Milkweed loss in agricultural fields because of herbicide
use: effect on the monarch butterfly population
JOHN M. PLEASANTS
1
and KAREN S. OBERHAUSER
2
1
Department of Ecology,
Evolution and Organismal Biology, Iowa State University, Ames, IA, USA and
2
Department of Fisheries, Wildlife and Con-
servation Biology, University of Minnesota, St Paul, MN, USA
Abstract. 1. The size of the Mexican overwintering population of monarch butter-
flies has decreased over the last decade. Approximately half of these butterflies come
from the U.S. Midwest where larvae feed on common milkweed. There has been a
large decline in milkweed in agricultural fields in the Midwest over the last decade.
This loss is coincident with the increased use of glyphosate herbicide in conjunction
with increased planting of genetically modified (GM) glyphosate-tolerant corn
(maize) and soybeans (soya).
2. We investigate whether the decline in the size of the overwintering population
can be attributed to a decline in monarch production owing to a loss of milkweeds in
agricultural fields in the Midwest. We estimate Midwest annual monarch production
using data on the number of monarch eggs per milkweed plant for milkweeds in dif-
ferent habitats, the density of milkweeds in different habitats, and the area occupied
by those habitats on the landscape.
3. We estimate that there has been a 58% decline in milkweeds on the Midwest
landscape and an 81% decline in monarch production in the Midwest from 1999 to
2010. Monarch production in the Midwest each year was positively correlated with
the size of the subsequent overwintering population in Mexico. Taken together, these
results strongly suggest that a loss of agricultural milkweeds is a major contributor
to the decline in the monarch population.
4. The smaller monarch population size that has become the norm will make the
species more vulnerable to other conservation threats.
Key words. Glyphosate, GMO, milkweed, monarch butterfly
Introduction
Monarch butterflies (Danaus plexippus L. Lepidoptera: Danai-
nae) in the Eastern North American migratory population
undergo a multi-generation annual cycle that includes wintering
in central Mexico. In the spring, adults that have overwintered
migrate north and reproduce in Texas and states to the north
and east. Their offspring move farther north into much of the
eastern half of the United States and southern Canada, and two
to three more generations are produced (Cockrell et al., 1993;
Malcolm et al., 1993; Prysby & Oberhauser, 2004). Most adults
that emerge after mid-August are in a state of reproductive dia-
pause (Herman, 1985; Goehring & Oberhauser, 2002) and
migrate from the summer breeding range to their wintering
grounds, where they remain until spring (Solensky, 2004).
Annual counts of the size of the overwintering population in
Mexico indicate that the monarch population has been declining
over the last decade and a half (Rendo
´
n-Salinas et al., 2011;
Brower et al., 2011b). One possible explanation for this decline
is that monarch production has been decreasing as a result of a
reduction in the availability of the larval host plant. Monarch
larvae feed primarily on milkweeds (genus Asclepias- Family
Apocynaceae, subfamily Asclepiodeae). On the basis of milkweed
cardenolide fingerprints, it has been estimated that 92% of the
monarchs wintering in Mexico had fed as larvae on the common
milkweed, Asclepias syriaca (Malcolm et al., 1993). Studies in
Iowa found a large reduction in A. syriaca in corn (maize, Zea
mays) and soybean (soya, Glycine max) fields from 1999 to 2009
(Hartzler & Buhler, 2000; Hartzler, 2010). It is likely that a simi-
lar reduction has occurred throughout the region where corn
and soybeans are predominantly grown. Eighty per cent of both
Correspondence: John M. Pleasants, Department of Ecology,
Evolution and Organismal Biology, Iowa State University, Ames,
IA 50011, USA. E-mail: jpleasan@iastate.edu
Insect Conservation and Diversity (2012) doi: 10.1111/j.1752-4598.2012.00196.x
2012 The Authors
Insect Conservation and Diversity 2012 The Royal Entomological Society 1
corn and soybeans are grown in the Midwest (USDA, National
Agricultural Statistics Service, 2011c), which is composed of the
states of North and South Dakota, Nebraska, Kansas, Mis-
souri, Iowa, Minnesota, Wisconsin, Illinois, Indiana, Michigan,
and Ohio. A study in 2000 (Oberhauser et al., 2001) found that
monarchs heavily used milkweeds in corn and soybean fields.
On the basis of stable isotope analysis, Wassenaar and Hobson
(1998) estimated that half of the monarchs overwintering in
Mexico in 1997 came from the Midwest. Thus, the Midwestern
United States is at the epicentre of a reduction in milkweeds in
agricultural fields and is also an area that has in recent history
contributed a large component of the monarch population. In
this study, we estimate the magnitude of this milkweed loss and
its consequences for monarch production.
Milkweed in agricult ural fields has long been a concern for
farmers as its presence reduces yield (Bhowmik, 1994). In the
1970s and 1980s, milkweed infestation in agricultural fields was
viewed to be on the increase with 10.5 million ha infested in the
north-central states (Martin & Burnside, 1980). Herbicides have
been increasingly used to control weeds in row crops. Many of
these herbicides produce only moderate control of milkweed,
but glyphosate, often referred to as Roundup
(Monsanto, St.
Louis, MO, USA), is more effective (Bhowmik, 1994; Pline
et al., 2000). However, it also has a detrimental effect on crop
plants, so until the development of genetically modified (GM)
glyphosate-tolerant (Roundup Ready
, Monsanto) crop plants,
herbicides other than glyphosate were used to control weeds.
Glyphosate-tolerant soybeans were introduced in 1996 and had
reached a 94% adoption level by 2011, and glyphosate-tolerant
corn was introduced in 1998 and had reached a 72% adoption
level by 2011 (USDA, Economic Research Service, 2011).
Glyphosate use in soybeans went from 1.4 million kg in 1994 to
41.7 million kg in 2006 (the last year for which data are available
and when adoption of glyphosate-tolerant soybeans was 89%)
and glyphosate use in corn went from 1.8 million kg in 2000 to
28.5 million kg in 2010 when the adoption level was 70%
(USDA, National Agricultural Statistics Service, 2011a,b).
The time period (1999–2009) over which the Iowa studies
found a large reduction in A. syriaca in corn and soybean fields
(Hartzler & Buhler, 2000; Hartzler, 2010) is coincident with the
period when use of glyphosate herbicide increased in conjunc-
tion with the increased adoption of glyphosate-tolerant corn
and soybeans. It is very probable that a similar milkweed reduc-
tion has occurred throughout the Midwest because adoption lev-
els of herbicide-tolerant crops are similar throughout this region
(USDA, Economic Research Service, 2011). How much milk-
weed loss does this represent on a landscape scale? To address
this question, we need information on the density of milkweeds
in different habitats and the landscape area covered by those
habitats. Common milkweed tends to be found in habitats with
a moderate degree of disturbance, including roadsides, pastures,
old fields, prairies and agricultural fields (Bhowmik, 1994). Mul-
tiple data sets provide information on the density of milkweeds
in different habitats over the last decade. The studies by Hartzler
and Buhler (2000) and Hartzler (2010) surveyed a number of
milkweed habitats in Iowa, including agricultural fields. Addi-
tionally, a number of Midwest volunteers in the Monarch Larva
Monitoring Project (2011), hereafter referred to as MLMP,
measured milkweed density in their non-agricultural observation
patches over several consecutive years. Milkweed density data
can be combined with published statewide land-use data to esti-
mate the number of milkweeds in different habitats. Some of the
data sets we use come from Iowa because for some parameters
only Iowa data are available. However, we use data from the
Midwest as a whole whenever possible and make the case that
the resulting estimates of monarch production are representative
of the Midwest.
What is the significance of the loss of milkweeds in agricul-
tural fields for monarchs? To address this issue, we need to esti-
mate annual monarch production in the Midwest over the last
decade to determine whether there has been a significant down-
ward trend. Obtaining data to estimate production is difficult,
despite the fact that the monarch butterfly is such a well-studied
species. One approach would be to use the number of migrants
that come out of the Midwest at the end of the summer as a mea-
sure of production. A monarch tagging programme begun
20 years ago (Monarch Watch, 2011) has been tracking migrat-
ing butterflies. The number of monarchs tagged shows a decline
from 2004 to 2010 (Brower et al., 2011a). However, it is difficult
to obtain accurate measures of production from this tagging
programme because of the variability among the years in the
number of person-hours involved in capture and tagging, the fall
conditions when tagging occured and the locations where tag-
ging occured. Alternatively, one could use counts of the number
of migrating monarchs passing particular locations where they
tend to be funnelled because of passage over water or geogra-
phy. Such counts have been made for over a decade in upper
Michigan and New Jersey (Davis, 2011) but these sites do not
monitor monarchs from the Midwest.
Rather than trying to count adults, another approach to
estimating Midwest monarch production is to focus on the
number of eggs and larvae found on milkweed plants. This
requires monitoring many patches of milkweed in different
habitats, including agricultural fields. Production can then be
estimated from the average number of monarchs per plant in
each habitat and the number of milkweeds in each habitat
on the landscape. We have combined several existing data
sets that provide this information. The MLMP (2011), which
has been operational for over a decade, provides data on egg
and larva density on milkweeds. MLMP volunteers are
located throughout the monarch breeding range and monitor
sites of their choosing weekly over the summer months,
reporting the number of plants (stems) monitored and the
number of eggs and larvae observed. They learn the proce-
dures of the project through workshops, by reading directions
on the project website (MLMP, 2011) and via communica-
tion with the project managers (Prysby & Oberhauser, 2004).
The sites they monitor, however, are not in agricultural fields.
But one of us (Pleasants) has monitored eggs and larvae on
milkweeds in both agricultural fields and non-agricultural
habitats for several years in central Iowa and a study with
larger spatial scale quantified monarch density in both agri-
cultural and non-agricultural habitats in 2000 (Oberhauser
et al., 2001). We will make the case that the relative use of
milkweeds in agricultural and non-agricultural habitats
observed over those years can be extrapolated to provide
2 John M. Pleasants and Karen S. Oberhauser
2012 The Authors
Insect Conservation and Diversity 2012 The Royal Entomological Society, Insect Conservation and Diversity
data on monarch use of agricultural milkweeds in years
where only MLMP data exist. There is a question of what
aspect of production to use to estimate monarch population
changes. The latest stage for which we have density data,
and thus which is closest to actual production of adult mon-
archs, is the fifth instar (L5, the last larval instar). However,
there are many factors that can affect survivorship from egg
to L5 that have nothing to do with milkweed availability,
such as predation and weather. Our goal was to examine the
effect of milkweed resource limitation on monarch produc-
tion. Consequently, we chose to focus on eggs per plant that
represents potential production.
Methods
Data sources for milkweed density
Habitats in which milkweeds are found include primarily
roadsides, corn fields, soybean fields, pastures, old fields, and
land set aside from farming and enrolled in the Conservation
Reserve Program (CRP). CRP land is typically planted to a vari-
ety of cover plants including grasses and forbs. To estimate milk-
weed densities in these habitats, we used data from several
sources: Iowa censuses carried out in 1999 and 2009 (Hartzler &
Buhler, 2000; Hartzler, 2010), and data from some MLMP vol-
unteers who measured milkweed density at their sites in several
Midwest states. To calculate monarch production for each year,
it is necessary to know how milkweed densities have changed
over the last decade in non-agricultural and agricultural
habitats.
Non-agricultural habitats. For roadsides, there was little
observed change in milkweed density in Iowa between 1999
and 2009 (Hartzler, 2010) so we have assumed that milk-
weed density did not change in that habitat over the entire
period of the analysis. Hartzler (2010) measured milkweed
densities for CRP land and pastures in 1999 but not
in 2009 so any change that may have occurred could not be
determined from the Iowa data. However, a subset of
MLMP volunteers (n = 16) measured milkweed density at
their sites (which included natural areas, CRP land, pastures
and old fields) for at least 4 years over this period (97
total observations). Measurements by individual MLMP
volunteers did not cover the entire period but there were
sufficiently long and overlapping sequences to provide a
complete picture. Volunteers either measured the area of
their site and did a complete count of milkweed stems, or
used a modified belt transect to sample milkweed density in
100 1 · 1 m plots. We have used those data to estimate the
change in milkweed density in CRP land and pasture land
over the last decade.
For the data from the MLMP volunteers, we used log of
milkweed density as the variate and used an SAS mixed model
and restricted maximum likelihoodestimationwithfixedeffects
being ‘habitat’, ‘year’ and ‘habitat by year’. We did not find a
‘habitat by year’ effect so we reran the analysis with this
removed. There was a significant ‘year’ effect (F
1,85
= 9.35,
P = 0.003). The slope of the regression (on a log scale) was
)0.0536, which corresponds to a decline in density of 5.2% per
year. We found no ‘habitat’ effect so we applied the same rate of
decline to both CRP land and pastures (Fig. 1).
Agricultural habitats. We have values for milkweed density
in Iowa agricultural fields for 1999 and 2009 (Hartzler & Buhler,
2000; Hartzler, 2010). To calculate milkweed density in fields for
the intervening years, we have to make an assumption about the
shape of the decline. Pleasants observed the change in the num-
ber of milkweeds in plots in seven agricultural fields in Iowa
from 2000 to 2008 (Fig. 1). The observed decline is best
described by an exponential decay function. Such a function is
also consistent with more acres of glyphosate-tolerant corn and
soybeans being planted each year over the last decade (USDA,
Economic Research Service, 2011). We have therefore assumed
that milkweed density in fields decreased as an exponential decay
function from its 1999 value to its 2009 value (see Table 1). This
corresponds to a 14.2% decline per year (Fig. 1). Other decline
functions, ranging from a linear decline to a more precipitous
exponential decay, had no significant effect on the overall
results.
Data sources for land use
We obtained data on the acres occupied by roadsides and pas-
tures on the Iowa landscape in 2002 from Lubowski et al. (2006)
and, because no more recent data exist, we have assumed the
Year
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
Percent of starting milkweed density
0
10
20
30
40
50
60
70
80
90
100
Non-ag
Lowa ag Hartzler
Lowa ag Pleasants
Fig. 1. Decline of milkweeds in agricultural and non-agricultural
habitats. The line depicting the decline in non-agricultural habi-
tats is based on a regression using data from MLMP volunteers.
The line depicting the decline in agricultural habitats is based on
an exponential decay function connecting the 1999 and 2009 val-
ues from the Iowa surveys (see Methods). Also shown is the pro-
portional change in the number of milkweed stems in all
monitored plots in seven agricultural fields in Iowa starting with
998 stems in 2000. The increase in milkweed stems observed in
the agricultural sites in 2005 was attributed to the influence of
fields where corn was planted 2 years in a row. Some agricultural
fields received glyphosate herbicide treatment and others non-
glyphosate treatment. No observations were made in 2006.
Herbicide use and monarch butterflies 3
2012 The Authors
Insect Conservation and Diversity 2012 The Royal Entomological Society, Insect Conservation and Diversity
acres in roadside and pasture have not changed substantially
over the last decade. Data on the acres planted to corn or soy-
beans by year were obtained from Iowa State Agricultural Sta-
tistics (2010) and the amount of Iowa CRP land from the
USDA Conservation Programs (2010).
Estimating monarch use of non-agricultural milkweeds
To estimate monarch use of non-agricultural milkweeds,
we used data on the number of monarch eggs per milk-
weed stem from the MLMP. We examined MLMP data
from 1999 to 2010 for sampling localities within the Mid-
west (eastern Kansas, eastern Nebraska, eastern North and
South Dakota, Minnesota, Iowa, Missouri, Wisconsin, Illi-
nois, Michigan, Indiana and western Ohio). Sites were
excluded in any given year if the average number of milk-
weedsmonitoredwas<25andiftherewerefewerthan
five sampling events in July and August. We also excluded
garden sites because they represent a minor component of
milkweeds on the landscape. Sites were excluded if volun-
teers observed more larvae than eggs because these volun-
teers may not have been able to discern monarch eggs
accurately. We initially divided sites into two groups based
on the habitats in which the milkweeds were found: ‘natu-
ral areas’ (prairies or nature preserves) and ‘other’ (pas-
tures, old fields, roadsides and CRP land); there were no
sites in agricultural fields. However, ‘natural areas’ and
‘other’ were not significantly different from each other in
egg density and were combined in the analysis into a single
‘non-agricultural’ category.
For any site, the number of eggs per plant varies over the
course of the season. However, there is a population build-up
during July and August when the second ⁄ third generation
Table 1. Estimates of the amount of milkweed in non-agricultural habitats, agricultural fields and total milkweeds in Iowa from 1999 to
2010.
Year
Milkweeds in non-agricultural habitats
CRP hectares* CRP mlkwds Pasture mlkwdsà Roadside mlkwds§ Total non-ag mlkwds–
1999 601 127.4 19.8 38.2 185.4
2000 647 130.1 18.8 38.2 187.1
2001 729 139.0 17.8 38.2 195.0
2002 755 136.4 16.9 38.2 191.5
2003 762 130.5 16.0 38.2 184.7
2004 767 124.5 15.2 38.2 177.9
2005 776 119.5 14.4 38.2 172.0
2006 793 115.7 13.6 38.2 167.5
2007 797 110.3 12.9 38.2 161.4
2008 733 96.2 12.3 38.2 146.6
2009 690 85.8 11.6 38.2 135.7
2010 663 78.2 11.0 38.2 127.4
Year Milkweeds in agricultural fields
Total milkweeds§§
Total ag hectares** Mlkwd density Total ag mlkwdsàà
1999 9267 23.00 213.2 398.5
2000 9308 19.75 183.8 370.9
2001 9186 16.92 155.4 350.4
2002 9166 14.55 133.4 324.8
2003 9267 12.49 115.8 300.4
2004 9267 10.73 99.4 277.3
2005 9247 9.21 85.2 257.2
2006 9207 7.91 72.8 240.3
2007 9247 6.79 62.8 224.2
2008 9328 5.83 54.4 201.0
2009 9389 5.00 46.9 182.6
2010 9389 4.29 40.3 167.6
*·1000; from USDA Conservation Programs (2010).
m
2
· 1000; CRP ha · 212 m
2
ha
)1
(milkweed density from H&B, 2000) · 0.948
x
(where x = 0 for 1999).
àm
2
· 1000; 1416 ha (Lubowski et al., 2006) · 14 m
2
ha
)1
(milkweed density from H&B, 2000) · 0.948
x
(where x = 0 for 1999).
§m
2
· 1000; 386 ha (Lubowski et al., 2006) · 99 m
2
ha
)1
(average milkweed density from H&B, 2000 and H, 2010).
–m
2
· 1000; Conservation Reserve Program (CRP) milkweeds + Pasture milkweeds + Roadside milkweeds.
** ·1000; from Iowa State Ag. Statistics (2010).
m
2
ha
)1
; 1999 value from H&B (2000), 2009 value from H (2010); others = 1999 value · 0.858
x,
where x = 0 for 1999.
ààm
2
· 1000; Ag ha · Milkweed density.
§§m
2
· 1000.
4 John M. Pleasants and Karen S. Oberhauser
2012 The Authors
Insect Conservation and Diversity 2012 The Royal Entomological Society, Insect Conservation and Diversity
occurs (MLMP, 2011). We used egg density at the peak of this
build-up as a metric of annual production. For each year, our
estimate of production was based on the average maximum egg
density over all MLMP sites. This metric does not include all of
the annual production but does allow us to examine the relative
differences in production among years.
Monarch use of milkweeds in agricultural fields
Pleasants monitored milkweed populations and monarch
activity in agricultural fields and non-agricultural habitats in
Iowa from 2000 to 2003. Initially six study sites were selected.
Each site included a field planted to soybeans, another field adja-
cent or nearby that was planted to corn and a nearby non-agri-
cultural habitat. Non-agricultural habitats included natural
areas, pastures, old fields and roadsides. CRP land was not
explicitly included as a habitat type but the non-agric ultural hab-
itats selected are similar in vegetative characteristics to CRP
land. Sites were all located within a 10 km radius of Ames, Iowa,
except for one site located 40 km south of Ames. Over the years
of study, a few sites were removed from monitoring for logistical
reasons and a few others added but in all years, both agricultural
and non-agricultural plots were examined. Within each site,
patches of milkweeds were marked (milkweed plots). These
patches were relatively discrete units that ranged in area from
3 · 3to6· 10 m and contained 10–150 milkweed stems. In
each field, approximately 10 milkweed plots were chosen and
mapped using a global positioning system device so they could
be relocated in subsequent years. Sites were visited at weekly
intervals: in 2000 from late May to late August; in 2001 from
early July through late August; and in 2002 and 2003 from early
June to late August. During each visit, every milkweed stem in
each milkweed plot was inspected for monarch eggs and larvae.
As described above, we used the maximum number of eggs
per stem observed during the weekly censuses from July through
August as the measure of production. Egg densities in different
non-agricultural habitat types were not statistically different, so
they were combined into one category. Egg densities on milk-
weed in corn and soybean fields in any year were not statistically
differentfromeachotherandwerecombinedintoasinglecate-
gory. The results are shown in Table 2. Egg densities on milk-
weeds in agricultural fields were significantly higher than on
milkweeds in non-agricultural habitats in each year by an aver-
age factor of 3.89.
Estimating potential monarch production
Potential monarch production for any year is equal to the
sum of egg production from two sources: non-agricultural and
agricultural milkweeds. To calculate production from non-agri-
cultural milkweeds, we first determined the number of milk-
weeds in non-agricultural habitats. This is equal to the area
occupied by each habitat type (CRP land, pasture and roadside)
multiplied by the density of milkweeds in that habitat (see
Table 1).We then multiplied the total number of non-agricul-
tural milkweeds by the average number of eggs per non-agricul-
tural milkweed plant for that year from the MLMP data (see
Table 3). To calculate production from agricultural fields, we
first determined the number of milkweeds in fields. This is equal
to the area occupied by agricultural land multiplied by the milk-
weed density in fields (see Table 1). The number of agricultural
milkweeds in each year was multiplied by the eggs per agricul-
tural milkweed plant. For the years 2000–2003, we used Iowa
data for the eggs per agricultural milkweed (from Table 2). For
each of the other years, the egg density on agricultural milk-
weeds was taken to be 3.89 times the MLMP value for that year
(see Table 3).
Results
Estimates of milkweed numbers on the Iowa landscape
(Table 1) show that milkweeds declined in both agricultural
fields and non-agricultural habitats from 1999 to 2010. There
was a 31% decline for non-agricultural milkweeds and an 81%
decline for agricultural milkweeds with a 58% overall decline for
total milkweeds. In 1999, milkweeds in agricultural fields consti-
tuted 53% of total milkweeds, but by 2010 were only 24% of the
total. The 58% loss of milkweeds on the landscape actually
underestimates the loss of resource for monarchs, because most
Table 2. Maximum eggs per milkweed stem July through August for agricultural and non-agricultural sites in Iowa where ‘n’ is the
number of fields examined. Egg densities on milkweeds in agricultural fields were significantly higher than on milkweeds in non-
agricultural habitats in each year (2000: t = 3.97, d.f. = 11; 2001: t = 2.90, d.f. = 4; 2002: t = 3.35, d.f. = 4: t = 4.54, d.f. = 5; all
P-values < 0.02).
Year
Maximum eggs per milkweed
Ratio ag ⁄ non-ag
Agricultural Non-agricultural
Avg. SE n Avg. SE n
2000 0.796 0.140 10 0.197 0.049 8 4.05
2001 1.661 0.459 5 0.329 0.021 3 5.05
2002 0.659 0.123 4 0.205 0.056 4 3.21
2003 1.125 0.108 5 0.345 0.133 3 3.26
Average ratio 3.89
Herbicide use and monarch butterflies 5
2012 The Authors
Insect Conservation and Diversity 2012 The Royal Entomological Society, Insect Conservation and Diversity
