model was then used in step 3 to assess primary factors of interest:
distance from the grassland/soybean edge (DIST) and whether insecticides were applied
via airplane or ground sprayer (SPRAY). The column ∆AICc compares models within
each step of model development. Models were linear mixed models, included site as a
random effect, and were fitted using the maximum likelihood method. Insecticide
residues were collected from arthropod samples with unknown masses, and calculations
converting parts per billion to ng/g included the maximum estimates of target chemical
residues per sample. .......................................................................................................... 47
Chapter 2
Table 1. Vegetation covariates and their hypothesized relationships to arthropod
population metrics. These covariates were used in models assessing effects of soybean
aphid insecticide spraying on arthropod abundance, consumable biomass, and richness in
grasslands in Minnesota’s farmland region. ..................................................................... 90
Table 2. Locations and sampling dates of study sites during the summers of 2017 and
2018 in Minnesota's farmland region. Arthropod samples were collected to assess
impacts of insecticide spraying on arthropod abundance, biomass, and richness in
grasslands. Regions of Minnesota sampled in this study include the southwest (SW), west
central (WC), and central (C) regions. Treatment sites were grasslands adjacent to
soybean fields that were sprayed with insecticides to control aphids; control sites were
grasslands adjacent to corn fields that were not sprayed for aphids. ................................ 91
Table 3. Welch’s two-sample t-test results comparing total abundance of arthropods
collected on secondary sampling transects and reference transects at treatment and
control sites during the summers of 2017 and 2018 in the farmland region of Minnesota.
Treatment sites consisted of grasslands adjacent to soybean fields that were sprayed with
soybean aphid insecticides; control sites were grasslands adjacent to unsprayed corn
fields. Secondary transects were located parallel to field edges and extended from the
field edges to grassland interiors; reference transects were located > 60 m from other
transects to minimize sampling disturbance. Bold values indicate significant differences
(p < 0.05). .......................................................................................................................... 92
Table 4. Welch’s two-sample t-tests results comparing total dry consumable biomass
(mg) of arthropods collected on secondary sampling transects and reference transects at
treatment and control sites during the summers of 2017 and 2018 in the farmland region
of Minnesota. Treatment sites consisted of grasslands adjacent to soybean fields that were
sprayed with soybean aphid insecticides; control sites were grasslands adjacent to
unsprayed corn fields. Secondary transects were located parallel to field edges and
extended from the field edges to grassland interiors; reference transects were located > 60
m from other transects to minimize sampling disturbance. Bold values indicate significant
differences (p < 0.05). ....................................................................................................... 93
Table 5. Coefficient estimates and 95% confidence intervals for covariates included in
model step 2 used to assess arthropod total abundance, total consumable dry biomass,
xi
bird prey (i.e., insects in orders Araneae, Coleoptera, and Orthoptera and Lepidoptera
larvae) abundance, bird prey consumable biomass, Araneae family richness, Coleoptera
family richness, Hemiptera family richness, and Orthoptera family richness. Bold
estimates indicate significance based on 95% confidence intervals not overlapping zero.
The covariate coefficient estimates of percent canopy cover, percent forb cover, and year
are listed only if these coefficients were included in best-supported models using data
collected prior to spraying at both treatment and control sites (model step 1). Covariates
that were included in best-supported models in step 1 were included in models using data
collected at treatment and control sites at 0 and 25 m from the field edge during pre-
spray, 3–5 days post-spray, and 19–21 days post-spray sampling periods (model step 2).
Arthropod samples were collected in grasslands in the farmland region of Minnesota
during the summers of 2017 and 2018. ............................................................................. 94
xii
List of Figures
Chapter 1
Figure 1. Locations of treatment (purple symbols) and control sites (green symbols) in
the farmland region of Minnesota during 2017 (square symbols) and 2018 (circle
symbols) field sampling efforts. Treatment sites were grasslands adjacent to soybean
fields sprayed for aphids; control sites were grasslands adjacent to corn fields that were
not sprayed with insecticides to control soybean aphids. Regions shown include: SW =
southwest, SC = south central, WC = west central, and C = central. ............................... 49
Figure 2. Field sampling design used to assess the exposure of grassland wildlife to
soybean aphid insecticides in the farmland region of Minnesota during the summers of
2017 and 2018. Sampling was conducted on grasslands adjacent to privately owned
soybean fields sprayed for aphid infestations. Black lines indicate primary sampling
transects established perpendicular to the field edge (orange line) and extending into the
grassland. Sampling stations (white circles) were placed 0, 5, 25, 50, 100, and 200 m
from the field edge. An additional station at 400 m was added if the size of the grassland
allowed. ............................................................................................................................. 50
Figure 3. Target chemical (i.e., chlorpyrifos, cyhalothrin, and bifenthrin) deposition on
passive sampling devices (PSDs; n = 368) by distance from field edge to grassland
interior at treatment sites and control sites. Sampling was conducted during the summers
of 2017 and 2018 in Minnesota’s farmland region. Control sites were grasslands adjacent
to corn fields that were not treated with insecticides during sampling; treatment sites
consisted of grasslands adjacent to soybean fields that were treated with insecticides
either by airplanes or ground sprayers. Negative values on the y-axis resulted when
calculating the logarithm of values between 0 and 1. ....................................................... 51
Figure 4. Chlorpyrifos deposition on passive sampling devices (PSDs; n = 368) by
distance from field edge to grassland interior at treatment sites and control sites. White
bars represent PSDs deployed at mid-canopy height (0.5 m above ground); gray bars
represent PSDs deployed at ground level (0 m above ground). The horizontal dashed line
represents the contact LD50 for honey bees (Apis mellifera; 66.67 ng/cm2, see Table 1).
Sampling was conducted during the summers of 2017 and 2018 in Minnesota’s farmland
region. Treatment sites consisted of grasslands adjacent to soybean fields that were
treated with insecticides either by airplanes or ground sprayers; control sites were
grasslands adjacent to unsprayed corn fields. Negative values on the y-axis resulted when
calculating the logarithm of values between 0 and 1. ....................................................... 52
Figure 5. Percentages of applied active ingredients captured as drift on passive sampling
devices (PSDs; n = 206) at treatment sites. Sampling was conducted during the summers
of 2017 and 2018 in Minnesota’s farmland region. Treatment sites consisted of
grasslands adjacent to soybean fields that were treated with insecticides either by
airplanes or ground sprayers. Codes in upper right corners of plots correspond to site IDs
xiii
(see Table 4). At treatment site tA, the landowner reported using thiamethoxam but this
insecticide was not detected on PSDs. Note that y-axes differ among plots. ................... 53
Figure 6. Target chemical (i.e., chlorpyrifos, cyhalothrin, and bifenthrin) deposition on
arthropod samples (n = 81) by distance from field edge to grassland interior at treatment
sites and control sites. Sampling was conducted during the summers of 2017 and 2018 in
Minnesota’s farmland region. Control sites were grasslands adjacent to corn fields that
were not treated with insecticides during sampling; treatment sites consisted of grasslands
adjacent to soybean fields that were treated with insecticides either by airplanes or
ground sprayers. Residues were collected from arthropod samples with unknown masses,
and calculations converting parts per billion to ng/g included the maximum estimates of
target chemical residues per sample. ................................................................................ 54
Figure 7. Chlorpyrifos deposition on arthropod samples (n = 81) by distance from field
edge to grassland interior at treatment sites and control sites. White bars represent
samples collected at control sites (grasslands adjacent to unsprayed corn fields); gray bars
represent samples collected at treatment sites (grasslands adjacent to soybean fields that
were treated with insecticides). The horizontal lines represent the acute oral LD50 for
house sparrows (Passer domesticus), acute oral LD50 for ring-necked pheasants
(Phasianus colchicus), acute oral LD50 for common grackles (Quiscalus quiscula), and
acute oral dose causing orientation impairment in white-crowned sparrows (Zonotrichia
leucophrys; Eng et al. 2017). Acute oral LD50 values are reported in Solomon et al.
(2001). Sampling was conducted during the summers of 2017 and 2018 in Minnesota’s
farmland region. Chlorpyrifos residues were collected from arthropod samples with
unknown masses, and calculations converting parts per billion to mg/kg included the
maximum estimates of chlorpyrifos residues per sample. Negative values on the y-axis
resulted when calculating the logarithm of values between 0 and 1................................. 55
Chapter 2
Figure 1. Locations of treatment (purple symbols) and control sites (green symbols)
during summer 2017 (square symbols) and 2018 (circle symbols) field sampling efforts in
the farmland region of Minnesota. Treatment sites were grasslands adjacent to soybean
fields sprayed for aphids; control sites were grasslands adjacent to corn fields that were
not sprayed with insecticides to control soybean aphids. Regions shown include: SW =
southwest, SC = south central, WC = west central, and C = central. ............................... 96
Figure 2. Field sampling design used to assess the effects of soybean aphid insecticides
on arthropods in the farmland region of Minnesota. Sampling was conducted in
grasslands adjacent to row crop fields during the summers of 2017 and 2018. Treatment
sites were bordered by soybean fields sprayed with insecticides to combat soybean
aphids; control sites were bordered by corn fields that were not treated with insecticides
during sampling efforts. Secondary arthropod collection transects were 20 m long and
were located 0, 25, and 100 m from the field edge. Reference transects were 20 m long
and were located > 60 m from other transects to minimize sampling disturbance.
xiv
Observers collected arthropods during 3 periods: before spraying (blue lines), 3–5 days
post-spraying (green lines) and 19–21 days post-spraying (yellow lines). ....................... 97
Figure 3. Mean arthropod abundances through time at treatment and control sites. Red
circles indicate samples from control site secondary arthropod collection transects and
blue triangles indicate samples from treatment site secondary arthropod collection
transects. Black circles indicate samples from reference transects at control sites and
black triangles indicate samples from reference transects at treatment sites. Error bars
represent standard errors. Treatment sites were grasslands bordered by soybean fields
sprayed with insecticides to combat soybean aphids; control sites were grasslands
bordered by corn fields that were not treated with insecticides. Secondary arthropod
collection transects were parallel to the field edge and were visited by researchers
repeatedly; reference transects were located > 60 m from other transects to minimize
sampling disturbance. Arthropod samples were collected during the summers of 2017 and
2018 in Minnesota’s farmland region. .............................................................................. 98
Figure 4. Mean arthropod consumable dry biomasses (mg) through time at treatment and
control sites. Red circles indicate samples from control site secondary arthropod
collection transects and blue triangles indicate samples from treatment site secondary
arthropod collection transects. Black circles indicate samples from reference transects at
control sites and black triangles indicate samples from reference transects at treatment
sites. Error bars represent standard errors. Treatment sites were grasslands bordered by
soybean fields sprayed with insecticides to combat soybean aphids; control sites were
grasslands bordered by corn fields that were not treated with insecticides. Secondary
arthropod collection transects were parallel to the field edge and were visited by
researchers repeatedly; reference transects were located > 60 m from other transects to
minimize sampling disturbance. Arthropod samples were collected during the summers
of 2017 and 2018 in Minnesota’s farmland region. .......................................................... 99
xv
List of Appendices
Chapter 1
Appendix A. Survey sent to landowners with fields immediately adjacent to potential
study sites in March and April 2017 to assess soybean aphid spraying practices and to
solicit cooperation for summer 2017 sampling efforts. .................................................... 56
Appendix B. Minimums, medians, maximums, means, and coefficients of variation (CV)
of all chemical residues on passive sampling devices deployed during the summers of
2017 and 2018 in Minnesota’s farmland region. Control sites were grasslands adjacent to
unsprayed corn fields; treatment sites consisted of grasslands adjacent to soybean fields
that were treated with insecticides either by airplanes or ground sprayers. Minimums,
medians, maximums, and means are reported in ng/cm2. ................................................. 58
Appendix C. Welch’s two-sample t-test results comparing chlorpyrifos residues (ng/cm2)
on passive sampling devices deployed at mid-canopy height (0.5 m above ground) and
ground level (0 m above ground). Samples were collected during the summers of 2017
and 2018 in Minnesota’s farmland region. Treatment sites consisted of grasslands
adjacent to soybean fields that were treated with insecticides either by airplanes or
ground sprayers; control sites were grasslands adjacent to unsprayed corn fields. .......... 60
Appendix D. Minimums, medians, maximums, means, and coefficients of variation (CV)
of all chemical residues on arthropod samples collected during the summers of 2017 and
2018 in Minnesota’s farmland region. Control sites were grasslands adjacent to
unsprayed corn fields; treatment sites consisted of grasslands adjacent to soybean fields
that were treated with insecticides either by airplanes or ground sprayers. Residues were
collected from arthropod samples with unknown masses, and calculations converting
parts per billion to ng/g included the maximum estimates of chemical residues per
sample. Minimums, medians, maximums, and means are reported in ng/g. .................... 61
Appendix E. Welch’s two-sample t-test results comparing chlorpyrifos residues (ng/g) on
arthropod samples at treatment sites and control sites. Samples were collected during the
summers of 2017 and 2018 in Minnesota’s farmland region. Treatment sites consisted of
grasslands adjacent to soybean fields that were treated with insecticides either by
airplanes or ground sprayers; control sites were grasslands adjacent to unsprayed corn
fields. Chlorpyrifos residues were collected from arthropod samples with unknown
masses, and calculations converting parts per billion to ng/g included the maximum
estimates of chlorpyrifos residues per sample. Bold values indicate significant differences
(p < 0.05). .......................................................................................................................... 63
Appendix F. Minimums, medians, maximums, means, and coefficients of variation (CV)
of target chemical (i.e., chlorpyrifos, cyhalothrin, and bifenthrin) residues on passive
sampling devices (PSDs) deployed during the summers of 2017 and 2018 in Minnesota’s
farmland region. Control sites were grasslands adjacent to unsprayed corn fields;
xvi
treatment sites consisted of grasslands adjacent to soybean fields that were treated with
insecticides either by airplanes or ground sprayers. PSDs were deployed at mid-canopy
height (0.5 m above ground) and ground level (0 m above ground). Minimums, medians,
maximums, and means are reported in ng/cm2. ................................................................ 64
Chapter 2
Appendix A. Percentages of total arthropod abundance and dry consumable biomass by
taxon for all samples collected at treatment and control study sites during the summers of
2017 and 2018 in the farmland region of Minnesota. Treatment sites were grasslands
bordered by soybean fields sprayed with insecticides to combat soybean aphids; control
sites were grasslands bordered by corn fields that were not treated with insecticides. .. 100
Appendix B. Means (
x
) and standard deviations (SD) for abundance of arthropods
collected via sweep-net and vacuum sampling during 3 sampling periods at treatment and
control study sites on secondary arthropod collection transects. Arthropod samples were
collected in grasslands in the farmland region of Minnesota during the summers of 2017
and 2018. Treatment sites were grasslands bordered by soybean fields sprayed with foliar
insecticides to control soybean aphids; control sites were bordered by corn fields that
were not treated with insecticides during sampling. Pre-spray samples were collected 1–3
days prior to the spraying event. Arthropods in orders important in the diets of grassland
nesting birds are listed under “Bird prey groups”; arthropods less important in their diets
are listed under “Other groups.” “Other” includes arthropods in any order not listed in
addition to arthropods not identified to order. ................................................................ 102
Appendix C. Means (
x
) and standard deviations (SD) for abundance of arthropods
collected via sweep-net and vacuum sampling during 2 sampling periods at treatment and
control study sites on additional reference transects. Arthropod samples were collected in
grasslands in Minnesota’s farmland region during the summers of 2017 and 2018.
Treatment sites were grasslands bordered by soybean fields sprayed with foliar
insecticides to control soybean aphids; control sites were bordered by corn fields that
were not treated with insecticides during sampling. Secondary arthropod collection
transects were parallel to the field edge and were visited by researchers repeatedly;
reference transects were located > 60 m from other transects to minimize sampling
disturbance. Pre-spray samples were collected 1–3 days prior to the spraying event.
Arthropods in orders important in the diets of grassland nesting birds are listed under
“Bird prey groups”; arthropods less important in their diets are listed under “Other
groups.” “Other” includes arthropods in any order not listed in addition to arthropods not
identified to order. ........................................................................................................... 106
Appendix D. Means (
x
) and standard deviations (SD) for consumable dry biomass
estimates (mg) of arthropods collected via sweep-net and vacuum sampling during 3
sampling periods at treatment and control study sites on secondary transects. Arthropod
samples were collected in grasslands in Minnesota’s farmland region during the summers
of 2017 and 2018. Treatment sites were grasslands bordered by soybean fields sprayed
xvii
with foliar insecticides to control soybean aphids; control sites were bordered by corn
fields that were not treated with insecticides during sampling. Secondary arthropod
collection transects were parallel to the field edge and were visited by researchers
repeatedly; reference transects were located > 60 m from other transects to minimize
sampling disturbance. Pre-spray samples were collected 1–3 days prior to the spraying
event. Arthropods in orders important in the diets of grassland nesting birds are listed
under “Bird prey groups”; arthropods less important in their diets are listed under “Other
groups.” “Other” includes arthropods in any order not listed in addition to arthropods not
identified to order. ........................................................................................................... 107
Appendix E. Means (
x
) and standard deviations (SD) for consumable dry biomass
estimates (mg) of arthropods collected via sweep-net and vacuum sampling during 2
sampling periods at treatment and control study sites on additional reference transects.
Arthropod samples were collected in grasslands in Minnesota’s farmland region during
the summers of 2017 and 2018. Treatment sites were grasslands bordered by soybean
fields sprayed with foliar insecticides to control soybean aphids; control sites were
bordered by corn fields that were not treated with insecticides during sampling.
Secondary arthropod collection transects were parallel to the field edge and were visited
by researchers repeatedly; reference transects were located > 60 m from other transects to
minimize sampling disturbance. Pre-spray samples were collected 1–3 days prior to the
spraying event. Arthropods in orders important in the diets of grassland nesting birds are
listed under “Bird prey groups”; arthropods less important in their diets are listed under
“Other groups.” “Other” includes arthropods in any order not listed in addition to
arthropods not identified to order. .................................................................................. 110
1
Chapter 1
Grassland wildlife exposure to insecticide spray drift on public lands in Minnesota’s
farmland landscape
OVERVIEW
Soybean aphid (Aphis glycines) insecticides are widely used in the farmland
region of Minnesota to combat insect pests. In Minnesota, the most commonly used
broad spectrum foliar insecticides have been shown to be toxic to wildlife in laboratory
settings. This is of concern to wildlife conservation because increasing evidence suggests
that insecticide exposure is a significant threat to grassland birds and pollinators.
However, little information exists regarding drift and deposition of insecticides in
grasslands in the farmland region of Minnesota. To address this information gap, I
measured insecticide drift and deposition onto passive samplers and arthropods in
grasslands adjacent to soybean fields. I collected samples immediately following
insecticide application at treatment sites and at control sites without insecticide
application. I detected insecticides in grasslands up to 400 m from field edges regardless
of whether adjacent fields were sprayed with insecticides, and deposition was greatest
within 25 m of field edges. The insecticide chlorpyrifos is especially toxic to wildlife, and
I measured residues that were above the contact LD50 for honey bees (Apis mellifera) up
to 25 m from field edges in grasslands. The masses of chlorpyrifos that birds could
consume in a day (if food items contained chlorpyrifos residues equivalent to those in my
arthropod samples) were below the acute oral lethal doses (LD50 values) for common
grackles (Quiscalus quiscula), house sparrows (Passer domesticus), northern bobwhites
(Colinus virginianus), red-winged blackbirds (Agelaius phoeniceus), and ring-necked
2
pheasants (Phasianus colchicus). I used linear mixed models in a hierarchical selection
approach to assess the importance of distance from field edge, spray method (plane or
ground sprayer), and sampler height (mid-canopy or ground) in explaining insecticide
deposition in grasslands. The best-supported model of deposition on passive sampling
devices included an inverse association of distance from the field edge with deposition (β
= -0.62, 95% CI = -1.30 – 0.06) and positive association of samplers being placed at the
mid-canopy level (β = 146.81, 95% CI = -28.99 – 322.60) compared to ground level.
Canopy cover of live vegetation had an inverse association with deposition (β = -6.02,
95% CI = -12.06 – 0.12). The best-supported model of insecticide deposition on
arthropods included effects of air temperature (β = -544.19, 95% CI = -937.41 – -150.98)
and maximum height of vegetation (β = 272.97, 95% CI = 2.10 – 543.84). Grasslands
with cover ≥25 m from row crop edges may provide wildlife habitat with lower exposure
to insecticides. Management regimes that increase the percent canopy cover in grasslands
also have the potential to reduce exposure of grassland wildlife to foliar insecticides.
Key Words: insecticides, chlorpyrifos, chemical drift, farmland landscape, row crops,
grasslands, grassland birds, non-target arthropods, direct exposure, indirect exposure
3
INTRODUCTION
Insecticides are widely used on soybeans throughout the farmland region of
Minnesota to control insect pests, but little is known about the environmentally relevant
impact of these chemicals on non-target grassland wildlife. Soybean aphids (Aphis
glycines) are common agricultural pests that were first discovered in Minnesota in 2000
and quickly spread throughout the state’s farmland region by 2001 (Venette and Ragsdale
2004). Given that large, untreated populations of this pest can decrease crop yields by
40% (Ragsdale et al. 2011), many producers apply broad spectrum foliar insecticides on
their soybeans to control aphid outbreaks. Between 2015-2018, approximately 41% of the
area planted to soybeans in Minnesota was treated with insecticides (U.S. Department of
Agriculture [USDA] National Agricultural Statistics Service [NASS] 2016, 2018, 2019).
With over 3 million ha of soybeans planted in Minnesota annually (USDA 2019a) and
much of the state’s grassland cover located in the farmland region (Minnesota Prairie
Plan Working Group 2018), grassland wildlife have the potential to be exposed to
soybean aphid insecticides in Minnesota.
Chlorpyrifos, lambda-cyhalothrin, and bifenthrin are the 3 most common active
ingredients applied on soybeans to control soybean aphids in Minnesota (USDA NASS
2016). Foliar applications of these chemicals using airplanes or ground sprayers are
common when aphids reach threshold levels. Because these ingredients have varying
withholding times and modes of action, landowners may use a single active ingredient
during a growing season, apply a product that combines insecticides with different modes
of action, or apply rotations of insecticides (Koch et al. 2016). Although these chemicals
can be very effective against soybean aphids when used individually, using multiple
4
active ingredients on a field during a single growing season is becoming increasingly
common to combat aphids’ resistance to lambda-cyhalothrin and bifenthrin (Koch and
Potter 2019).
Growing evidence suggests that insecticide exposure is a threat to grassland
wildlife, especially birds (Avery et al. 2004, Mineau and Whiteside 2006, 2013).
Chlorpyrifos, lambda-cyhalothrin, and bifenthrin have all been shown to be toxic to non-
target organisms in laboratory settings (National Pesticide Information Center 2001,
Christensen et al. 2009, Johnson et al. 2010). These chemicals disrupt nervous systems of
organisms and can cause mortality at high doses. At sublethal dosages, insecticides have
been shown to have negative effects on the behavior and physiology of animals.
Symptoms of sublethal exposure include increased susceptibility to predation, lost
breeding opportunities, impaired development of offspring, loss of mobility and
orientation, and impaired feeding and breeding behavior (Mitra et al. 2011, Moore et al.
2014, Eng et al. 2017).
Chlorpyrifos is a broad-spectrum organophosphate insecticide that disrupts
nervous system functioning of target and non-target organisms through direct contact,
ingestion, and inhalation (Christensen et al. 2009). This chemical is very highly toxic to
bird species including common grackles (Quiscalus quiscula), house sparrows (Passer
domesticus), northern bobwhites (Colinus virginianus), red-winged blackbirds (Agelaius
phoeniceus), and ring-necked pheasants (Phasianus colchicus) with acute oral lethal
doses (LD50 values) of 8.5, 29.5, 32, 13.2, and 12.2 mg/kg, respectively (Solomon et al.
2001, Table 1). Chlorpyrifos is also highly toxic to honey bees (Apis mellifera) with a
contact LD50 of 100 ppb or 66.67 ng/cm2 (Tomlin 2000, Ostiguy et al. 2019, see Table 1
5
for further explanation). LD50 values of chlorpyrifos in contaminated pollen and nectar of
adult honey bees are also representative of non-Apis bee species (e.g., bumblebees; Cutler
et al. 2014, U.S. Environmental Protection Agency [EPA] et al. 2014). At sublethal
levels, chlorpyrifos has been shown to impair orientation in white-crowned sparrows
(Zonotrichia leucophrys) at doses equivalent to 10% of the oral LD50 for house sparrows
(2.95 mg/kg; Solomon et al. 2001, Eng et al. 2017).
Lambda-cyhalothrin is a broad-spectrum pyrethroid insecticide that also disrupts
the nervous systems of target and non-target organisms (National Pesticide Information
Center [NPIC] 2001). This chemical is low in toxicity to birds but highly toxic to
pollinators including bees (World Health Organization [WHO] 1990, NPIC 2001).
Lambda-cyhalothrin is reported to have an oral LD50 of >3,950 mg/kg for mallards (Anas
platyrhynchos), and cyhalothrin’s oral LD50 for domestic hens (Gallus domesticus) is
>10,000 mg/kg (WHO 1990). This chemical is very highly toxic to honey bees with an
oral LD50 of 0.97 µg/bee and contact LD50 of 0.051 µg/bee (WHO 1990, NPIC 2001).
Bifenthrin is a broad-spectrum pyrethroid insecticide that affects the central and
peripheral nervous systems of organisms by direct contact or ingestion (Johnson et al.
2010). This chemical is low in toxicity to birds including northern bobwhites and
mallards with oral LD50 values of 1,800 mg/kg and 2,150 mg/kg, respectively (Tomlin
2000). However, terrestrial insects are especially susceptible to this chemical, with
bifenthrin being highly toxic to bumblebees. One study showed that direct exposure to
bifenthrin killed 100% of worker bumblebees (Besard et al. 2010). Reported toxicity
values for bumblebees include an oral LD50 value of 0.1 µg/bee and contact LD50 of
0.01462 µg/bee (Tomlin 2000).
6
One important avenue for grassland birds and other wildlife to be exposed to
insecticides is through drift associated with routine airplane or ground-based applications
to prevent and control soybean aphid outbreaks. Drift occurs when insecticides are
sprayed on crops but environmental factors (e.g., temperature, wind speed, or wind
direction) result in their transport to areas beyond the targeted application area. The
distances over which drift occurs vary widely, with reported distances ranging from 1 m
to 2,000 m (Davis and Williams 1990, Langhof et al. 2005, Carlsen et al. 2006,
Antuniassi et al. 2014, Holterman et al. 2017, Runquist et al. 2018, Baio et al. 2019).
There is little information about drift and environmentally relevant exposure of wildlife
in grasslands for standard foliar insecticide application regimes in Midwestern farmland
landscapes. This information is necessary to effectively design grasslands that are
protected and managed for wildlife.
Knowledge of wildlife’s exposure to insecticides will help managers with
grassland conservation efforts in Minnesota. Several conservation plans aim to add
grassland cover to Minnesota’s landscape in the coming years. For example, Minnesota’s
Pheasant Action Plan (Minnesota Department of Natural Resources [MNDNR] 2020) and
Prairie Conservation Plan (Minnesota Prairie Plan Working Group 2018) both aim to
offset grassland cover losses due to declining Conservation Reserve Program (CRP)
enrollments by establishing grassland/wetland habitat complexes within the farmland
region of the state. Additionally, a 2016 Minnesota law that requires perennial vegetation
buffers averaging 15-m wide along public lakes, rivers, and streams and 5-m wide along
public ditches has resulted in the addition of narrow strips of grassland cover to farmland
landscapes in recent years. Understanding the potential for grassland wildlife to be
7
exposed to insecticide drift will provide managers and landowners with better
information for managing grasslands and buffers in the farmland region of the state.
My objective was to quantify drift of soybean aphid insecticides into grasslands in
the farmland region of Minnesota. To address this objective, I measured the deposition of
insecticides onto filter paper to assess the potential for grassland wildlife to be directly
exposed to these chemicals through drift. I also quantified insecticide deposition on
arthropods to assess the potential for grassland birds and other insectivores to be exposed
to these chemicals indirectly through consumption of contaminated arthropod prey. I
hypothesized that the distance from the grassland/soybean field edge, spraying method
used, and sample height would influence the amount of insecticide residues measured in
grasslands. In particular, I predicted that samples collected closer to the field edge would
collect more residues, airplane spraying would result in greater measures of drift in
grasslands than ground spraying, and filter paper samples placed above the ground would
contain more residues than samples collected at ground level. I also hypothesized that
chlorpyrifos residues would exceed the contact LD50 for honey bees and acute oral LD50
values for birds in grasslands near the edges of sprayed fields.
METHODS
Study Area
I conducted this study in the southwest (SW), west-central (WC), and central (C)
regions of Minnesota (Fig. 1). Corn and soybeans accounted for approximately 90%,
67%, and 71% of the landscape in these 3 regions, respectively (USDA 2019a, b).
Grassland cover on public and private land accounted for 6.9%, 10.0%, and 5.6% of the
landscape in these regions (Messinger and Davros 2018). Since 2003, these areas have
8
also experienced some of the greatest estimated uses of chlorpyrifos and lambda-
cyhalothrin in Minnesota (Minnesota Department of Agriculture [MDA] 2005, 2012,
2014, 2016).
My study sites consisted of public Wildlife Management Areas (WMAs)
comprised of reconstructed grasslands or grassland/wetland complexes. These sites were
managed by the Minnesota Department of Natural Resources (MNDNR) with the intent
of providing high quality habitat for wildlife. I selected study sites in ArcGIS (version
10.6.1, ESRI 2021) by first choosing WMAs that were bordered by row crop fields and
were of sufficient size. I focused on potential treatment sites that were predicted to be
downwind (i.e., east or north) from cooperators’ soybean fields based on archived
National Weather Service data (TWC Product and Technology LLC 2015).
I visited potential study site WMAs to examine the vegetation diversity and to
identify the crops planted in adjacent fields. I chose sites dominated by a diverse mesic
tallgrass prairie mix containing warm-season grasses and forbs because this assemblage
is commonly used by MNDNR managers and agency partners to restore habitat for
grassland birds and pollinators. In my study sites, predominant grass species included big
bluestem (Andropogon gerardii), smooth brome (Bromus inermis), Canada wild rye
(Elymus Canadensis), and Kentucky bluegrass (Poa pratensis). Dominant forb species
were wild bergamot (Monarda fistulosa), smooth oxeye (Heliopsis helianthoides), and
Canada goldenrod (Solidago canadensis). Canada thistle (Cirsium arvense) was also
commonly present but was not planted by the MNDNR.
9
Landowner Contact
Landowner cooperation was vital to timing my field sampling efforts. To request
the cooperation of landowners and learn about their insecticide spraying practices, I
mailed surveys to 206 landowners who owned land bordering 29 potential study sites in
March and April 2017 (Appendix A). Although the mailed surveys helped me gather
useful information about common spraying habits, I ultimately found that soliciting
landowner cooperation through in-person visits and phone calls was more effective.
Therefore, I did not mail surveys to landowners in 2018. Once I secured landowner
cooperation, I kept in contact with them throughout the growing season to determine if
and when they would be applying insecticides on their soybean fields. Several
landowners rented their land and/or hired farming cooperatives to spray their fields, so I
contacted combinations of landowners, renters, agronomists, cooperative representatives,
and pilots to determine the exact time of spraying and to obtain additional relevant data
after spraying (e.g., insecticide product used, application rate, and tank pressure).
Experimental Design
I conducted sampling in July-August 2017 and 2018, the peak period in which
insecticides were used to control soybean aphids. Each of my treatment study sites
consisted of upland grassland cover directly adjacent to a soybean field. The soybean
field adjacent to each treatment study site was treated with foliar insecticides used to
control soybean aphids. Insecticide applicators treated fields using ground sprayers or
airplanes. I worked closely with cooperating landowners to learn the exact dates of
spraying and to verify the chemical formulations applied to soybean fields bordering my
sites. My control study sites had similar site characteristics except that they were adjacent
10
to corn fields. I was not in contact with the landowners of these corn fields and did not
observe foliar pesticide spraying on these fields during sampling. Standard management
practices for corn in this region do not include foliar insecticide applications in late
summer.
Within each study site, I established 3 primary transects 90-100 m apart that
extended perpendicular from the soybean field edge to the grassland interior (Fig. 2). I
conducted sampling at stations placed at 6 distances (0, 5, 25, 50, 100, and 200 m) along
each of these 3 primary transects. If the site was large enough, I also established a station
at 400 m along each transect. I created transects with the same orientation at control sites.
Therefore, I established 18-21 drift sampling stations at each study site. At treatment
sites, I aligned primary transects perpendicular to the cooperator’s soybean field edge. At
control sites, I established primary transects perpendicular to a grassland edge that was
east or north of a corn field.
Measuring Insecticide Deposition with Passive Sampling Devices
To assess the potential for birds and other wildlife to be directly exposed to
soybean aphid insecticide drift, I deployed passive sampling devices (PSDs) to collect
drift residues. PSDs consisted of WhatmanTM Qualitative Filter Paper (grade 2; GE
Healthcare U.K. Ltd., Little Chalfont, U.K.) attached to 1.27-cm hardware cloth formed
to a cylinder shape. This structure approximated the size and shape of a large songbird or
a gamebird chick. The surface area of the filter paper on each PSD was 354.75 cm2: the
top contained a 7-cm-diamter circle and the vertical plane was covered by a 11.5 by 28.5
cm sheet of filter paper with 1 cm of overlap for attachment.
11
I deployed PSDs in treatment study sites ≤4.5 hours before insecticide
application. I placed PSDs at ground level (0 m) and mid-canopy height (0.5 m) at each
sampling station for a total of 36–42 PSDs per site. Ground-level sampling measured
potential insecticide drift exposure for ground-nesting birds and other ground-dwelling
wildlife (e.g., gamebirds, spiders, beetles, ants, and small mammals). Mid-canopy
sampling measured potential exposure of above-ground nesting birds and canopy-
dwelling species of spiders and insects to insecticide drift. I retrieved the PSDs from the
field ≤2.25 hours after insecticide spraying in the adjacent soybean field ended. At
control sites, I also placed PSDs at ground and mid-canopy levels at each sampling
station. I allowed the PSDs to be exposed to air for a similar amount of time as PSDs at
treatment sites. Upon PSD collection, I wrapped the pieces of filter paper in aluminum
foil, enclosed them in airtight plastic bags, and placed them in a cooler with dry ice in the
field. This prevented chemical degradation by sunlight and heat. I then stored these bags
in a -80 °C freezer until I shipped them to the laboratory for analysis.
During 2018 only, I deployed PSDs 1–3 days prior to spraying at mid-canopy and
ground height at each 0-m sampling station. I conducted this sampling at both treatment
and control sites. PSDs were exposed to the air for 1–3 hours. These samples served as a
secondary field-based control to determine whether insecticides were present prior to
known spraying events at treatment sites.
Measuring Insecticide Deposition on Arthropods
To assess the potential for birds and other insectivorous wildlife to be exposed to
insecticides indirectly via consumption of contaminated arthropod prey, I collected
arthropod samples ≤4 hours after insecticide application in adjacent soybean fields. In
12
each study site, I established secondary arthropod sampling transects beginning at the
sampling stations at 0, 5, and 25 m from the soybean field edge along each of the 3
primary transects. The secondary arthropod sampling transects ran 30 m to the right
(when facing the adjacent crop field) of each PSD sampling station and parallel to the
field edge. Two observers simultaneously collected arthropod samples along each
secondary transect: 1 observer used a sweep net while another used a vacuum sampler
(Southwood and Henderson 2000). Observers walked each secondary transect down and
back for a total of 60 m per collection method per sample. Observers walked unique paths
1.25 m apart to minimize disturbance from sampling and to maximize the likelihood that
the arthropod communities being sampled were similar (Doxon et al. 2011).
I collected sweep-net samples using a standard 38-cm diameter canvas net that
was swung 60 times per sample, and the same observer collected all sweep-net samples
in this study. A second observer collected vacuum samples using a modified hand-held
vacuum with a 15-cm long nozzle held 15 cm above the ground (BioQuip Products Inc.,
Rancho Dominguez, CA, U.S.A.). Sweep-net sampling collected canopy-dwelling
arthropods whereas vacuum sampling collected ground-dwelling arthropods. I combined
sweep-net and vacuum samples from each transect into 1 sample for a total of 9 samples
per study site. I immediately placed arthropod samples in airtight plastic bags and froze
them on dry ice to prevent chemical degradation by sunlight and heat. I then stored them
in a -80 °C freezer until later analysis. I collected arthropod samples at control sites using
the same transect layout and methods, with the timing of collection based on when I
deployed PSDs.
13
Arthropod samples contained varying amounts of vegetation (e.g., grass seed
heads and/or small pieces of stem). This plant material was not separated from arthropods
before sending to the laboratory because post-sampling processing could have caused
contamination or UV degradation of insecticides. I did not weigh samples before
shipment and estimated the maximum mass of any arthropod + vegetation sample to be
10.6 g. This value is the maximum mass of 7 samples that I collected and weighed while
evaluating arthropod sampling techniques (i.e., sweep netting and vacuum sampling)
along 60-m transects at a WMA not used as one of my study sites.
Laboratory Analysis
I sent filter paper and arthropod samples to the U.S. Department of Agriculture -
Agricultural Marketing Service National Science Laboratory (USDA-AMS NSL;
Gastonia, NC, U.S.A.). Laboratory staff analyzed the samples using a solvent-based
extraction method and tested for several insecticides and fungicides. They concentrated
extracts by evaporation and then analyzed them using gas chromatography/mass
spectrometry-negative chemical ionization (GC/MS-NCI) or another appropriate method
(MET-104). The laboratory reported chemical residues of chlorpyrifos, cyhalothrin
(total), and bifenthrin in parts per billion (ppb), which I converted to ng/cm2 for PSDs
and ng/g or mg/kg for arthropod samples (see Residue Conversions in the Data Analysis
section). The laboratory’s limit of detection (LOD) was 2 ppb for chlorpyrifos, 1 ppb for
cyhalothrin, and 4 ppb for bifenthrin. As an additional control, I sent 5 filter paper
samples to the USDA-AMS NSL for chemical residue analysis. I did not deploy these
samples in the field, but I had attached them to hardware cloth frames and stored them in
an airtight bin in the back of a field vehicle prior to shipment to the lab.
14
Weather Measurements
I used Kestrel 5500AG agricultural weather meters (Nielsen-Kellerman Co.,
Boothwyn, PA, U.S.A.) mounted on tripods and equipped with weather vanes to measure
relevant weather data including ambient temperature, relative humidity, wind speed, and
wind direction every 20 seconds during PSD deployment and arthropod sample
collection. I also measured weather data while sampling at control sites. I placed the
weather meters along the center primary transect at 0 m, 100 m, and 200 m from the field
edge.
Vegetation Measurements
I measured ground cover, canopy cover, litter depth, maximum height of live and
dead vegetation, vertical density, and species richness of vegetation in 30 by 55 cm plots
at each PSD station and at the endpoints of arthropod sampling transects. I measured
vegetation 1–3 d prior to the spraying event at treatment sites and 1–3 d prior to PSD
deployment at control sites. Using a modified point-intercept method, I categorized
ground cover into bare ground, litter, or other (e.g., woody debris, rock, or gopher
mound; Bureau of Land Management 1996). I calculated canopy cover from nadir digital
photographs taken of each plot from 1.5 m above the ground with the program
SamplePoint (Booth et al. 2006). Canopy cover categories included grass, forb, dead
vegetation, woody vegetation, and other. I measured litter depth to the nearest 0.1 cm at 1
point within the plot that subjectively represented the average condition of the plot. I
recorded the maximum height of live and dead vegetation within each plot to the nearest
0.5 dm. I measured vertical vegetation density by placing a Robel pole in the center of
each plot and estimating the visual obstruction reading (VOR) from 4 m away and 1 m
15
above the ground from each of the 4 cardinal directions (Robel et al. 1970). Finally, I
counted the number of unique forb and grass species to determine species richness in the
plot.
Researcher Safety
Long-term exposure to organophosphate and pyrethroid insecticides has been
linked to detrimental health effects in humans, particularly chemical applicators. These
chronic health risks include adverse respiratory effects (e.g., asthma and wheezing) and
lung cancer (Lee et al. 2007, Hoppin et al. 2017). Bifenthrin is listed by the EPA as a
possible human carcinogen (Johnson et al. 2010). Exposure to chlorpyrifos, lambda-
cyhalothrin, and bifenthrin can cause short-term side effects including eye, skin, nose,
and throat irritation, headaches, nausea, and dizziness (Dow AgroSciences LLC 2014a,
Syngenta Crop Protection LLC 2014).
To reduce exposure to these insecticides, researchers followed the Personal
Protective Equipment (PPE) recommendations listed on the specimen labels of mixes
containing chlorpyrifos. This chemical is associated with the most severe health risks of
insecticides used on soybeans in my study area. Researchers were equipped with more
PPE than necessary because the PPE recommendations on specimen labels are intended
for applicators who spend several days and many hours per year working with these
chemicals (D. Herzfeld, University of Minnesota, personal communication). Our overall
exposure levels were very low as researchers spent ≤4 hours in grasslands adjacent to
sprayed fields on 1 day per treatment study site. Researchers did not enter the fields
where insecticides were sprayed. We were equipped with Tychem® QC 127 series
hooded coveralls (DuPont, Wilmington, DE, U.S.A.), StanSolv® 15 mil nitrile gloves
16
(MAPA Professional, Colombes, FR), and rubber boots while collecting samples in
treatment sites immediately after insecticide application. Researchers had chemical-
resistant goggles and half-mask air-purifying respirators on-hand in the event that they
experienced eye, skin, nose, or throat irritation while in the field.
Data Analysis
Residue Conversions
To convert the ppb of chemicals reported by the laboratory to ng/cm2 on PSDs or
ng/g on arthropod samples, I used the following equations:
= ng/cm2
or
= ng/g
where:
Ca = concentration of analytical sample (ppb, ng/mL)
Va = volume of analytical sample = 1 mL
Vtx = volume of total extract = 15 mL
Vxs = volume of extract subsample prior to concentration for analysis = 5 mL
Wt = weight of total frozen sample = 3.53 g for PSDs weighing ≥3 g;
or
= known weight (g) of PSD or arthropod samples weighing <3 g;
or
= 3.0 g (minimum) or 10.6 g (maximum) for arthropod samples with
unknown mass ≥3 g
17
Ws = weight of frozen subsample = 3 g for PSD and arthropod samples
weighing ≥3 g;
or
= Wt (g) when Wt < 3 g
Ap = surface area of filter paper from PSD prior to freezer mill processing =
354.75 cm2
Wa = weight of arthropod sample prior to freezer mill processing = 15 g
I did not weigh samples before shipping them to the laboratory. To calculate
minimum residue estimates on arthropod samples with unknown weights, I used 3.0 g for
the total frozen sample weight (Wt), because the laboratory provided me a list of any
sample weighing <3 g. Therefore, 3 g reflected their minimum sample weight threshold.
To calculate maximum residue estimates on arthropod samples with unknown weights, I
used 10.6 g. I obtained this value from the maximum weight of samples collected during
practice arthropod sampling (see Measuring Insecticide Deposition on Arthropods
section). The maximum residue estimates represent “worst-case scenarios” of the
deposition of insecticides on arthropod samples.
I used insecticide residue estimates on arthropods to calculate the mass of
insecticides that birds could consume in a day and the mass of contaminated arthropods
required to reach the LD50 for several bird species. In these calculations, I used values of
bird body mass, mass of food that birds eat in a day, and LD50 values reported in
Solomon et al. (2001). I assumed that my samples were representative of food items birds
could be consuming. Although samples contained arthropods and some vegetation from
18
sweep netting, birds in farmland regions include both invertebrate and plant materials in
their diets (McNicol et al. 1982, Moreby 2004).
Model Covariate Selection
I measured multiple characteristics of weather and vegetation at study sites, but
subsequently reduced the number of covariates used in models assessing insecticide
deposition on PSDs and arthropod samples. To select weather covariates to use in
models, I first considered the aspects of weather that may influence drift: air temperature,
relative humidity, mean wind speed, maximum wind speed, and wind direction. I
considered mean air temperature and mean wind speed in models of insecticide
deposition because I observed no substantive fluctuations in these measures during each
spraying event. I derived values of mean air temperature and wind speed by first
calculating the average of simultaneous measurements at 3 weather meters; every
weather meter was recording measurements at 20 s intervals. Then, I calculated the mean
of values between the start and end time of the spraying event at treatment sites or PSD
deployment at control sites. I deemed a site to be downwind when the average wind
direction was within ±62° of the transect orientation. I then tested whether any of the
remaining weather covariates were highly correlated with any other weather covariate
(i.e., |r| > 0.7; Dormann et al. 2013). Temperature and relative humidity were highly
correlated (r = -0.78) and I considered temperature and not relative humidity in models of
insecticide deposition. Higher air temperatures are associated with higher measures of
drift, as higher temperatures promote evaporation and result in smaller droplets that are
readily transported by the wind (Nuyttens et al. 2017).
19
I reduced the list of all potential vegetation covariates to those that were
associated with distinct aspects of vegetation structure above the ground level: height,
density, and canopy cover. Although I measured the height and canopy cover of both live
and dead vegetation separately, I only used measurements from live vegetation in my
models, as the majority of plants within vegetation plots were alive. I then tested for
correlations between the remaining covariates: maximum height of live vegetation,
vertical obstruction readings, and canopy cover of live plants. These covariates were not
highly correlated (i.e., |r| < 0.7; Dormann et al. 2013).
Models of Insecticide Deposition
I used linear mixed models to assess the potential effects of distance from field
edge, spray method, and PSD height on insecticide deposition on PSDs. To assess
insecticide deposition on arthropod samples, I used linear mixed models with all of the
same covariates except PSD height. The response variables in these models were the
summed residues of chlorpyrifos, lambda-cyhalothrin, and bifenthrin in each sample
(hereafter, “target chemicals”). The units of these residues were ng/cm2 for PSD samples
and ng/g for arthropod samples. Because the arthropod samples that I collected had
unknown weights, my calculations converting lab-reported parts per billion to ng/g
resulted in estimates of target chemical residues per sample (see the Residue Conversions
section). I used maximum residue estimates in models of insecticide deposition on
arthropod samples.
Using data from treatment sites only, I used a hierarchical model selection
approach similar to the methods of Daly et al. (2015). The first set of models
incorporated weather conditions during the spraying event, the second set incorporated
20
vegetation covariates, and the final step incorporated my primary factors of interest (i.e.,
distance, spray method, and height). This approach allowed me to examine how these
primary factors of interest influenced insecticide deposition after accounting for other
environmental factors that I expected to affect drift based on the literature (Table 2). I
constructed models using package nlme (Pinheiro et al. 2021) in program R (version
3.6.0, R Core Team 2021). I fitted models using the maximum likelihood method with
study site as a random effect.
The first set of models included the reduced set of weather covariates: mean air
temperature (TEMP; °C), mean wind speed (WSP; m/s), and a binary covariate for
whether transects were downwind of the sprayed field (WDIR). I retained weather
covariates from the model with the lowest Akaike’s Information Criteria corrected for
sample size (AICc; Burnham and Anderson 2002) and included them in all models in the
next step of analysis.
In the second step of model selection, I added vegetation covariates to the best-
supported model considering weather covariates to account for additional variation in the
data. I included continuous vegetation covariates for maximum height of live vegetation
(MHL; dm), VOR reading from the direction of the sprayed field (VOR; dm), and
percentage of the canopy consisting of live vegetation (CCLIVE; summed percent cover
of grasses and forbs). I used vegetation measurements recorded at each PSD sampling
station for models of insecticide deposition on PSDs. For models of insecticide
deposition on arthropod samples, I used the averaged vegetation measurements from the
start and end of arthropod collection transects. I retained weather and vegetation
covariates from the model with the lowest AICc for inclusion in the final modeling step.
21
I incorporated distance from field edge (DIST; m), spray method (SPRAY; i.e.,
ground or airplane application), and PSD height (HT; i.e., ground or mid-canopy) in the
final step of model selection of insecticide deposition on PSDs. I added DIST, SPRAY,
and HT to the best-supported model of insecticide deposition on PSDs from step 2. I also
included the interaction of spray method × distance as a covariate in these models. I
followed the same approach for models of insecticide deposition on arthropods except
that I did not include HT. The inclusion of distance from field edge, spray method, and/or
PSD height in the best-supported model would suggest that these factors influenced
insecticide deposition, after accounting for other environmental factors (i.e., weather and
vegetation).
RESULTS
Landowner Surveys
Of the 206 surveys I initially mailed to landowners who owned land adjacent to
potential study sites in spring 2017, 24.4% were returned. I sent a second round of 164
letters and had a response rate of 6.1%. Not all landowners filled out the survey
completely because many rented their land and did not have information on aphid
insecticide spraying practices. Approximately 13.6% of landowners completed the survey
in its entirety and 11 landowners indicated that they would be planting soybeans adjacent
to a WMA in 2017. These landowners were willing to be contacted during the growing
season; however, I did not deem these WMAs as feasible treatment sites after visiting
them in-person. More landowners reported spraying their fields by airplanes (n = 12) than
by ground sprayers (n = 8) in the previous 3-5 years. The majority of landowners that
22
sprayed aphid insecticides in past years reported using products with the trade names
Lorsban® (chlorpyrifos; n = 9) and Warrior® (lambda-cyhalothrin; n = 4).
Insecticide Spraying
I collected samples at 5 treatment study sites and 4 control sites between 29 July –
24 August 2017 and 18 July – 18 August 2018, coinciding with peak activity for aphid
insecticide spraying (Table 3). I collected 368 PSD samples and 81 arthropod samples
combined across years. Additionally, I collected 30 pre-spraying PSD samples as
secondary field-based controls in 2018. Soybean fields bordering my treatment study
sites were treated with chlorpyrifos (n = 4) and cyhalothrin (n = 3; Table 4). No
insecticide applicators used bifenthrin. One applicator reported using the insecticide
thiamethoxam in their tank mix, but I did not detect this chemical on PSDs at treatment
sites (Appendix B). Both airplanes (n = 3) and ground sprayers (n = 2) were used to apply
aphid insecticides at treatment study sites. Two of 5 treatment study sites were downwind
at the time of spraying.
Insecticide Deposition on Passive Sampling Devices
I detected target chemicals on PSDs at all distances (0–400 m) at both treatment
and control sites (Table 5, Fig. 3). I detected chlorpyrifos and cyhalothrin at treatment
sites, but did not detect bifenthrin (Fig. 3, Appendix B). The mean value of target
chemical deposition on PSDs at treatment sites was greatest at 0 m from the field edge (x
= 351.44 ng/cm2, Table 5). At every distance from the field edge at treatment sites, mean
target chemical deposition on mid-canopy PSDs was greater than deposition on ground
level PSDs (Table 5). Treatment sites’ target chemical residues at 400 m from the field
edge had a greater mean (x
= 6.96 ng/cm2) than residues I measured at 50, 100, and 200
23
m from the field edge (x
= 3.83, 0.37, and 0.13 ng/cm2, respectively), likely due to the
high mean residue value I measured at 400 m at treatment sites bordered by fields treated
by ground sprayers (x
= 22.43 ng/cm2, Table 5). At treatment sites, means and
coefficients of variation (CVs) of target chemical deposition on PSDs were greater when
bordering fields were sprayed by airplanes compared to ground sprayers at most
distances, with the greatest differences occurring 0–50 m from the field edge (Table 5).
I detected all 3 target chemicals on PSDs at control sites (Fig. 3). Means of target
chemical deposition were generally much lower at control sites compared to treatment
sites; however, mean deposition at 200 m at control sites was 0.06 ng/cm2 greater than at
treatment sites (Table 5). At all distances from the field edge at control sites, target
chemical deposition CVs were similar (CV = 1.85, 0.95, 0.90, 0.97, 1.03, 0.98, 1.02;
Table 5).
Chlorpyrifos had the greatest mean deposition values of all the chemical residues
I measured at both treatment and control sites, with the exception of diethyltoluamide
(DEET; Appendix B). I summarized chlorpyrifos residues separately from target
chemicals to compare residues to levels shown to be toxic to birds and pollinators. Mean
chlorpyrifos residues exceeded 66.67 ng/cm2 (the acute contact LD50 for honey bees,
Table 1) on PSDs at 0, 5, and 25 m from the field edge at treatment sites (x
= 346.99,
141.88, 263.58, respectively; Table 6). Mean chlorpyrifos residues at treatment sites also
exceeded 66.67 ng/cm2 at sites bordered by soybean fields sprayed by airplane at 0–25 m
from the field edge (Table 6). Mean chlorpyrifos residues on PSDs did not exceed 66.67
ng/cm2 at any distance from the field edge at control sites (Table 6, Fig. 4).
24
There were consistent differences in mean chlorpyrifos deposition associated with
PSD height at treatment sites, with greater deposition on mid-canopy PSDs (Table 6).
The greatest differences in residues between mid-canopy and ground-height samplers
occurred 0–25 m from field edges (Table 6). These differences were not statistically
significant based on results of Welch’s two-sample t-tests, however, and did not exhibit a
trend as distance from the field edge increased (Appendix C). PSDs from treatment study
sites contained residue means > 66.67 ng/cm2 at both ground and mid-canopy heights at 0
and 25 m from field edges; mid-canopy PSDs at 5 m from field edges also had a mean
chlorpyrifos residue > 66.67 ng/cm2 (Table 6).
To compare target chemical deposition on PSDs to the concentrations of
insecticides applied in the soybean field at treatment sites, I used landowner-reported
application rates (Table 4) and insecticide label information (Syngenta Canada Inc. 2012;
Cheminova Inc. 2013; Dow AgroSciences LLC 2014b, a; Syngenta Crop Protection LLC
2014) to calculate the active ingredient concentrations of target chemicals applied in
ng/cm2. The landowner at site tC did not provide application rate information, so I used
the label-recommended application rate for calculations pertaining to this site. PSDs
captured very low percentages of active ingredients applied in the field, with the
exception of site tB (Fig. 5). At site tB, 63.9% and 25.5% of the applied concentrations of
chlorpyrifos and cyhalothrin, respectively, were deposited on PSDs at 0 m from the field
edge.
The PSDs that I deployed pre-spraying in 2018 contained very low levels of the
target chemicals in control sites (x
= 0.054 ng/cm2, CV = 0.41) and treatment sites (x
=
0.071 ng/cm2, CV = 0.61). Chlorpyrifos was the predominant insecticide detected in
25
these control samples. The samples that I sent to the lab that were not deployed in the
field but had been attached to hardware cloth frames and stored in the back of a field
vehicle contained detectable levels of bifenthrin (x
= 0.35 ng/cm2, CV = 0.98). Bifenthrin
was the only target chemical detected in these samples.
Insecticide Deposition on Arthropod Samples
I detected target chemicals on arthropod samples at all distances (0–25 m) at both
treatment and control sites (Fig. 6). I detected chlorpyrifos and cyhalothrin at treatment
sites, but did not detect bifenthrin (Fig. 6, Appendix D). I detected all 3 target chemicals
at control sites (Fig. 6, Appendix D).
Similar to PSD samples, I analyzed chlorpyrifos residues separately from other
target chemicals. Chlorpyrifos was the insecticide with the greatest deposition values on
arthropods (Appendix D). Mean chlorpyrifos residues on arthropods collected at 0 and 5
m from the field edge were greater at control sites than treatment sites (Table 7), and
these values were different based on results of Welch’s two-sample t-tests (Appendix E).
Conversely, I found that mean chlorpyrifos residues on samples collected 25 m from the
field edge were greater at treatment sites than control sites (Table 7) but Welch’s two-
sample t-tests did not indicate a statistically significant difference (Appendix E).
I compared chlorpyrifos residues on arthropods to the acute oral LD50 values for
house sparrows, ring-necked pheasants, and common grackles (Table 1). I also compared
residues to the level at which sublethal exposure to chlorpyrifos has been shown to cause
orientation impairment in white-crowned sparrows (Eng et al. 2017). Arthropods
collected in grasslands 0–25 m from field edges at treatment sites contained mean
chlorpyrifos residues lower than acute oral LD50 values for all bird species I considered,
26
and levels were below those causing sublethal effects of orientation impairment in white-
crowned sparrows (Table 7, Fig. 7). However, at some treatment sites at 25 m from field
edges, samples contained residues greater than the acute oral LD50 values for ring-necked
pheasants and common grackles (Fig. 7). These samples also contained residues above
the level associated with orientation impairment in white-crowned sparrows (Fig. 7).
I calculated the mass of chlorpyrifos a bird could consume in a day if every food
item contained the amount of chlorpyrifos residues found in my arthropod samples.
These values resulted from multiplying chlorpyrifos residues on arthropod samples
(mg/kg) by the mass of food birds are reported to eat per day (kg; Solomon et al. 2001).
When comparing the amount of chlorpyrifos that could be contained in birds’ daily food
to LD50 values (mg/bird; Solomon et al. 2001), using maximum residue estimates resulted
in values below oral LD50 values for all bird species I considered: common grackles,
house sparrows, northern bobwhites, red-winged blackbirds, and ring-necked pheasants
(Table 8). When I used minimum residue estimates, the amounts of chlorpyrifos that
could have been in birds’ daily food intakes were further below the LD50 values for every
species (Table 9).
Using chlorpyrifos residue estimates on arthropod samples, I calculated the
masses of contaminated arthropods that would be needed to reach LD50 values for various
bird species. I divided the LD50 values for birds (mg/bird; Solomon et al. 2001) by the
chlorpyrifos residues on arthropod samples (mg/kg). Using maximum residue estimates
on arthropods resulted in arthropod masses much greater than what birds are reported to
consume in a single day (Table 8). These calculated arthropod masses increased when I
used minimum chlorpyrifos residue estimates in calculations (Table 9).
27
Model Selection: Insecticide Deposition on Passive Sampling Devices
The best-supported model of target chemical deposition on PSDs at treatment
sites after accounting for weather and vegetation covariates included distance from the
field edge and PSD height (Table 10). There was an inverse association between distance
from grassland field edge and deposition (β = -0.62, 95% CI = -1.30 – 0.06). Deposition
was greater on PSDs placed at the mid-canopy level than ground level (β = 146.81, 95%
CI = -28.99 – 322.60). Spray method (i.e., ground or airplane) was not included in the
best-supported model. Wind direction during insecticide spraying events, mean air
temperature, and percent canopy cover of live vegetation were the environmental factors
included in the best-supported models of insecticide deposition on PSDs. Sites downwind
of sprayed fields had a positive association with target chemical deposition (β = 123.19,
95% CI = -298.89 – 545.27). Air temperature had an inverse association with deposition
(β = -54.14, 95% CI = -112.21 – 3.94). Canopy cover of live vegetation also had an
inverse association with deposition (β = -6.02, 95% CI = -12.16 – 0.12).
Model Selection: Insecticide Deposition on Arthropod Samples
The best-supported model of target chemical deposition on arthropod samples at
treatment sites after accounting for weather and vegetation covariates did not include
distance from grassland/field edge or spray method (Table 11). Mean air temperature and
the maximum height of live vegetation were included in the best-supported models of
insecticide deposition that considered weather and vegetation variables. Air temperature
had an inverse association to deposition (β = -544.19, 95% CI = -937.41 – -150.98). The
association between maximum height of vegetation and target chemical deposition was
positive (β = 272.97, 95% CI = 2.10 – 543.84).
28
DISCUSSION
In the farmland region of Minnesota, I found detectable levels of target chemicals,
particularly chlorpyrifos, within both treatment and control sites, suggesting that some
background levels of insecticide drift were occurring across this landscape. Amounts of
chemical drift occurring from agricultural fields can vary greatly, with one study finding
foliar insecticide residues on PSDs located up to 2,000 m away from aerially-treated
fields (Baio et al. 2019). Runquist et al. (2018) sampled grassland vegetation near
sprayed fields in Minnesota and detected chlorpyrifos residues in all samples, with the
greatest residue measuring 2,290 ppb along a grassland edge. Insecticide spraying was
not observed in their vicinity when they collected their samples, indicating that
chlorpyrifos residues were present in grasslands even in the absence of spraying on
bordering fields (Runquist et al. 2018). Chlorpyrifos has been shown to persist in the
environment after its initial application: its half-life is 4.2 hours in the atmosphere and 7-
120 days in soils, and residues can remain on plant surfaces up to 14 days post-
application, potentially accumulating through time (Solomon et al. 2001, Christensen et
al. 2009).
The insecticide residues I measured at control study sites likely did not drift from
the corn fields bordering these sites. Chlorpyrifos, cyhalothrin, and bifenthrin are used as
foliar insecticides against corn pests in Minnesota; however, over 84% of the area planted
to corn in Minnesota in 2018 contained seeds genetically modified to protect against
insect pests (Potter et al. 2018). With these modifications, the need for foliar insecticides
on corn has decreased considerably in recent years (L. Stahl, University of Minnesota
Extension, personal communication). Thus, the likely sources of insecticide residues that
29
I measured on samples at control sites and the extremely low residues I detected on pre-
spraying PSDs were sprayed soybean fields in the vicinity of my study sites.
Contaminated arthropods could also have immigrated to my control sites after receiving
sub-lethal doses of insecticides from nearby treated fields (Longley et al. 1997). The
bifenthrin residues I detected on PSDs that were not deployed were likely due to
contamination from insecticide application in the Minnesota farmland landscape in which
my study sites were located. Because these samples were stored in a vehicle used for
fieldwork, exposure to bifenthrin could have occurred as I was driving through areas
where this insecticide was recently applied.
Insecticide deposition on PSDs in my study decreased as distance from the
soybean field edge increased. Several other studies have also documented edge effects of
chemical drift from agricultural fields (Threadgill and Smith 1975, Bui et al. 1998,
Langhof et al. 2005, Carlsen et al. 2006, Nsibande et al. 2015, Holterman et al. 2017,
Baio et al. 2019). However, I found that distance from the field edge was not related to
insecticide deposition on arthropods, likely because I collected arthropods only ≤25 m
from the field edge. Twenty-five meters was likely insufficient for detecting trends in
insecticide deposition over the gradient from field edge to grassland interior.
Alternatively, arthropods could have received lethal doses of insecticides and died before
they could be captured by my sampling efforts.
Mean chlorpyrifos residues on PSDs exceeded the contact LD50 for honey bees 0–
25 m from the field edge at treatment sites on mid-canopy height samplers and at sites
bordered by fields sprayed via airplane (Table 6). Insecticides can also have sublethal
effects that negatively impact arthropod physiology and behavior at levels below the
30
contact LD50 (Desneux et al. 2007). Birds’ dermal exposure to insecticides is not
considered in the EPA’s registration process, and contact LD50 values for birds have not
been established. Thus, I was unable to compare residues on PSDs to levels that may
cause harmful effects to birds resulting from dermal exposure.
The mass of chlorpyrifos a bird could consume in a day (if every food item
contained chlorpyrifos residues equivalent to those found in my arthropod samples) was
below the acute oral LD50 values for several farmland species including common
grackles, house sparrows, northern bobwhites, red-winged blackbirds, and ring-necked
pheasants (Table 8, Table 9). Arthropods in grasslands 0–25 m from sprayed fields
contained chlorpyrifos levels lower than acute oral LD50 values for all bird species I
considered (Table 7, Fig. 7). Solomon et al.'s (2001) risk assessment also indicated that
chlorpyrifos residues on invertebrates in agricultural systems were below oral acute LD50
values for birds. Notably, although I collected chlorpyrifos residues from arthropod
samples with unknown masses and made conversions from chlorpyrifos concentrations
(ppb) using sample mass estimates, even using maximum estimates of chlorpyrifos
residues per sample did not result in estimates of exposure that approached oral LD50
values. My results, therefore, represent best estimates of worst-case scenarios of
chlorpyrifos exposure to birds at my study sites.
The masses of contaminated arthropods in my study that would be needed to
reach oral LD50 values for various bird species were much greater than what birds could
plausibly consume in a single day (Table 8, Table 9). I made these estimates with the
assumption that birds would be consuming arthropods with chlorpyrifos residues on
them, but Bennett, Jr. and Prince (1981) found that ring-necked pheasants avoided food
31
items treated with insecticides. In that study, pheasants detected the presence of
insecticides on treated seeds, avoided treated food items when alternatives were
available, and consumed less food when it was treated with insecticides (Bennett, Jr. and
Prince 1981). Thus, the presence of insecticide residues could cause birds to avoid
contaminated arthropod food items. Future studies considering birds’ potential for
indirect mortality from insecticides would benefit from measuring residues on individual
food items and studying birds’ avoidance of arthropods contaminated with chlorpyrifos.
Acute oral LD50 values have several limitations in assessing potential effects of
insecticide exposure in birds. First, acute oral LD50 values do not reflect chronic exposure
or multiple exposures to insecticides. Second, some mortality can still occur at exposures
below LD50 values. Third, the doses I considered were for adult animals, and in general,
juveniles are more susceptible to negative effects from exposure to chemicals than adults
(Solomon et al. 2001). Finally, negative effects can occur at sublethal doses, although
such effects are difficult to quantify. Acute oral LD50 values are determined by
administering single doses to birds and evaluating mortality at increasing concentrations
of insecticides in laboratory testing (U.S. EPA 2012). Effects of chronic or repeated oral
exposure are less understood, and widespread use of soybean aphid insecticides in
farmland landscapes could contribute to chronic exposure with direct and indirect effects
on birds. Symptoms of exposure to sublethal doses of chlorpyrifos have been documented
in birds in both lab and field studies (e.g., altered brain cholinesterase activity, altered
behaviors, reduced weight gain, and impaired migratory ability; McEwen et al. 1986,
Richards et al. 2000, Al-Badrany and Mohammad 2007, Moye 2008, Eng et al. 2017).
Across breeding biomes, grassland birds have experienced the greatest population losses
32
in recent years with 74% of grassland species reported to be declining (Rosenberg et al.
2019). Chronic exposure to insecticides and detrimental sublethal effects could be
contributing to these declines, and it would be beneficial to consider these potential
population-level effects in future research.
Canopy cover of live vegetation had an inverse association to deposition on PSDs
with higher deposition where vegetation was less dense. Other studies have also found
that canopy cover is related to insecticide deposition resulting from drift (Praat et al.
2000, Donkersley and Nuyttens 2011, Holterman et al. 2017). Higher deposition in areas
with less dense vegetation, in addition to mid-canopy PSDs having greater target
chemical residues than ground-level PSDs, indicated that ground-nesting birds and other
ground-dwelling wildlife may experience less direct exposure to insecticide drift than
organisms in the upper canopy layer of grasslands. Height of vegetation had a positive
association with deposition on arthropod samples. Taller vegetation causes greater air
turbulence intensity that can prevent droplets from settling out of the air (Fogarty et al.
2018), which suggests that higher vegetation height would be associated with lower
insecticide deposition. It is not clear why I did not observe this pattern in my study.
There is some evidence that higher temperatures promote drift by increasing
evaporation rates and diminishing spray droplet size (Nuyttens et al. 2007, Donkersley
and Nuyttens 2011). However, I observed an inverse association between mean air
temperature and insecticide deposition on PSDs and arthropod samples. A previous study
found that downwind spray deposition had a positive association with temperature up to
15° C whereas deposition appeared to decrease at temperatures above 15° C (Holterman
et al. 2017). The mean air temperature at my treatment study sites ranged from 17–28° C
33
during spraying events. In temperatures >15° C, spray clouds can become more buoyant
and spread vertically, causing dilution and decreasing deposition immediately downwind
of the chemical application site (Holterman et al. 2017).
The method of insecticide application (i.e., airplane or ground sprayer) was not
included in the best-supported models of insecticide deposition on PSDs and arthropods.
I suspected that models including the interaction of spray method × distance would best
explain drift because my analyses showed that target chemical residues on PSDs had
more variation (greater CVs) at treatment sites bordering fields sprayed by plane than
those sprayed on the ground. However, this interaction term was not included in the best-
supported model of insecticide deposition on PSDs or arthropods. Several other factors
related to spraying equipment can impact drift, including spray droplet size, nozzle type,
operating pressure, driving speed, boom height, and uncontrolled boom movements
(Threadgill and Smith 1975; Nuyttens et al. 2007, 2017; Arvidsson et al. 2011;
Donkersley and Nuyttens 2011; Nsibande et al. 2015). Furthermore, the application rate
of insecticides and other pesticides has been shown to be an important predictor of drift
(Donkersley and Nuyttens 2011, Nsibande et al. 2015, Baio et al. 2019). Although I
requested information from cooperators regarding the spraying equipment and rate they
used, I did not control for these variables in my study design. These factors likely
influenced the amount of insecticides deposited on PSDs and arthropod samples.
Very low percentages of active ingredients applied in the field were deposited as
drift on my PSDs at treatment sites. This is consistent with Carlsen et al. (2006), who
showed 0.1–9% of applied amounts of herbicides were deposited on passive samplers
near a field edge. However, PSDs contained very high percentages of applied ingredients
34
at site tB. The field bordering this site was sprayed by airplane and the site was
downwind of the sprayed field. During the spraying event, the wind speed averaged 0.93
m/s and the ambient air temperature was 17 °C. The relative humidity was 90.53%, but
high humidity values have been shown to correlate with lower drift rates due to low
evaporation rates (Nuyttens et al. 2007). Thus, the spray method (i.e., via airplane) likely
contributed more to these high drift measures than weather conditions.
Few field studies have documented direct mortality of birds or other grassland
wildlife from insecticide exposure associated with foliar insecticide application in
farmland landscapes. The ecotoxicological risk assessment of chlorpyrifos performed by
Solomon et al. (2001) did not support the contention that the use of chlorpyrifos in
agroecosystems results in extensive mortality of wildlife at the landscape level. However,
my results indicate that insecticide deposition occurs in grasslands and on arthropods that
inhabit these grasslands in a farmland landscape in Minnesota, suggesting that wildlife
are being exposed to these insecticides in grasslands adjacent to row crops treated with
foliar insecticides.
Management Implications
Although increasing the amount of grassland cover in the farmland region of
Minnesota is a priority for natural resource managers, little is known about the exposure
of grassland wildlife to insecticides applied to control insect pests in these landscapes.
My results indicated that chemical deposition in grasslands was greatest ≤ 25 m from
edges of soybean fields sprayed with foliar insecticides, suggesting that interior grassland
cover ≥ 25 m from row crop edges may provide habitat with lower exposure to
insecticides. Additionally, chlorpyrifos levels exceeded contact LD50 values for honey
35
bees up to 25 m from the edges of treated soybean fields. Narrow strips of grassland
cover (e.g., buffers created by the 2016 Minnesota Buffer Law) may be ecological traps
or population sinks for organisms that inhabit them, in part due to potential exposure to
agricultural insecticides. My results also suggest that management regimes that increase
the percent canopy cover of vegetation have the potential to decrease exposure of
grassland wildlife to insecticides. These results are relevant to tallgrass prairie systems
beyond Minnesota to areas that share similar climates, topographies, and vegetation
communities.
36
Table 1. LD50 values for chlorpyrifos for various species of birds (Solomon et al. 2001)
and pollinators. The acute contact LD50 for honey bees (Apis mellifera) is reported as 70
ng/bee in Tomlin (2000) and was converted to ng/cm2 by dividing this value by a honey
bee’s apparent exposure surface area (1.05 cm2; Poquet et al. 2014). The acute LD50 is a
common measure of acute toxicity and represents the lethal dose that causes death in 50%
of animals from a single brief exposure. Exposure to chlorpyrifos in contaminated pollen
and nectar of adult honey bees is also representative of non-Apis bee species (e.g.,
bumblebees; Cutler et al. 2014, U.S. EPA et al. 2014).
Species
Acute
oral LD50
(mg/kg)
Acute oral LD50
(mg/bird)
Acute toxicity
classification
Mass of food
eaten per day
(kg)
Avian species
Common grackle
8.5
0.97
Very highly toxic
0.034
House sparrow
29.5
0.83
Highly toxic
0.008
Northern bobwhite
32
5.70
Highly toxic
0.053
Red-winged blackbird
13.2
0.69
Highly toxic
0.016
Ring-necked pheasant
12.2
13.85
Very highly toxic
0.114
Acute contact
LD50 (ng/cm2)
Acute toxicity
classification
Pollinator species
Honey bee
66.67
Highly toxic
37
Table 2. Weather, vegetation, and primary factors of interest and their hypothesized relationships to chemical deposition. Each set of
covariates constitute steps in the hierarchical modeling process used to assess insecticide deposition on passive sampling devices
(PSDs) and arthropods in Minnesota’s farmland region. Height was included in models of insecticide deposition on PSDs but not in
models of deposition on arthropods.
Acronym
Hypothesized relationship to chemical deposition
Rationale
Weather
TEMP
Ambient air temperature (°C)
Greater deposition associated with greater temperatures
Nuyttens et al. 2007, Arvidsson et al.
2011, Donkersley and Nuyttens 2011
WDIR
Wind direction (whether the study
site was downwind of the sprayed
field or not)
Greater deposition associated with sites that were downwind
of sprayed fields
Holterman et al. 2017
WSP
Wind speed (m/s)
Greater deposition associated with faster wind speeds
Arvidsson et al. 2011, Nsibande et al.
2015, Holterman et al. 2017, Nuyttens et
al. 2017, Baio et al. 2019
Vegetation
CCLIVE
Percentage of the canopy consisting
of live vegetation (%)
Greater deposition associated with lesser percent canopy
cover
Praat et al. 2000, Donkersley and
Nuyttens 2011, Holterman et al. 2017
MHL
Maximum height of live vegetation
(dm)
Greater deposition associated with lower vegetation height
Fogarty et al. 2018
VOR
Visual Obstruction Reading (VOR)
Greater deposition associated with lesser vegetation density
Donkersley and Nuyttens 2011
Primary factors of interest
DIST
Distance from grassland/field edge
(m)
Greater deposition associated with shorter distance from
edge
Threadgill and Smith 1975, Bui et al.
1998, Langhof et al. 2005, Carlsen et al.
2006, Nsibande et al. 2015, Holterman et
al. 2017, Baio et al. 2019
HT
Height of PSD (ground or mid-
canopy)
Greater deposition associated with mid-canopy PSDs
Langhof et al. 2005
SPRAY
Spray method (airplane or ground
sprayer)
Greater deposition associated with airplane sprayers
Nuyttens et al. 2007
38
Table 3. Locations and sampling dates of study sites during the summers of 2017 and
2018 in Minnesota's farmland region. Filter paper and arthropod samples were collected
to assess insecticide deposition from adjacent soybean fields. Regions of Minnesota
sampled in this study include the southwest (SW), west central (WC), and central (C)
regions. Treatment sites were grasslands adjacent to soybean fields that were sprayed
with insecticides to control aphids; control sites were grasslands adjacent to corn fields
that were not sprayed for aphids.
Site ID
Region
County
Site type
Year
Dates when field
sampling
occurred
tA
SW
Jackson
Treatment
2017
29 July
tB
SW
Murray
Treatment
2017
11 August
cA
SW
Jackson
Control
2017
24 August
cB
SW
Lyon
Control
2017
12 August
tC
WC
Lac qui Parle
Treatment
2018
8–10 August
tD
C
Stearns
Treatment
2018
27–28 July
tE
WC
Yellow Medicine
Treatment
2018
6–7 August
cC
C
Kandiyohi
Control
2018
17–18 August
cD
WC
Lac qui Parle
Control
2018
18–21 July
39
Table 4. Spraying methods and other application information for soybean aphid spraying events by cooperating landowners.
Insecticide applications occurred on soybean fields adjacent to treatment sites that were sampled during the summers of 2017 and
2018 in Minnesota's farmland region. Formulated products are commercially available that contain active ingredients and inert
ingredients, and their labels provide recommended application rates and percentages of active ingredients. Sprayer application rate
refers to the rate of tank mix applied to the field; a tank mix could include combinations of insecticides, other pesticides, adjuvants,
and solvents. Some landowners declined to provide some information.
a Values were calculated using landowner-reported formulated product application rates and product label information.
b Value was calculated using an estimated formulated product application rate of 1.75 L/ha based on label recommendations.
Site
ID
Spray
method
Trade name
of
formulated
product
Insecticide
active
ingredients
Formulated
product
application
rate (L/ha)
Active
ingredient
application
rate
(ng/cm2)a
Sprayer
application
rate (L/ha)
Application
speed (m/s)
Boom
height
(m)
Tank
pressure
(kPa)
tA
Ground
Endigo™
lambda-
cyhalothrin +
thiamethoxam
0.26
271.1 +
360.6
140.3
4
0.2–0.3
275.8
tB
Airplane
Bolton™
chlorpyrifos +
gamma-
cyhalothrin
0.88
2627.0 +
87.2
18.7
67.9
1.5
275.8
tC
Ground
Lorsban®-4E
chlorpyrifos
NA
8406.4b
93.5
NA
NA
137.9–
206.8
tD
Airplane
Lorsban®
Advanced
chlorpyrifos
1.17
5261.0
18.7
55.9
2.7–4.0
275.8
tE
Airplane
Lorsban®
Advanced;
Warrior II®
chlorpyrifos;
lambda-
cyhalothrin
0.44; 0.22
1972.9; 546.4
NA
NA
NA
NA
40
Table 5. Means and coefficients of variation (x
[CV]) of target chemicals (i.e., chlorpyrifos, lambda-cyhalothrin, and bifenthrin)
detected on passive sampling devices (PSDs) by distance from soybean field edge to grassland interior. Samples were collected in
Minnesota’s farmland region in 2017 and 2018. Treatment sites consisted of grasslands adjacent to soybean fields that were treated
with insecticides; control sites were grasslands adjacent to unsprayed corn fields. PSD height refers to samplers placed at mid-canopy
height (0.5 m above ground) or ground level (0 m above ground) and these calculations included samples from treatment sites with
both spray methods. Spray method refers to whether the bordering soybean field was sprayed using an airplane or ground boom and
these calculations included samples from both mid-canopy and ground height PSDs. Mean values are reported in ng/cm2.
Distance from soybean field edge
Site type
0 m
5 m
25 m
50 m
100 m
200 m
400 m
Treatment
351.44 (4.11)
143.86 (3.82)
265.81 (3.48)
3.83 (2.35)
0.37 (1.64)
0.13 (1.44)
6.96 (5.02)
PSD height
Mid-canopy
616.68 (3.29)
257.4 (2.99)
383.14 (3.28)
5.66 (2.11)
0.47 (1.56)
0.18 (1.35)
13.8 (3.58)
Ground
86.2 (2.8)
30.31 (2.36)
148.47 (2.68)
2 (2.16)
0.27 (1.68)
0.09 (1.38)
0.11 (0.96)
Spray method
Airplane
569.15 (3.25)
239.57 (2.93)
442.84 (2.65)
6.26 (1.77)
0.47 (1.61)
0.07 (1.31)
0.08 (0.99)
Ground
24.89 (2.22)
0.28 (1.05)
0.25 (1.05)
0.19 (1.1)
0.23 (1.18)
0.22 (1.17)
22.43 (2.81)
Control
0.41 (1.85)
0.2 (0.95)
0.2 (0.9)
0.21 (0.97)
0.22 (1.03)
0.19 (0.98)
0.29 (1.02)
41
Table 6. Means and coefficients of variation (x
[CV]) of chlorpyrifos residues detected on passive sampling devices (PSDs) by
distance from soybean field edge to grassland interior. Samples were collected in Minnesota’s farmland region in 2017 and 2018.
Treatment sites consisted of grasslands adjacent to soybean fields that were treated with insecticides; control sites were grasslands
adjacent to unsprayed corn fields. PSD height refers to samplers placed at mid-canopy height (0.5 m above ground) or ground level (0
m above ground) and these calculations included samples from treatment sites with both spray methods. Spray method refers to
whether the bordering soybean field was sprayed using an airplane or ground boom and these calculations included samples from both
mid-canopy and ground height PSDs. Mean values are reported in ng/cm2.
Distance from soybean field edge
Site type
0 m
5 m
25 m
50 m
100 m
200 m
400 m
Treatment
346.99 (4.15)
141.88 (3.86)
263.58 (3.5)
3.71 (2.36)
0.36 (1.62)
0.13 (1.44)
6.96 (5.02)
PSD height
Mid-canopy
611.42 (3.31)
254.2 (3.02)
380.84 (3.29)
5.48 (2.12)
0.46 (1.53)
0.18 (1.35)
13.8 (3.58)
Ground
82.55 (2.91)
29.56 (2.37)
146.33 (2.7)
1.93 (2.16)
0.26 (1.65)
0.09 (1.38)
0.11 (0.96)
Spray method
Airplane
561.72 (3.29)
236.28 (2.95)
439.14 (2.66)
6.05 (1.78)
0.45 (1.59)
0.07 (1.31)
0.08 (0.99)
Ground
24.88 (2.22)
0.28 (1.05)
0.25 (1.05)
0.19 (1.1)
0.23 (1.18)
0.22 (1.17)
22.43 (2.81)
Control
0.37 (1.93)
0.19 (1.02)
0.19 (1)
0.21 (0.99)
0.21 (1.1)
0.18 (1.05)
0.23 (0.93)
42
Table 7. Means and coefficients of variation (x
[CV]) of chlorpyrifos residues detected
on arthropod samples by distance from soybean field edge to grassland interior in
Minnesota’s farmland region in the summers of 2017 and 2018. Treatment sites consisted
of grasslands adjacent to soybean fields that were treated with insecticides either by
airplanes or ground sprayers; control sites were grasslands adjacent to unsprayed corn
fields. Means are reported in mg/kg. Residues were collected from arthropod samples
with unknown masses, and calculations converting parts per billion to mg/kg included the
maximum estimates of target chemical residues per sample.
Distance from soybean field edge
Site type
0 m
5 m
25 m
Treatment
0.044 (1.14)
0.034 (0.73)
2.15 (2.64)
Control
0.44 (1.11)
0.50 (1.22)
0.55 (1.32)
43
Table 8. Minimums, maximums, and coefficients of variation (CV) of the masses of chlorpyrifos that would be in a birds’ daily food
intake if consuming items with arthropod sample residue amounts estimated in this study, and masses of arthropods needed to reach
the LD50 values for various bird species. Chlorpyrifos residues were collected from arthropod samples with unknown masses and
calculations included the maximum estimates of chlorpyrifos residues per sample. Chlorpyrifos residues (mg/kg) were multiplied by
the masses of food that various bird species typically consume in one day (kg; Solomon et al. 2001) to calculate masses of
chlorpyrifos in daily food intake. LD50 values for birds (mg/bird; Solomon et al. 2001) were divided by the chlorpyrifos residues on
arthropod samples (mg/kg) to calculate masses of arthropods needed to reach LD50 values. Treatment sites consisted of grasslands
adjacent to soybean fields that were treated with insecticides either by airplanes or ground sprayers; control sites were grasslands
adjacent to unsprayed corn fields. Arthropod samples were collected during the summers of 2017 and 2018 in Minnesota's farmland
region.
Species
Mass of food
eaten/day (kg)
LD50
(mg/bird)
Mass of chlorpyrifos in daily
food (mg) [min–max (CV)]
Mass of arthropods to reach
LD50 (kg) [min–max (CV)]
Treatment sites
Common grackle
0.034
0.97
0–0.65 (4.51)
0.05–162 (0.97)
House sparrow
0.008
0.83
0–0.16 (4.51)
0.04–138 (0.97)
Northern bobwhite
0.053
5.70
0–1.01 (4.51)
0.3–949 (0.97)
Red-winged blackbird
0.016
0.69
0–0.29 (4.51)
0.04–114 (0.97)
Ring-necked pheasant
0.114
13.85
0–2.14 (4.51)
0.73–2,308 (0.97)
Control sites
Common grackle
0.034
0.97
0–0.08 (1.2)
0.41–969 (2.21)
House sparrow
0.008
0.83
0–0.02 (1.2)
0.35–826 (2.21)
Northern bobwhite
0.053
5.70
0–0.13 (1.2)
2.4–5,696 (2.21)
Red-winged blackbird
0.016
0.69
0–0.04 (1.2)
0.29–686 (2.21)
Ring-necked pheasant
0.114
13.85
0–0.27 (1.2)
5.83–13,847 (2.21)
44
Table 9. Minimums, maximums, and coefficients of variation (CV) of the masses of chlorpyrifos that would be in a birds’ daily food
intake if consuming items with arthropod sample residue amounts estimated in this study, and masses of arthropods needed to reach
the LD50 values for various bird species. Chlorpyrifos residues were collected from arthropod samples with unknown masses, and
calculations included the minimum estimates of chlorpyrifos residues per sample. Chlorpyrifos residues (mg/kg) were multiplied by
the masses of food that various bird species typically consume in one day (kg; Solomon et al. 2001) to calculate masses of
chlorpyrifos in daily food intake. LD50 values for birds (mg/bird; Solomon et al. 2001) were divided by the chlorpyrifos residues on
arthropod samples (mg/kg) to calculate masses of arthropods needed to reach LD50 values. Treatment sites consisted of grasslands
adjacent to soybean fields that were treated with insecticides either by airplanes or ground sprayers; control sites were grasslands
adjacent to unsprayed corn fields. Arthropod samples were collected during the summers of 2017 and 2018 in Minnesota's farmland
region.
Species
Mass of food
eaten/day (kg)
LD50
(mg/bird)
Mass of chlorpyrifos in daily
food (mg) [min–max (CV)]
Mass of arthropods to reach
LD50 (kg) [min–max (CV)]
Treatment sites
Common grackle
0.034
0.97
0–0.18 (4.43)
0.18–346 (0.86)
House sparrow
0.008
0.83
0–0.04 (4.43)
0.15–295 (0.86)
Northern bobwhite
0.053
5.70
0–0.29 (4.43)
1.07–2,034 (0.86)
Red-winged blackbird
0.016
0.69
0–0.08 (4.43)
0.13–245 (0.86)
Ring-necked pheasant
0.114
13.85
0–0.61 (4.43)
2.59–4,945 (0.86)
Control sites
Common grackle
0.034
0.97
0–0.02 (1.17)
1.44–1,211 (1.78)
House sparrow
0.008
0.83
0–0.01 (1.17)
1.23–1,033 (1.78)
Northern bobwhite
0.053
5.70
0–0.04 (1.17)
8.48–7,120 (1.78)
Red-winged blackbird
0.016
0.69
0–0.01 (1.17)
1.02–858 (1.78)
Ring-necked pheasant
0.114
13.85
0–0.08 (1.17)
20.61–17,309 (1.78)
45
Table 10. Number of parameters (K), Akaike's Information Criterion corrected for
sample size (AICc; n = 206), conditional R2 value (R2; variation explained by the entire
model including random effects), deviance (d), and model weight (ω) for models of target
chemical deposition (ng/cm2) onto passive sampling devices (PSDs) at treatment study
sites in the farmland region of Minnesota during the summers of 2017 and 2018. PSDs
were used to assess direct exposure of wildlife to drift from target insecticides (i.e.,
chlorpyrifos, lambda-cyhalothrin, and bifenthrin) sprayed to control soybean aphids. A
hierarchical model selection approach was used in which the first set of models assessed
weather conditions during the spraying event: whether the study site was downwind of
the sprayed field (WDIR), ambient air temperature (TEMP), and wind speed (WSP). The
best-supported weather model was then used as a base model to assess vegetation
covariates in step 2: percentage of the canopy consisting of live vegetation (CCLIVE),
maximum height of live vegetation (MHL), and the vertical density (visual obstruction
reading) from the direction of the sprayed field (VOR). The best-supported weather +
vegetation model was then used in step 3 to assess primary factors of interest: distance of
the PSD from the grassland/soybean edge (DIST), whether the PSD was placed at mid-
canopy or ground level (HT), and whether insecticides were applied via airplane or
ground sprayer (SPRAY). The column ∆AICc compares models within each step of
model development. Models were linear mixed models, included site as a random effect,
and were fitted using the maximum likelihood method.
Model
K
AICc
∆AICc
R2
d
ω
Step one:
WDIR + TEMP
5
3,266.75
0
0.096
3,256.45
0.44
TEMP
4
3,267.55
0.81
0.083
3,259.35
0.29
WSP + WDIR + TEMP
6
3,268.86
2.12
0.096
3,256.44
0.15
WSP + TEMP
5
3,269.56
2.82
0.083
3,259.26
0.11
WDIR
4
3,274.32
7.58
0.072
3,266.12
0.0099
WSP + WDIR
5
3,276.42
9.68
0.072
3,266.12
0.0034
WSP
4
3,276.73
9.99
0.072
3,268.54
0.0029
Step two:
WEATHERa +
CCLIVE
6
3,264.79
0
0.11
3,252.37
0.30
WEATHER + MHL +
CCLIVE
7
3,265.97
1.18
0.12
3,251.40
0.17
WEATHER + MHL
6
3,266.37
1.58
0.11
3,253.95
0.14
WEATHER
5
3,266.75
1.95
0.096
3,256.45
0.11
46
WEATHER + VOR +
CCLIVE
7
3,266.92
2.13
0.11
3,252.35
0.10
WEATHER + MHL +
VOR + CCLIVE
8
3,267.62
2.83
0.12
3,250.89
0.074
WEATHER + MHL +
VOR
7
3,268.39
3.60
0.11
3,253.82
0.050
WEATHER + VOR
6
3,268.54
3.75
0.097
3,256.12
0.047
Step three:
VEGb + DIST + HT
8
3,263.28
0
0.14
3,246.55
0.21
VEG + DIST
7
3,263.81
0.53
0.13
3,249.24
0.16
VEG + HT
7
3,264.28
1.00
0.13
3,249.71
0.13
VEG
6
3,264.79
1.51
0.11
3,252.37
0.10
VEG + HT + SPRAY *
DIST
10
3,265.07
1.79
0.15
3,243.94
0.087
VEG + DIST + SPRAY
+ HT
9
3,265.17
1.89
0.14
3,246.25
0.083
VEG + SPRAY * DIST
9
3,265.59
2.31
0.14
3,246.67
0.067
VEG + DIST + SPRAY
8
3,265.68
2.40
0.13
3,248.95
0.064
VEG + SPRAY + HT
8
3,266.19
2.91
0.13
3,249.46
0.050
VEG + SPRAY
7
3,266.68
3.40
0.11
3,252.11
0.039
a WEATHER = covariates in the top-ranked Weather model (WDIR + TEMP) from step 1.
b VEG = covariates in the top-ranked Weather and Vegetation model (WDIR + TEMP + CCLIVE) from
step 2.
47
Table 11. Number of parameters (K), Akaike's Information Criterion corrected for
sample size (AICc; n = 45), conditional R2 value (R2; variation explained by the entire
model including random effects), deviance (d), and model weight (ω) for models of target
chemical deposition (ng/g) on arthropod samples collected from treatment study sites in
the farmland region of Minnesota during the summers of 2017 and 2018. Arthropods
were used to assess potential for indirect exposure of wildlife to drift from target
insecticides (i.e., chlorpyrifos, lambda-cyhalothrin, and bifenthrin) sprayed to control
soybean aphids. A hierarchical model selection approach was used in which the first set
of models assessed weather conditions during the spraying event: whether the study site
was downwind of the sprayed field (WDIR), ambient air temperature (TEMP), and wind
speed (WSP). The best-supported weather model was then used as a base model in step 2
to assess vegetation covariates: percentage of the canopy consisting of live vegetation
(CCLIVE), maximum height of live vegetation (MHL), and vertical density (visual
obstruction reading) from the direction of the sprayed field (VOR). The best-supported
weather + vegetation model was then used in step 3 to assess primary factors of interest:
distance from the grassland/soybean edge (DIST) and whether insecticides were applied
via airplane or ground sprayer (SPRAY). The column ∆AICc compares models within
each step of model development. Models were linear mixed models, included site as a
random effect, and were fitted using the maximum likelihood method. Insecticide
residues were collected from arthropod samples with unknown masses, and calculations
converting parts per billion to ng/g included the maximum estimates of target chemical
residues per sample.
Model
K
AICc
∆AICc
R2
d
ω
Step one:
TEMP
4
855.21
0
0.25
846.21
0.41
WDIR + TEMP
5
855.57
0.36
0.29
844.03
0.35
WSP + TEMP
5
857.68
2.47
0.25
846.14
0.12
WSP + WDIR + TEMP
6
858.23
3.02
0.29
844.02
0.092
WDIR
4
861.67
6.46
0.19
852.67
0.016
WSP
4
863.93
8.72
0.19
854.93
0.0053
WSP + WDIR
5
864.21
9.00
0.19
852.67
0.0046
Step two:
WEATHERa + MHL
5
853.77
0
0.31
842.24
0.32
WEATHER + MHL +
VOR
6
854.11
0.34
0.35
839.9
0.27
WEATHER
4
855.21
1.44
0.25
846.21
0.15
48
WEATHER + MHL +
CCLIVE
6
856.21
2.44
0.32
842.00
0.094
WEATHER + MHL +
VOR + CCLIVE
7
856.92
3.15
0.35
839.90
0.066
WEATHER + VOR
5
857.72
3.95
0.25
846.19
0.044
WEATHER + CCLIVE
5
857.74
3.97
0.25
846.20
0.044
WEATHER + VOR +
CCLIVE
6
860.37
6.60
0.25
846.16
0.012
Step three:
VEGb
5
853.77
0
0.31
842.24
0.46
VEG + DIST
6
854.95
1.18
0.34
840.74
0.25
VEG + SPRAY
6
856.10
2.33
0.32
841.89
0.14
VEG + DIST + SPRAY
7
857.29
3.52
0.34
840.27
0.078
VEG + SPRAY * DIST
8
857.46
3.68
0.38
837.46
0.072
a WEATHER = covariates in the top-ranked Weather model (TEMP) from step 1.
b VEG = covariates in the top-ranked Weather and Vegetation model (TEMP + MHL) from step 2.
49
Figure 1. Locations of treatment (purple symbols) and control sites (green symbols) in
the farmland region of Minnesota during 2017 (square symbols) and 2018 (circle
symbols) field sampling efforts. Treatment sites were grasslands adjacent to soybean
fields sprayed for aphids; control sites were grasslands adjacent to corn fields that were
not sprayed with insecticides to control soybean aphids. Regions shown include: SW =
southwest, SC = south central, WC = west central, and C = central.
50
Figure 2. Field sampling design used to assess the exposure of grassland wildlife to
soybean aphid insecticides in the farmland region of Minnesota during the summers of
2017 and 2018. Sampling was conducted on grasslands adjacent to privately owned
soybean fields sprayed for aphid infestations. Black lines indicate primary sampling
transects established perpendicular to the field edge (orange line) and extending into the
grassland. Sampling stations (white circles) were placed 0, 5, 25, 50, 100, and 200 m
from the field edge. An additional station at 400 m was added if the size of the grassland
allowed.
51
Figure 3. Target chemical (i.e., chlorpyrifos, cyhalothrin, and bifenthrin) deposition on
passive sampling devices (PSDs; n = 368) by distance from field edge to grassland
interior at treatment sites and control sites. Sampling was conducted during the summers
of 2017 and 2018 in Minnesota’s farmland region. Control sites were grasslands adjacent
to corn fields that were not treated with insecticides during sampling; treatment sites
consisted of grasslands adjacent to soybean fields that were treated with insecticides
either by airplanes or ground sprayers. Negative values on the y-axis resulted when
calculating the logarithm of values between 0 and 1.
52
Figure 4. Chlorpyrifos deposition on passive sampling devices (PSDs; n = 368) by distance from field edge to grassland interior at
treatment sites and control sites. White bars represent PSDs deployed at mid-canopy height (0.5 m above ground); gray bars represent
PSDs deployed at ground level (0 m above ground). The horizontal dashed line represents the contact LD50 for honey bees (Apis
mellifera; 66.67 ng/cm2, see Table 1). Sampling was conducted during the summers of 2017 and 2018 in Minnesota’s farmland region.
Treatment sites consisted of grasslands adjacent to soybean fields that were treated with insecticides either by airplanes or ground
sprayers; control sites were grasslands adjacent to unsprayed corn fields. Negative values on the y-axis resulted when calculating the
logarithm of values between 0 and 1.
53
Figure 5. Percentages of applied active ingredients captured as drift on passive sampling
devices (PSDs; n = 206) at treatment sites. Sampling was conducted during the summers
of 2017 and 2018 in Minnesota’s farmland region. Treatment sites consisted of
grasslands adjacent to soybean fields that were treated with insecticides either by
airplanes or ground sprayers. Codes in upper right corners of plots correspond to site IDs
(see Table 4). At treatment site tA, the landowner reported using thiamethoxam but this
insecticide was not detected on PSDs. Note that y-axes differ among plots.
54
Figure 6. Target chemical (i.e., chlorpyrifos, cyhalothrin, and bifenthrin) deposition on
arthropod samples (n = 81) by distance from field edge to grassland interior at treatment
sites and control sites. Sampling was conducted during the summers of 2017 and 2018 in
Minnesota’s farmland region. Control sites were grasslands adjacent to corn fields that
were not treated with insecticides during sampling; treatment sites consisted of grasslands
adjacent to soybean fields that were treated with insecticides either by airplanes or
ground sprayers. Residues were collected from arthropod samples with unknown masses,
and calculations converting parts per billion to ng/g included the maximum estimates of
target chemical residues per sample.
55
Figure 7. Chlorpyrifos deposition on arthropod samples (n = 81) by distance from field
edge to grassland interior at treatment sites and control sites. White bars represent
samples collected at control sites (grasslands adjacent to unsprayed corn fields); gray bars
represent samples collected at treatment sites (grasslands adjacent to soybean fields that
were treated with insecticides). The horizontal lines represent the acute oral LD50 for
house sparrows (Passer domesticus), acute oral LD50 for ring-necked pheasants
(Phasianus colchicus), acute oral LD50 for common grackles (Quiscalus quiscula), and
acute oral dose causing orientation impairment in white-crowned sparrows (Zonotrichia
leucophrys; Eng et al. 2017). Acute oral LD50 values are reported in Solomon et al.
(2001). Sampling was conducted during the summers of 2017 and 2018 in Minnesota’s
farmland region. Chlorpyrifos residues were collected from arthropod samples with
unknown masses, and calculations converting parts per billion to mg/kg included the
maximum estimates of chlorpyrifos residues per sample. Negative values on the y-axis
resulted when calculating the logarithm of values between 0 and 1.
56
Appendix A. Survey sent to landowners with fields immediately adjacent to potential
study sites in March and April 2017 to assess soybean aphid spraying practices and to
solicit cooperation for summer 2017 sampling efforts.
57
58
Appendix B. Minimums, medians, maximums, means, and coefficients of variation (CV)
of all chemical residues on passive sampling devices deployed during the summers of
2017 and 2018 in Minnesota’s farmland region. Control sites were grasslands adjacent to
unsprayed corn fields; treatment sites consisted of grasslands adjacent to soybean fields
that were treated with insecticides either by airplanes or ground sprayers. Minimums,
medians, maximums, and means are reported in ng/cm2.
Site type
Chemical
Minimum
Median
Maximum
Mean
CV
Control
Acephate
0
0
0
0
NA
Azoxystrobin
0
0
0.02
0.00018
9.46
Bifenthrin
0
0
0.57
0.01
7.31
Chlorothalonil
0
0
0
0
NA
Chlorpyrifos
0.03
0.1
3.38
0.23
1.48
Clothianidin
0
0
0
0
NA
Cyfluthrin
0
0
0
0
NA
Cyhalothrin
0
0
0.14
0.0087
3.36
Cypermethrin
0
0
0
0
NA
DEET
1.84
26.32
97.62
32.2
0.76
Deltamethrin
0
0
0
0
NA
Dimethoate
0
0
0
0
NA
Dinotefuran
0
0
0
0
NA
Esfenvalerate
0
0
0
0
NA
Fluoxastrobin
0
0
0
0
NA
Fluxapyroxad
0
0
0
0
NA
Imidacloprid
0
0
0
0
NA
Metconazole
0
0
0
0
NA
Methomyl
0
0
0
0
NA
Propiconazole
0
0
0
0
NA
Pyraclostrobin
0
0
0
0
NA
Sulfoxaflor
0
0
0
0
NA
Tebuconazole
0
0
0
0
NA
Tefluthrin
0
0
0
0
NA
Tetraconazole
0
0
0.03
0.0015
3.7
Thiamethoxam
0
0
0
0
NA
Trifloxystrobin
0
0
0
0
NA
Treatment
Acephate
0
0
0
0
NA
Azoxystrobin
0
0
0.025
0.00092
3.76
Bifenthrin
0
0
0
0
NA
Chlorothalonil
0
0
1.14
0.0055
14.35
Chlorpyrifos
0
0.06
7841.13
111.07
6.2
Clothianidin
0
0
0
0
NA
Cyfluthrin
0
0
0
0
NA
Cyhalothrin
0
0
40.8
1.28
4.11
59
Cypermethrin
0
0
0
0
NA
DEET
1.82
32.29
157.22
39.84
0.87
Deltamethrin
0
0
0
0
NA
Dimethoate
0
0
0
0
NA
Dinotefuran
0
0
0
0
NA
Esfenvalerate
0
0
0
0
NA
Fluoxastrobin
0
0
0
0
NA
Fluxapyroxad
0
0
0.15
0.0036
4.43
Imidacloprid
0
0
0
0
NA
Metconazole
0
0
1.52
0.036
4.64
Methomyl
0
0
0
0
NA
Propiconazole
0
0
0.059
0.0012
5.07
Pyraclostrobin
0
0
3.71
0.097
4.35
Sulfoxaflor
0
0
0
0
NA
Tebuconazole
0
0
0
0
NA
Tefluthrin
0
0
0
0
NA
Tetraconazole
0
0
0
0
NA
Thiamethoxam
0
0
0
0
NA
Trifloxystrobin
0
0
0.0085
0.000041
14.35
60
Appendix C. Welch’s two-sample t-test results comparing chlorpyrifos residues (ng/cm2)
on passive sampling devices deployed at mid-canopy height (0.5 m above ground) and
ground level (0 m above ground). Samples were collected during the summers of 2017
and 2018 in Minnesota’s farmland region. Treatment sites consisted of grasslands
adjacent to soybean fields that were treated with insecticides either by airplanes or
ground sprayers; control sites were grasslands adjacent to unsprayed corn fields.
Distance from
field edge (m)
t
df
p-value
Treatment sites
0
1.01
14.39
0.331
5
1.13
14.23
0.277
25
0.69
16.76
0.499
50
1.12
17.56
0.280
100
0.94
23.07
0.358
200
1.27
20.83
0.216
400
1.00
12.00
0.337
Control sites
0
-0.44
15.22
0.669
5
0.68
20.15
0.503
25
0.59
19.93
0.562
50
0.95
20.78
0.353
100
0.23
22.00
0.821
200
0.42
19.83
0.680
400
0.90
12.89
0.383
61
Appendix D. Minimums, medians, maximums, means, and coefficients of variation (CV)
of all chemical residues on arthropod samples collected during the summers of 2017 and
2018 in Minnesota’s farmland region. Control sites were grasslands adjacent to
unsprayed corn fields; treatment sites consisted of grasslands adjacent to soybean fields
that were treated with insecticides either by airplanes or ground sprayers. Residues were
collected from arthropod samples with unknown masses, and calculations converting
parts per billion to ng/g included the maximum estimates of chemical residues per
sample. Minimums, medians, maximums, and means are reported in ng/g.
Site type
Chemical
Minimum
Median
Maximum
Mean
CV
Control
Acephate
0
0
0
0
NA
Azoxystrobin
0.71
19.79
124.37
31.35
1.02
Bifenthrin
0
0
418.35
72.63
1.76
Chlorothalonil
0
0
0
0
NA
Chlorpyrifos
0
254.75
2,374.40
497.07
1.2
Clothianidin
0
0
0
0
NA
Cyfluthrin
0
0
0
0
NA
Cyhalothrin
0
0
75.61
6.11
2.6
Cypermethrin
0
0
0
0
NA
DEET
0
49.2
375.95
88.18
1.17
Deltamethrin
0
0
0
0
NA
Dimethoate
0
0
0
0
NA
Dinotefuran
0
0
0
0
NA
Esfenvalerate
0
0
0
0
NA
Fluoxastrobin
0
0
0
0
NA
Fluxapyroxad
0
0
37.45
2.34
2.84
Imidacloprid
0
0
0
0
NA
Metconazole
0
0
0
0
NA
Methomyl
0
0
0
0
NA
Propiconazole
0
12.72
174.55
27.89
1.33
Pyraclostrobin
0
6.58
47.35
8.81
1.16
Sulfoxaflor
0
0
0
0
NA
Tebuconazole
0
0
11.31
0.31
6
Tefluthrin
0
0
0
0
NA
Tetraconazole
0
0
92.57
10.17
2.1
Thiamethoxam
0
0
0
0
NA
Trifloxystrobin
0
0.8
4.95
1.15
1.19
Treatment
Acephate
0
0
0
0
NA
Azoxystrobin
0
9.19
52.29
13.39
0.92
Bifenthrin
0
0
0
0
NA
Chlorothalonil
0
0
0
0
NA
Chlorpyrifos
0
30.39
18,868
744.29
4.51
Clothianidin
0
0
0
0
NA
Cyfluthrin
0
0
0
0
NA
62
Cyhalothrin
0
0
1,858.53
202.06
2.37
Cypermethrin
0
0
0
0
NA
DEET
0
61.83
496.79
117.49
1.25
Deltamethrin
0
0
0
0
NA
Dimethoate
0
0
1.6
0.16
2.69
Dinotefuran
0
0
0
0
NA
Esfenvalerate
0
0
0
0
NA
Fluoxastrobin
0
0
0
0
NA
Fluxapyroxad
0
0
17.67
0.94
3.28
Imidacloprid
0
0
18.37
1.16
3.32
Metconazole
0
0
243.8
21.59
2.36
Methomyl
0
0
0
0
NA
Propiconazole
0
9.19
99.6
15.22
1.27
Pyraclostrobin
0
9.89
608.44
60.76
1.91
Sulfoxaflor
0
0
0
0
NA
Tebuconazole
0
0
0
0
NA
Tefluthrin
0
0
0
0
NA
Tetraconazole
0
0
3.53
0.079
6.71
Thiamethoxam
0
0
0
0
NA
Trifloxystrobin
0
1.41
4.95
1.48
0.96
63
Appendix E. Welch’s two-sample t-test results comparing chlorpyrifos residues (ng/g) on
arthropod samples at treatment sites and control sites. Samples were collected during the
summers of 2017 and 2018 in Minnesota’s farmland region. Treatment sites consisted of
grasslands adjacent to soybean fields that were treated with insecticides either by
airplanes or ground sprayers; control sites were grasslands adjacent to unsprayed corn
fields. Chlorpyrifos residues were collected from arthropod samples with unknown
masses, and calculations converting parts per billion to ng/g included the maximum
estimates of chlorpyrifos residues per sample. Bold values indicate significant differences
(p < 0.05).
Distance from
field edge (m)
t
df
p-value
0
-2.80
11.19
0.017
5
-2.65
11.03
0.022
25
1.08
14.57
0.297
64
Appendix F. Minimums, medians, maximums, means, and coefficients of variation (CV)
of target chemical (i.e., chlorpyrifos, cyhalothrin, and bifenthrin) residues on passive
sampling devices (PSDs) deployed during the summers of 2017 and 2018 in Minnesota’s
farmland region. Control sites were grasslands adjacent to unsprayed corn fields;
treatment sites consisted of grasslands adjacent to soybean fields that were treated with
insecticides either by airplanes or ground sprayers. PSDs were deployed at mid-canopy
height (0.5 m above ground) and ground level (0 m above ground). Minimums, medians,
maximums, and means are reported in ng/cm2.
Site type
Distance
Minimum
Median
Maximum
Mean
CV
Control
0
0.04
0.1
3.38
0.41
1.85
5
0.04
0.096
0.7
0.2
0.95
25
0.05
0.12
0.66
0.2
0.9
50
0.03
0.14
0.67
0.21
0.97
100
0.04
0.12
0.78
0.22
1.03
200
0.04
0.11
0.67
0.19
0.98
400
0.034
0.12
0.93
0.29
1.02
Treatment
0
0
0.44
7,860.53
351.44
4.11
5
0
0.22
2,907.74
143.86
3.82
25
0
0.18
4,884.19
265.81
3.48
50
0
0.046
38.07
3.83
2.35
100
0
0.05
2.67
0.37
1.64
200
0
0.032
0.7
0.13
1.44
400
0
0.046
178.12
6.96
5.02
PSD height
Mid-canopy
Ground
Mid-canopy
Ground
Mid-canopy
Ground
Mid-canopy
Ground
Mid-canopy
Ground
Mid-canopy
Ground
Mid-canopy
Ground
0
0
0.51
7,860.53
616.68
3.29
0
0.37
936.51
86.2
2.8
5
0
0.47
2,907.74
257.4
2.99
0
0.22
249.02
30.31
2.36
25
0
0.46
4,884.19
383.14
3.28
0
0.06
1,456.19
148.47
2.68
50
0
0.042
38.07
5.66
2.11
0
0.05
14.14
2
2.16
100
0
0.13
2.67
0.47
1.56
0
0.05
1.36
0.27
1.68
200
0
0.05
0.7
0.18
1.35
0
0.03
0.41
0.088
1.38
400
0
0.04
178.12
13.8
3.58
0
0.07
0.36
0.11
0.96
Spray method
Airplane
Ground
0
0.02
0.44
7,860.53
569.15
3.25
0
0.3
146.27
24.89
2.22
Airplane
5
0.05
0.51
2,907.74
239.57
2.93
Ground
0
0.14
0.81
0.28
1.05
Airplane
25
0.03
0.33
4,884.19
442.84
2.65
Ground
0
0.17
0.83
0.25
1.05
Airplane
50
0
0.045
38.07
6.26
1.77
Ground
0
0.11
0.54
0.19
1.1
Airplane
100
0.02
0.05
2.67
0.47
1.61
Ground
0
0.13
0.79
0.23
1.18
Airplane
200
0
0.03
0.34
0.072
1.31
Ground
0
0.1
0.7
0.22
1.17
65
Airplane
400
0
0.045
0.25
0.082
0.99
Ground
0
0.11
178.12
22.43
2.81
66
Chapter 2
Impacts of insecticide spray drift on arthropod prey resources of birds in grasslands
in Minnesota
OVERVIEW
Soybean aphid (Aphis glycines) insecticides are used throughout the farmland
region of Minnesota to combat insect pests. However, these foliar spray insecticides have
the potential to drift beyond target fields into nearby grassland cover where birds and
other insectivores forage. Arthropods serve important roles in grassland ecology and are
susceptible to mortality from exposure to broad-spectrum insecticides. My objective was
to assess impacts of soybean aphid insecticides on arthropods in grasslands, especially
those that are important in grassland bird diets. I measured the abundance, consumable
biomass, and family richness of insects and spiders in grasslands adjacent to soybean
fields that were treated with chlorpyrifos, lambda-cyhalothrin, and bifenthrin—the 3 most
common insecticides used to treat soybean aphids in Minnesota. I compared these
measures to samples collected at control sites adjacent to corn fields not sprayed for
aphids during 3 periods: 1–3 days before spraying, 3–5 days post-spraying, and 19–21
days post-spraying. Short-term reductions in total arthropod abundance, bird prey
abundance, and Coleopteran family richness occurred in grasslands bordered by fields
sprayed with foliar insecticides. The total abundance of arthropods in grasslands
bordering sprayed soybean fields was lower 3–5 days after insecticide applications (β = -
49.06, 95% CI = -89.84 – -8.28). The abundance of arthropods important in grassland
bird diets (specifically, Araneaens, Coleopterans, Orthopterans, and Lepidopteran larvae)
was also lower after nearby spraying, with lower abundance measured in treatments sites
67
19–21 days post-spraying (β = -23.94, 95% CI = -44.99 – -2.88). Coleopteran family
richness at treatment sites was lower than control sites 3–5 days after insecticide
applications (β = -0.94, 95% CI = -1.82 – -0.06). Measures of total consumable dry
biomass, bird prey biomass, family richness of Araneaens, family richness of
Hemipterans, and family richness of Orthopterans were not different between treatment
and control sites post-spraying. My results suggest that reductions in arthropod food
abundance for grassland birds are associated with insecticide spraying up to 21 days after
the spraying event.
Key Words: insecticides, farmland landscape, row crops, grasslands, non-target
arthropods, grassland bird diets
68
INTRODUCTION
Insecticides are used on soybeans throughout the farmland region of Minnesota to
control insect pest populations. Over 3 million ha of soybeans are planted in Minnesota
annually (U.S. Department of Agriculture [USDA] 2019a), and in 2015–2018,
approximately 41% of the area planted to soybeans was treated with insecticides (U.S.
Department of Agriculture National Agricultural Statistics Service 2016, 2018, 2019).
Producers apply broad-spectrum foliar insecticides on their soybeans to control soybean
aphids (Aphis glycines). If left untreated, these pests can decrease crop yields by 40%
(Ragsdale et al. 2011). Grasslands in Minnesota’s farmland regions are highly
fragmented and often share borders with row crop fields (Minnesota Prairie Plan
Working Group 2018). Thus, grassland wildlife, including beneficial insects and spiders,
have the potential to be exposed to soybean aphid insecticides due to chemical drift in
these landscapes (see Chapter 1).
Chlorpyrifos, lambda-cyhalothrin, and bifenthrin are the most commonly used
foliar insecticides applied to soybeans in Minnesota (USDA National Agricultural
Statistics Service [NASS] 2016). When soybean aphids reach threshold levels, these
chemicals are applied in a liquid form using airplanes or ground sprayers. These
insecticides kill insects by disrupting nervous system function and are designed to be
effective by direct contact, ingestion, and inhalation (National Pesticide Information
Center [NPIC] 2001, Christensen et al. 2009, Johnson et al. 2010). The modes of action
of these chemicals are similar for both target and non-target organisms (Christensen et al.
2009, Johnson et al. 2010). Thus, these insecticides can kill beneficial arthropods
including pollinators and predators of soybean aphids (Minnesota Department of
69
Agriculture [MDA] 2018). Sublethal doses of insecticides can also be harmful to
beneficial arthropods. Symptoms of sublethal exposure include increased susceptibility to
predation, depressed immune system capacity, reduced fecundity, impaired development
of offspring, loss of mobility, and impaired feeding and breeding (Desneux et al. 2007).
Although insecticide applications target arthropods in row crops, insecticides can
drift beyond fields into nearby grassland cover. Chemical drift occurs when liquid foliar
pesticides are sprayed on crops and wind or other environmental factors transport them
beyond the application site. Foliar pesticides have been shown to drift beyond target crop
fields under typical application conditions (see Chapter 1; Threadgill and Smith 1975,
Bui et al. 1998, Langhof et al. 2005, Carlsen et al. 2006, Nsibande et al. 2015, Holterman
et al. 2017, Baio et al. 2019). Drift can occur over large distances, even up to 2,000 m
beyond targeted areas (Baio et al. 2019). Many factors can influence drift, including
environmental factors (e.g., wind speed, wind direction, and air temperature) and
spraying equipment (e.g., boom height, tank pressure, and nozzle design). Insecticide
product labels contain best management practices with recommendations for equipment
that reduces chemical drift, but it is the applicator’s responsibility to choose appropriate
methods. Labels also include information on suitable wind, temperature, and humidity
conditions for insecticide application (Dow AgroSciences LLC 2014a).
Grassland areas near row crop fields are especially susceptible to insecticide
deposit via drift and are important to non-target arthropods in farmland regions.
Insecticide product labels contain information on buffer zones where insecticides must
not be applied around water bodies and residential areas, but no such regulations exist
around grasslands or other natural areas (Dow AgroSciences LLC 2014a). Grassland
70
cover in farmland landscapes is important to non-target arthropods because it provides
diverse vegetation composition and structure that is non-existent in monoculture row
crops, serves as a refuge during tilling and harvest, and harbors source populations that
contribute to the recolonization of insecticide-treated fields (Tscharntke and Greiler 1995,
Longley et al. 1997).
Arthropods represent a large proportion of the total biodiversity in tallgrass prairie
ecosystems and maintaining or enhancing their populations is an important conservation
goal (Dietrich et al. 1998, Harper et al. 2000). Minnesota’s Prairie Conservation Plan
aims to establish and enhance grassland cover within the farmland region of the state to
support diverse populations of birds, arthropods, and other wildlife (Minnesota Prairie
Plan Working Group 2018). Beyond their importance to biodiversity, diverse arthropod
communities are crucial for their role in ecosystem function (e.g., pollination and nutrient
cycling; Harper et al. 2000). With insects being the primary pollinators of native plants,
management that maintains a diverse arthropod fauna is crucial to the conservation of
tallgrass prairie ecosystems (Dietrich et al. 1998). Thus, reduced arthropod populations
resulting from insecticide application may pose a threat to grasslands and the wildlife they
support.
Reductions in arthropod abundance and biomass due to foliar insecticide drift
could negatively affect insectivorous grassland birds by reducing their food supply. The
majority of breeding grassland birds’ diets incorporate insects, and insects are the
primary item fed to nestlings (Wiens and Rotenberry 1979, Kaspari and Joern 1993).
Protein-rich arthropods are especially important for breeding grassland birds during egg-
laying, nestling, and fledgling periods. Insect diversity and abundance have been shown
71
to be lower in crop fields exposed to lambda-cyhalothrin (Langhof et al. 2003, Galvan et
al. 2005). There is correlative evidence that reduced insect food supplies are associated
with reduced nesting success for birds in fragmented habitats surrounded by cultivated
fields (Zanette et al. 2000). High pesticide use on landscapes has also been shown to
correlate with insectivorous bird population declines, with reduced insect food supplies cited
as a cause of this relationship (Hallmann et al. 2014).
My objective was to assess the effects of soybean aphid insecticides on arthropod
prey of grassland nesting birds and other insectivorous wildlife. Specifically, I measured
the abundance, consumable biomass, and family richness of insects and spiders in
grasslands adjacent to row crop fields in Minnesota’s farmland region. I hypothesized
that arthropod samples at treatment sites where bordering soybean fields were sprayed
with foliar insecticides would have lower abundance, biomass, and richness compared to
control sites where foliar insecticides were not applied to the adjacent field. I predicted
that these measures of arthropod communities would be lower at the field edge than in
the grassland interior after insecticide spraying events at treatment sites.
METHODS
Study Area
I selected study sites in the southwest, west-central, and central regions of
Minnesota (Fig. 1). Corn and soybeans accounted for approximately 90%, 67%, and 71%
of the landscape in these 3 regions, respectively (USDA 2019a, b). Grasslands covered
6.9%, 10.0%, and 5.6% of the landscape in these regions on public and private land
(Messinger and Davros 2018). These areas have experienced some of the greatest
72
estimated uses of chlorpyrifos and lambda-cyhalothrin in Minnesota since 2003 (MDA
2005, 2012, 2014, 2016).
My study sites consisted of public Wildlife Management Areas (WMAs) managed
by the Minnesota Department of Natural Resources (MNDNR) as habitat for grassland
and wetland wildlife. I first selected study sites using ArcGIS (version 10.6.1, ESRI
2021) and chose WMAs consisting of grasslands or grassland/wetland complexes
bordered by row crops. I chose potential treatment sites that were predicted to be
downwind (east or north) from soybean fields based on typical wind direction patterns
determined from archived National Weather Service data (TWC Product and Technology
LLC 2015). I then visited these WMAs to observe their plant diversity and to identify the
crops planted in adjacent fields. I chose sites dominated by a diverse mesic tallgrass
prairie mix containing warm-season grasses and forbs. This assemblage is increasingly
used by MNDNR managers and agency partners to restore gras