Asclepias dynamics on US rangelands: implications for conservation of monarch butter ﬂ ies and other insects

,


INTRODUCTION
Milkweed species (Asclepias spp.) play a critical role in the life cycle of the monarch butterfly (Danaus plexippus L.) and other insects.For example, monarch immature instars are obligate specialists and feed on the leaves of various species of milkweed and 10 other insect species that utilize various parts of the plant.The status of milkweed plants in the United States (US), including plant densities, distribution, and species hierarchy and associated stressors such as agricultural practices, has been the subject of considerable debate and controversy (Semmens et al. 2016, Agrawal 2017, Pleasants 2017, Thogmartin et al. 2017, Zaya et al. 2017, Pelton et al. 2019, Warner et al. 2019, Taylor et al. 2020).Milkweed dynamics have been at the forefront of discussions and research concerning the downward population trends of monarch butterflies and other insects throughout the US (Warner et al. 2019, USFWS 2020a, b).Yet, there have been no recent extensive regional or national assessments of the distribution and density of milkweed plants in the US (Agrawal 2017).
About 350 million years ago, two separate branches of insects diverged, one being the true bugs lacking metamorphosis, and the other being the megagroup, Holometabola, evolved complete metamorphosis, which included the beetles (Coleoptera ~300 M years ago), flies (Diptera ~175 M years ago), and the butterflies and moths (Lepidoptera ~150 M years ago) (Agrawal 2017).Each of these insect orders developed distinct ecologies and feeding traits.Today, 11 species of specialist insects are herbivores on milkweed: three aphid species extracting phloem sap; two lygaeid bugs consuming seeds; three beetle species boring through roots, tunneling through stems, and eating leaves; one fly mine invading leaf layers; and one moth and one butterfly species (the monarch) feeding on the leaves as caterpillars (Agrawal 2017).These diverse taxonomic groups of insect species that utilize milkweed shared a common ancestor (protoinsect) about 350 M years ago (Agrawal 2017).
Both the leaves and flowers of milkweed provide sustenance to monarch butterflies at different stages of their life cycle (USFWS 2020a).The leaves provide food to developing caterpillars, and the flowers later serve as excellent nectar sources for monarch adults throughout their migration (USDA-NRCS 2016, USFWS 2020a, b, Taylor et al. 2020).As monarch larvae ingest milkweed, they acquire toxic cardenolides, which deter predation (Roeske et al. 1976, Agrawal 2017).Malcolm and Brower (1989) identify at least 27 milkweed species that provide food for monarch larvae.Monarch larvae in North America also feed on native milkweed vines including honeyvines (Cynanchum laeve), twinevines (Funastrum spp.), and milkvines (Matelea spp.), but feed most often (by a wide margin) on Asclepias (Ackery and Vane-Wright 1984, Diggs et al. 1999, Luna and Dumroese 2013, Pocius et al. 2017).Milkweeds occur in many different habitats such as wetlands, streambanks, prairies, and other more semiarid and arid rangelands, woodlands, cropland, irrigation ditches, roadsides, and abandoned fields (Benson 1959, McGregor et al. 1986).The USDA Plants database identified 130 milkweed species that occur in North America (USDA-NRCS 2020).Most milkweed species are caulescent perennial herbs with milky latex and erect to prostrate stems.Some milkweed species are toxic to livestock and other animals, although mice and quail appear to be less sensitive (Burrows and Tyrl 2013).Toxicologists recognize two distinct types of milkweed (Asclepias spp.): a neurologic type (e.g., A. fascicularis and A. subverticillata) and a digestive tract/cardiac type (e.g., A. speciosa and A. syriaca).
The monarch butterfly occurs globally (distributed throughout 90 countries, islands, and island groups); however, the subspecies Danaus plexippus plexippus occurs in North America where there are three monarch populations: an eastern population that overwinters in the state of Michoac an, Mexico, a smaller western population west of the Rocky Mountains that overwinters on the coast of California, USA (Hill et al. 1976, Pelton et al. 2019, USFWS 2020a), and a small population that spends the entire year in southern Florida, USA.Both the western and eastern populations are distributed across millions of hectares of rangelands.Although the western population overwinters in Pacific coastal forests, western monarchs spend much of the breeding season and migratory seasons in open habitats with milkweeds and other nectar v www.esajournals.orgsources; most of these open habitats are rangelands (Xerces Society 2018).Eastern monarchs overwinter in Mexico's Oyamel fir forests (Abies religiosa), Mexico's most endangered forest type where only 2% of the original forest exists (Anderson and Brower 1996).Bioclimate predictions by S aenz-Romero et al. (2012) show that suitable climate niches for A. religiosa will rapidly decrease over the course of the century (69.2% by the decade surrounding 2030, 87.6% by that surrounding 2060, and 96.5% by that surrounding 2090).The Oyamel fir forest is characterized by a unique set of climate and physiographic features (Calvert and Brower 1986) where the cool climate supports the maintenance of monarch lipid reserves and suspends their physiological state and reproductive cycle (Herman 1985, Masters et al. 1988, USFWS 2020a).
The eastern monarch population cycle includes at least four post-wintering generations with most of the first generation completing their development in US states of Texas and Oklahoma-some first-generation butterflies complete their development farther north in the US states Missouri and southern Illinois (Miller et al. 2012, Flockhart et al. 2013, Agrawal 2017) (Fig. 1).Common dominant Texas milkweed species (in order of dominance) such as A. asperula, A. latifolia A. viridis, A. prostrata, A. stenophylla, A. tuberosa, A. syriaca, A. viridiflora, and A. speciosa have emerged and are usually in leaf stage by April, but the inflorescence is not fully formed until May-June (Diggs et al. 1999, and author observations), and occur sporadically in late summer and fall (only with sufficient late summer rainfall).When autumn migration occurs, the Texas milkweeds are usually dormant or have minimal green-up; however, vegetative reemergence may occur (without flowering) if late summer rainfall has occurred; therefore, they offer little resources as nectar plants for monarchs.The majority of the first-generation monarchs are the bulk of the northward progression to the Midwest and Northeast US, and some succeeding generations migrate to Florida, North and South Carolina (USA), and farther north into Canada.During mid-and late summer, the US Midwestern corn belt is home to more eastern monarchs than any other region (Hartzler and Buhler 2000, Prysby et al. 2004, Pleasants and Oberhauser 2013, Small 2015), but the US Great Plains rangelands also support millions of monarchs each year, and geographically is the most important fall US migratory pathway to Mexico (USDA-NRCS 2015).During the autumn migration to the overwintering grounds in Mexico, numerous southern Great Plains forbs provide nectar sources for monarchs.Forb species that are rated "very high" as preference plants are as follows: Conoclinium spp., Eupatorium serotinum, Grindelia papposa, Helianthus spp., Liatris spp., Solidago spp., and Verbesina spp.(USDA-NRCS 2016).
Over the last 20 yr, the western and eastern monarch populations have experienced significant population declines (Brower et al. 2012, Espeset et al. 2016, Pocius et al. 2017, Agrawal 2017, Thogmartin et al. 2017, Pelton et al. 2019, Voorhies et al. 2019, Warner et al. 2019, USFWS 2020a, b) and, in response, concern in the cause or causes has stimulated considerable interest resulting in various hypotheses and milkweed conservation efforts (Zaya et al. 2017, Taylor et al. 2020, USFWS 2020a).In December 2020, the USFWS listed the monarch butterfly in the US Federal Register as a candidate species under the Endangered Species Act of 1973.Regarding monarch population declines, Pleasants and Oberhauser (2013) estimated a 58% decline in milkweeds on the Midwest US landscapes and an associated 81% decline in monarch production from 1999 to 2010.Semmens et al. (2016) estimated that monarch populations have declined by ~80% in the last decade.On overwintering sites in California, monarch populations declined from 1.2 million monarchs in 1997 to 234,000 monarchs in 2014, an 81% decrease (Jepsen et al. 2015).The Xerces Society (2020) estimates that only 29,418 monarchs overwintered in California during the winter of 2019-2020. Pelton et al. (2019) ) stated that the western monarch population has plummeted and "may now be hovering at its quasi-extinction threshold." The USFWS (2020a) conducted a species status assessment (SSA) that "delineated the historical number and distribution of monarch populations, assessed the status and health of the currently extant populations, identified the key drivers of their health, forecasted the likely future change in these drivers and monarch population responses to these changes, and evaluated the consequences of the population v www.esajournals.orgresponses to monarch viability."In the scientific literature, various causes for the decline of monarch populations overwintering in Mexico and subsequent migration to and from the US have been proposed in two main competing hypotheses.Hypothesis 1 is the milkweed limitation hypothesis where declining milkweed populations are attributed to multiple factors such as urbanization, agricultural practices, and use of herbicides (Pleasants and Oberhauser 2013, Flockhart et al. 2015, Small 2015, Pleasants 2017, Zaya et al. 2017, Saunders et al. 2018, Warner et al. 2019, USFWS 2020a,b, Taylor et al. 2020).The USFWS (2020a) summarized that herbicide toxicity studies show that lepidopterans exposed to recommended field application rates may be detrimental-herbicides have been detected in milkweed plants growing in proximity to agricultural fields, but the direct effects are "unknown and likely to be highly variable."Hypothesis 2 is the migration mortality hypothesis where monarch populations are not limited because of the amount of habitat (Ries et al. 2015a, b, Inamine et al. 2016, Agrawal 2017, Agrawal and Inamine 2018), but due to stress, mortality, lack of nectar resources, and parasitic infection during migration over the last two decades.Controversy and research have promoted and disputed both hypotheses.With regard to hypothesis 1, Shultz et al. (2021) reported that the "toxicity of applied insecticides to aquatic invertebrates and pollinators has increased considerably-in sharp contrast to the applied amount-and that this increase has been driven by highly toxic pyrethroids and neonicotinoids, respectively."Regarding hypothesis 2, Inamine et al. (2016) discount the lack of milkweed plants as the driving factor of monarch population declines and hypothesize that the lack of nectar sources in addition to habitat fragmentation, perturbations associated with fall migration, and degradation of overwintering sites are the main causes for declining monarch populations.Taylor et al. (2020) cite a recent lack of supporting evidence for the migration mortality hypothesis; however, it remains a viable hypothesis and continues to be advanced as the cause for monarch population decline (Agrawal 2019, Popkin 2020).In the USFWS (2020a) species status assessment report, they summarize the following: "The primary drivers affecting the health of the two North American migratory populations are primarily: loss and degradation of habitat (from conversion of grasslands to agriculture), widespread use of herbicides, logging/thinning at overwintering sites in Mexico, plant senescence, and incompatible management of overwintering sites in California, urban development, and drought, and continued exposure to insecticides."Rankings of extent of impact of various factors on monarch decline for the eastern population based on expert-elicited data were as follows: (1) availability, spatial distribution, and quality of milkweed (avg.contribution 25%); (2) availability and quality of overwintering habitat (20%); (3) climate (12.5%); (4) availability, quality, and spatial distribution of migration resources (12%); (5) disease and natural enemies (9.5%); (6) insecticides (8%); (7) availability, spatial distribution, and quality of nectar resources (breeding) (5%); (8) road mortality and pollutants (3%); (9) biogeographical scrambling of milkweed spp.(includes non-native spp.) (2%); (10) other (2%); and (11) monarch releases, captive breeding, and translocation (1.5%) (USFWS 2020a).Expertelicited ranks for the western North American population were as follows: (1) availability, spatial distribution, and quality of milkweed (22%); (2) availability, spatial distribution, and quality of nectar resources (breeding) (18%); (3) insecticides (18%); (4) climate change effects via impacts to habitat (17%); (5) availability and quality of overwintering habitat (16%); (6) climate change via non-habitat mediated effects; and (7) all others (9%) (USFWS 2020a).
Both specific milkweed and butterfly species have been observed to be in decline.Twenty-one milkweed species have been listed as endangered, threatened, vulnerable, or of special concern (USDA-NRCS 2020).For many pollinating insects, both native and exotic introduced plant species have important roles in plant-pollinator networks (Vitt et al. 2020).Brummitt et al. (2008) estimate that "one of every five plant species are threatened by extinction … and preventing the loss of these species, and the services they provide are global conservation priorities."Dirzo et al. (2014) cite that lepidopterans have declined by 40% during the last 40 yr.There are currently 20 species of butterflies listed as either endangered or threatened (USFWS 2020a, b), and each species has certain requirements as to habitat conditions, plant community composition, and preferred host plants.
There is considerable interest in pollinators and the status of forbs that are integral to their life cycle; however, little is known about actual host plant populations, extent, and ecological status.Variations in weather patterns are thought to be responsible for short-term changes in butterfly abundance, whereas long-term changes are associated with habitat quality (Thomas 1984, Agrawal 2017).There is limited information about more specific environmental factors that affect community structure of butterflies (Gilbert 1989, Kocher andWilliams 2000).Kocher and Williams (2000) examined butterfly species richness and individual abundance across geographic gradients and disturbed habitats and found that richness of butterflies was greater at low US latitudes and near Rocky Mountain longitudes with greater topographical relief.Total butterfly abundance was highest in northern US latitudes and Great Plains longitudes, and butterfly species richness and diversity were lower in disturbed habitats.With the advent of the USDA-National Resources Conservation Service (NRCS) Rangeland National Resource Inventory (NRI) data set, it is now possible to study plant composition, distribution, and environmental gradients on non-federal rangeland ecosystems and plant communities in the US (Spaeth et al. 2003, 2005, USDA-NRCS 2018b).In this study, we used the USDA-NRCS on-site rangeland NRI data set v www.esajournals.org(hereafter NRI data set) to examine (1) extent and distribution of milkweed species on non-federal rangelands across the US, and (2) rangeland health associations and environmental gradients associated with milkweed species, a primary food source for monarch butterflies and other insects.The analyses provide a novel and unique characterization of milkweed plant coverages on western US rangelands.It is our hope that this updated assessment of milkweed dynamics will be useful for guiding monarch conservation efforts in North America.

Study plots
The spatial coverage of the NRI data set includes all 17 western states west of 95-degree latitude.The USDA-NRCS in cooperation with Iowa State University's Center for Survey Statistics and Methodology, Ames, Iowa, USA, has conducted such large-scale resource inventories for >65 yr to assess US natural resources on nonfederal lands.The inventory process has evolved from qualitative assessments in the early 1980s to robust quantitative field methods used since 2003 (Spaeth et al. 2003(Spaeth et al. , 2005)).The NRI rangeland on-site study is scientifically based, using a randomized statistical methodology where each NRI sample point has a calculated weighting factor, which represents the number of hectares the sample point represents on the landscape (Nusser and Goebel 1997, Nusser et al. 1998, Breidt and Fuller 1999, Yu et al. 2017).Each NRI sample point is located using latitude-longitude coordinates in a global coordinate system derived from the official NRI spatial database.A computerassisted simulation instrument with a Global Positioning System (GPS) is used to navigate to sample points and record data in the field.NRI plot data selected for and utilized in this study were acquired for all NRI points across the 17 western states for the years 2009-2018 (includes the years where full plant census was conducted).Fundamental NRI field sampling protocols for this data set are summarized below, followed by descriptions of all variable derivations and analyses used in the current study.Detailed explanations of NRI sampling protocols are provided in the USDA-NRCS NRI handbook (USDA-NRCS 2018a).

NRI field sampling protocols: Point attributes and vegetation
Each NRI sample is a circular macroplot with a diameter of 45.7 m centered around the associated georeferenced NRI sample point with two transects 45.7 m in length and oriented 45 degrees in both directions from magnetic north intersecting at the NRI sample point (USDA-NRCS 2018a).Sampling of NRI points occurs during the period of optimum development and emergence of the majority of plant species (i.e., set dates are established for each US state and particular rangeland community type).Photographs are taken at each end of the two transects to provide a visual record at the time of sampling.At each NRI sample point, the soil map unit and component, ecological site, hillslope angle and aspect, and vertical and horizontal slope shapes are identified.A soil map unit is a "collection of soil areas or non-soil areas (miscellaneous areas) delineated in a soil survey.Each map unit is given a name that uniquely identifies the unit in a particular soil survey area" (USDA-NRCS 2013a).The soil map unit name represents the dominant soil; however, more finely delineated soil components with individual names are classified with the map unit.An ecological site is a conceptual classification of the landscape (USDA-NRCS 2013b).It is a distinctive land unit based on a recurring landform with distinct soils (chemical, physical, and biologic attributes), kinds and amounts of vegetation, hydrology, geology, climatic characteristics, ecological resistance and resiliency, successional dynamics and pathways, natural disturbance regimes, geologic and evolutionary history including herbivore and other animal impacts, and responses to particular management actions.These discrete characteristics separate one ecological site from another.Individual ecological sites for US rangelands are described (ecological site descriptions) through a federal interagency program managed by NRCS (USDA-NRCS 2013b).Ecological site descriptions include narratives and dynamics of the historic plant community (HPC) or reference community that describes the ecological potential and natural range of variability associated with natural disturbance regimes of the ecological site, definitions of biophysical features, plantcommunity-scale dynamics, endemic plant species with expected production, interpretations v www.esajournals.orgfor land use and responses to management, a reference sheet for 17 rangeland health indicators, and a state and transition model depicting inherent vegetation state changes and ecological thresholds of community change.State changes and ecological thresholds may be permanent depending on the dynamics of the plant species and resiliency of species to disturbance (Weltz andSpaeth 2012, USDA-NRCS 2013b).
All plants on each NRI macroplot are identified to species (except when plant senescence inhibits identification) using taxonomic conventions from the USDA Plants Database (USDA-NRCS 2020b).In the field, plant census density estimates are made using five class designations for each species.The five density classes are as follows: (1) 1-10 plants, (2) 11-100 plants, (3) 101-500 plants, (4) 501-1000 plants, and (5) >1000 plants (USDA-NRCS 2018a).To address individual sampler bias, density estimates are compared among the field team members and any major discrepancies are discussed and corrected.Continued robustness of ocular census density estimates for repeat sampling is achieved by repeated yearly and seasonal calibration and training.
A suite of rangeland health indicators are assessed on each NRI macroplot using a specific protocol, Interpreting Indicators of Rangeland Health (IIRH) (Pellant et al. 2005(Pellant et al. , 2020)).The respective 17 indicators of rangeland health are evaluated to assess three ecosystem attributes (soil and site stability, hydrologic function, and biotic integrity).Table 1 shows the respective indicators used to assess soil and site stability, hydrologic function, and biotic integrity (Pellant et al. 2005(Pellant et al. , 2020)).For each NRI macroplot, the 17 indicators are evaluated on degree of departure (none to slight, slight to moderate, moderate, moderate to extreme, and extreme to total) from reference conditions identified in the associated ecological site description (Pellant et al. 2005(Pellant et al. , 2020)).
Apparent rangeland trend, grazing intensity, and active erosion and downcutting are also recorded for each NRI sample point.Apparent rangeland trend is defined as the direction of change in an existing plant community relative to the HPC.It is described as: toward-moving toward the HPC; not apparent-no change detectable; and away from-moving away from the HPC.Apparent trend classes are categorically coded as toward = 3, not apparent = 2, and away = 1.It is only applicable on rangelands that have ecological site descriptions identifying HPC composition.Grazing use is recorded for the following categories: currently grazed light <30% harvest = 1; currently grazed moderate 30-50% harvest = 2; and currently grazed heavy >50% harvest = 3. Active cutting related to erosion is evaluated at each NRI sample point and classified as one of the following: none = 1, stable = 2, unstable = 3, and downcutting = 4.

Variable derivations and analyses applied in current study
All summary and statistical analyses were conducted using SAS software, version 9.4 (SAS Institute, Inc. 2013).We derived a similarity index for each NRI macroplot by comparing current species reconstructed production with the ecological site reference state production (USDA-NRCS 1997, 2021).Current year's production for calculating the similarity index is obtained from five clipped quadrats (herbaceous spp.) systematically spaced along the two transects (NE-SW  1982, 1987, 1991, and 2009-2014 NRI data were calculated and compared for trends on non-federal rangeland.The reference state in the ecological site description represents the HPC species biomass production (USDA-NRCS 2018a, 2021).On western North American rangelands, the ecological site reference plant community composition or HPC is based on a 500-yr or shorter period immediately preceding European settlement (USDA-NRCS 1997, 2021, Winthers et al. 2005).As the similarity index increases, it signifies a higher degree of plant composition similarity to HPC plant species composition, while a low index is associated with dissimilarity.
We utilized the National Soil Information System (NASIS) (USDA-NRCS 2020a) to identify a suite of soil component parameters for each NRI sample point, including taxonomic soil order, textural class, salinity class, pH range, soil organic matter, drainage classes, and hydraulic conductivity (Ksat class).For each NRI sample point, these parameters were merged with respective soil component information in the NRI data set to fully characterize soils for our subsequent analyses described below.
Since the field NRI plant census data represent a range for each of the five density classes in the macroplot, we calculated a low, midpoint, and high value for each milkweed species on a per hectare basis.For example, if milkweed density was estimated at class 1 (1-10 plants), then low = 1, midpoint = 5, and high = 10 plants in the macroplot (0.164 ha).We then extrapolated milkweed plants on a per hectare basis.For density class five (>1000 plants), no low, midpoint, or high estimates were calculated.Final milkweed plant numbers were determined by multiplying the plant density per hectare for low, midpoint, and high by the statistical weighting factor (hectares) determined for the NRI point as provided in the NRI data set (Nusser andGoebel 1997, Nusser et al. 1998).
For this study, each of the assessed rangeland health indicators (Table 1) for each NRI point was converted to a numeric rating: none to slight = 1, slight to moderate = 2, moderate = 3, moderate to extreme = 4, and extreme to total = 5.For example, the ecological site rangeland health reference sheet defines the none-to-slight category (=1) for invasive plants as: "Nonnative invasive plants not present.If native invasive species are present, composition matches that expected for the ecological site" (Pellant et al. 2020).If the field evaluation documents invasive plants as "common throughout," the rating would be moderate to extreme (=4).Using the individual indicator rankings, a mean was calculated for the three rangeland health ecosystem attributes (soil and site stability, hydrologic function, and biotic integrity (Table 1)) at each NRI point.Non-metric multidimensional scaling (NMS) with Sorensen distance measure (Mather 1976

Milkweed plant densities
Twenty-two milkweed species were identified in the rangeland NRI data set and ranked in order of constancy (% of field samples) with representative state occurrence (Table 2).The occurrence of the specific milkweed species was verified using the Biota of North America Program (BONAP) (Kartez 2015, USDA-NRCS 2020b) for the individual species.In the survey, field teams sometimes encountered senescenced milkweeds with no flowers or pods, thus if they were unable to identify milkweeds to species, genus was recorded (Asclepias spp.); this v www.esajournals.orgrepresented 21.3% of the total sample.Ranking of milkweed species is shown in Fig. 2a, b, and the dominant seven species, based on frequency of occurrence >5% of the total sample, were A. viridis, A. syriaca, A. verticillata, A. speciosa, A. asperula, A. viridiflora, and A. latifolia.These seven species represent about 60.6% of estimated hectares with milkweed presence.When milkweed species were ranked on the basis of plant density (number of plants, excluding identification to genus only >100 M plants, midpoint estimate), the following dominant species comprised 57.8, 62.0, and 62.8% (low, midpoint, and high estimates) of the total milkweed population, respectively: A. speciosa, A. verticillata, A. viridis, A. syriaca, A. pumila, A. asperula, and A. viridiflora (Fig. 2a).Note that A. speciosa was the dominant species from the standpoint of plant density, whereas A. viridis was the most frequently occurring species on NRI sample points.
We also summarized milkweed plants for all non-federal rangeland, for two US regions (central and west), and for individual US states (Table 3).According to the NRI study, there were 1,303,372,193 plants based on the low-density estimate, 4,100,309,324 plants for the midpoint, and at least 6,919,517,259 plants for the high estimate on 13.2 M ha (Table 3).On a regional scale, the central region accounted for 84% of the hectares with milkweed species presence, and the west region, 16% (Table 3).Eighty-nine percent of the estimated milkweed plants (midpoint estimates) were found in seven states: Kansas (22.2%),Texas (19.4%),Nebraska (15.6%),South Dakota (10.7%),North Dakota (8.5%), Oklahoma (7.1%), and Montana (5.3%).
When examining latitude gradients, milkweed population numbers (based on midpoint estimates) exhibit a bell-shaped curve with the highest milkweed populations in latitude ranges N35-40 (1.46 B plants, 35.7% of the total, 5.8 M ha); N40-45 (1.01 B plants, 25.9% of the total, 2.9 M ha); and N30-35 (745.7 M plants, 18.2% of the total, 2.6 M ha), respectively (Fig. 3A).Milkweed population numbers decreased correspondingly to the far north and south latitudes with the mid-latitude exhibiting higher milkweed densities.v www.esajournals.orgOn the longitude scale, milkweed plant population numbers decreased from W-95-100 (2.0 B plants, 49.2% of the total, 6.6 M ha) to W-120-125 (13.4 M plants, 0.3% of the total, 0.71 M ha) (Fig. 3b).The density of milkweed plants and corresponding hectares for the top six milkweed species varied among the five latitude gradients (Fig. 4a).Based on latitude, N35-40 and N40-45 contain some of the highest plant densities and total hectares of A. viridis, A. syriaca, A. verticillata, A. speciosa, A. asperula, and A. viridiflora.In examining the longitudinal gradient, representative midpoint densities of milkweed species were highest in W-95-100 and W-100-105 for A. viridis, A. syriaca, A. verticillata, A. speciosa, A. asperula, and A. viridiflora (Fig. 4b).

Rangeland health summary
Rangeland health assessments can provide information on types, patterns, and severity of problems in rangeland ecosystems (Pellant et al. 2005(Pellant et al. , 2020)).The assessments are determined from the preponderance of evidence from 17 indicators (see Table 1), and individual indicators associated with the three attribute can vary along the matrix from none to slight to extreme to total departure from ecological site reference.In a USDA-NRCS (2018b) report, on 164.3 M ha of US non-federal rangeland (2011-2015 sample), the NRI sample showed that 12.7% of the hectares exceeded moderate departure from reference for soil and site stability (an increase of 3.4% from 2004 to 2010 sample).Conditions exceeding moderate departure for hydrologic function were 16% of the 164.3 M ha (an increase of 4.1% from 2004 to 2010) (USDA-NRCS 2018b).For biotic integrity, conditions exceeding moderate departure from reference were 22.7% (a 7.5% increase from 2004 to 2010).
As a comparison to the USDA-NRCS (2018b) report, the status of rangeland health for sites with milkweed presence (census survey initiated in 2009(census survey initiated in to present, 2009(census survey initiated in -2018 NRI data) NRI data) was determined for the three rangeland health attributes.At moderate or higher departure from reference, soil and site stability, and hydrologic function equaled 3% of the 13.2 M ha represented in the sample, and biotic integrity, 6%.In order to explore gradients of particular rangeland health and environmental quality indicators, we conducted an NMS ordination for seven dominant milkweed species (Fig. 6) using rangeland health indicator ratings for the main ordination matrix.Joint plot correlation arrows indicate strength and direction (+ vs. À correlation with axes 1 and 2) of the respective variables (Fig. 6).
The NMS ordination of rangeland health indicators converged on three axes for the final solution.The final stress for the two-dimensional solution was 6.3, and final instability was zero with 30 iterations.McCune and Grace (2002) state that NMS final stress and instability not exceeding 10 provide a "good ordination with no real risk of drawing false inferences."The number of iterations represents the steps that NMS performed to find the final solution.

Milkweed species correlations with environmental factors
Four flexible beta groupings were identified among the matrix of seven dominant milkweed species with NMS symbols, variable descriptions, and correlation of variables with ordination axes 1 and 2 (Fig. 6).Two rangeland health indicators, invasive plants, and functional structural groups were highly correlated with two milkweed species (A.speciosa and A. syriaca) along axis 1 (cluster group 2) (Fig. 6; See Table 5 for an explanation of the variable abbreviations and the Pearson r values).As explained in Methods, we scaled the rangeland health indicators and three attributes as follows: 1 = none to slight departure … 5 = extreme to total departure; therefore, the direction of the arrow shows a trend away from class 1 (none to slight) toward class 5 (extreme to total departure from HPC).Therefore, on sites containing these two species, there was a trend toward increasing invasive species and a shift in plant functional groups.Plant functional groups are identified in respective ecological site descriptions, which are used to rank rangeland health indicators.Pellant et al. (2005Pellant et al. ( , 2020) ) define functional structural groups as a suite of species groupings that are based on plant growth form (sod-forming, caespitose), similar shoot dynamics (height and volume), root morphology (fibrous, tap), photosynthetic pathways (C3, C4, CAM), and nitrogen-fixing ability (legumes, non-leguminous), and life cycle (annuals, perennials).Functional plant composition changes can have significant impacts on ecosystem processes that are indicative of plant successional changes and shifts away from HPC composition as depicted in ecological site state and transition models (Tilman et al. 1997, Briske et al. 2005, 2008).It is interesting to note that A. syriaca and A. speciosa were associated with  A. syriaca, A. verticillata, A. speciosa, A. asperula, and A. viridiflora) by US latitude, secondary y-axis is estimated v www.esajournals.orgincreasing changes in functional plant groups and invasive plants (also decreasing similarity index and apparent trend); however, they were correlated with more stable soil surface stability, hydrologic function, soil loss, and less compaction and bare ground.
Asclepias viridis (cluster group 1) appeared to be a transitional species midway between group 3 and group 4 and was correlated with decreasing annual production due to plant composition, plant life-form shifts, and grazing use (Fig. 6).In rating annual production as an indicator in the rangeland health model, none to slight departure represents > 80% of potential production where extreme to total departure is <20% of potential production.
Asclepias latifolia and A. asperula were classified into cluster group 4 with a suite of rangeland health indicators and overall assessments trending away from reference conditions.Grazing use was correlated with a trend of use in excess of 50% harvest.Improper grazing and overutilization are often associated with decreased production and plant cover over time (DiTomaso 2000, Pellant et al. 2005, Mysterud 2006).Rangeland health indicators: litter movement, litter amount, soil surface loss and degradation, active cutting, bare ground, soil compaction, plant mortality- (Fig. 4. Continued) v www.esajournals.orgdead and dying plants, and reproductive capability of perennial plants, were all correlated with axes 1 and 2 (Fig. 6, Table 5).A downward trend of all three rangeland health attributes (soil and site stability, hydrologic function, and biotic integrity) was correlated with axes 1 and 2.
Asclepias viridiflora and A. verticillata (cluster group 3) were correlated with conditions more closely related to HPC plant composition as described in the respective ecological site descriptions.Two environmental quality assessments, similarity index and apparent trend were correlated with NMS axis 2, which suggests that these two species grow on sites with plant species and composition more closely aligned with HPC conditions.Recall that the numeric scale for similarity index is 0 to 100%, whereas the apparent trend categorical classes are toward HPC = 3, not apparent = 2, and away from HPC = 1.
As a point of reference, we compared the change in non-federal rangeland area from the 1982, 1987, and 1992 NRI studies.According to USDA-NRI studies from 1982 to 2018, the average decline in rangeland hectares among the dominant states with high milkweed populations (Kansas, Texas, Nebraska, South Dakota, North Dakota, Oklahoma, and Montana) was 2.8%.Recorded total non-federal hectares in 1982 was 169.6 M ha; in 2018, total hectares equaled 164.3 M ha.With the decline of monarch populations, we questioned whether plant community composition has changed over the last 30 yr, which may have an effect on milkweed populations?Average similarity index estimates for the above seven US states between 1982 and 1992 were 58% compared with 49% for the 2004-2018 NRI sample.Since rangeland hectares has only declined by 3.1% and similarity index by 9% over   v www.esajournals.orgthe last 30 yr, we can only surmise that rangeland conditions on non-federal rangeland have been somewhat consistent with yearly temporal environmental fluctuations.
About 42% of the NRI segments with milkweed species occurred at 0-25%, 37% at 26-50%, 18% at 51-75%, and 3% at 76-100% HPC similarity (Table 6).As a comparison, 46% of the entire data set (sites with and without milkweed) was <25% similarity to HPC.In Fig. 6, A. viridiflora and A. verticillata were correlated with higher similarity to reference conditions where A. latifolia, A. asperula, A. syriaca, and A. speciosa were correlated with lower similarity index and apparent trend-rangeland health indicators and attributes were also concomitantly trending toward increasing site disturbance and a preponderance of evidence away from reference conditions (HPC).Three on-site situations can be associated with percent similarity and affect calculations: (1) Similarity indices as calculated by NRCS consider the percentage of the allowable composition for each native species (the actual on-site composition may be lower than allowable or the target production rate for the species, which will lower the calculation.Even though native species may still persist on the site, their presence or production can be low, which can affect the final percentage); ( 2  v www.esajournals.orgnative plants that are not endemic to the ecological site [e.g., juniper (Juniperus spp.), mesquite (Prosopis spp.), and creosote bush (Larrea spp.) that are native shrub species], but are not listed as part of the native plant composition for HPC (they would be considered native invasive species); and/or (3) the site has a high incidence and composition of non-native plant species.
Apparent rangeland trend is a qualitative determination of direction of change of the existing plant community compared with the HPC.Forty-four percent of the sample segments with milkweed plants were recorded as trending away from HPC, 35% as trend not apparent, 18% trending toward HPC, and 3% not applicable (where no description of native species composition associated with an ecological site can be determined, e.g., California annual grassland) (Table 6).Comparatively, 40% of the entire rangeland sample was recorded as trending away, 36% as trend not apparent, and 13% as trending toward HPC.

DISCUSSION
In the 2009-2018 rangeland NRI data set, we calculated estimated milkweed plants from the census protocol and found the milkweed population on non-federal rangeland (17 western states) at 4.1 B plants on 13.2 M ha (the midpoint value estimate from the study).Evaluation of milkweed populations and geographic locations shows that tallgrass, mixed-grass, and shortgrass prairie grasslands contain the bulk of milkweed across non-federal rangelands in the western and central US.The dominant level III Omernik ecoregions with the greatest milkweed populations were the Central Great Plains-27, Northwestern Great and the Cross Timbers-29, which collectively account for almost 77% of the estimated milkweed plants and hectare presence.The lat-long cell 1 (Kansas and northern Oklahoma) had the greatest milkweed density and is geographically associated with approximately 1/3 each of tallgrass, mixedgrass, and shortgrass prairie grassland; lat-long cell 2 (western Nebraska and South Dakota, southwest North Dakota, northeastern Colorado, and eastern Wyoming) with the second greatest milkweed density is about 2/3 mixed-grass and 1/3 shortgrass prairie; and lat-long cell 3 (northeast Texas and southern Oklahoma) with the third highest milkweed density is about 2/3 tallgrass (note: includes savanna grassland types as well) and 1/3 mixed-grass prairie grassland (Fig. 7).
We found that individual milkweed species were associated with varying habitat dynamics and disturbance regimes as indicated by v www.esajournals.orgsimilarity index (accession and departure from HPC), apparent rangeland trend, and three rangeland health attributes and individual indicators.Among the seven more dominant milkweed species represented in the NMS ordination, two species, A. verticillata and A. viridiflora, were associated with higher ecological condition exemplified by higher similarity indices and apparent trend toward HPC.Asclepias viridis was a transitional species between A. verticillata and A. viridiflora A. asperula and A. latifolia-the two latter species-were associated with higher disturbance and declining ecological conditions.The two remaining species, A. speciosa and A. syriaca, were associated with greater invasive plant composition and changes in plant functional structural groups.Both these species have been categorized as having invasive properties in the US (although both native species) (Adams 1983, Cronquist et al. 1984, Whitson et al. 1996, Stevens 2000, Lym and Travnicek 2010, Young and Eldredge 2012).As a side note, A. syriaca is an introduced exotic invasive species throughout Europe (Balogh et al. 2007, Csontos et al. 2009, Gall e et al. 2015, Kelemen et al. 2016, Gud zinskas et al. 2019, Szilassi et al. 2019), and although a rare alien species in Lithuania, A. speciosa is likely to occur in other regions of Europe (Gud zinskas et al. 2019).Asclepias syriaca is characterized as a fast-growing species with both clonal and seed-dispersing abilities in anthropogenic and natural habitats (Bhowmik and Bandeen 1976, Whitson et al. 1996, Agrawal 2004).Seed survivability is high, buried seeds remain viable for at least five years (Bagi 2008), and shoots are produced from the root system within a few weeks after germination, all of which are advantageous traits of invasive species and impact and decrease native plant species diversity (Goodwin et al. 2001, Richardson and Pysek 2006, Pysek et al. 2012, Kelemen et al. 2016, Follak et al. 2021).Asclepias speciosa can also be weedy or invasive (Adams 1983, Cronquist et al. 1984, Uley 2005); however, it is showy (hence the common name) and is planted in horticultural settings as the species has fragrant blossoms and a long boom period of two to three months (Uley 2005).
The decline of the monarch butterfly is well documented in the literature, and the persistence and well-being of monarchs is largely dependent on milkweed species (Ackery and Vane-Wright 1984, Malcolm et al. 1993, Hartzler and Buhler 2000, Luna and Dumroese 2013, Zaya et al. 2017, USFWS 2020a, b) and a variety of nectar plants during the fall migration (USDA-NRCS 2016, Agrawal 2017, 2019, USFWS 2020a).The majority of research and information on milkweed populations come from land types other than rangeland throughout the Midwest and eastern states.As a side note, in a 2016 census by the American Farmland Trust (AFT 2016), zero rangeland acres were recorded for Iowa.In this study, the milkweed midpoint estimate on US non-federal rangeland was 4.1 B plants (the high estimate, 6.9 B) on 13.2 M ha and milkweeds were located predominantly in prairie ecosystems (Fig. 7).However, Texas and Oklahoma are key states associated with the first migrating monarch generation, and milkweed estimates on non-federal rangeland in these states were as follows: low (322.9M), midpoint (1.1 B), and high (1.8B) estimates of milkweed plants on 3.9 M ha.Consider that these estimates does not include other land types and uses such as cropland, pastureland, forested areas, urban gardens, roadsides, and abandoned areas.
We evaluated plant community composition (based on similarity index) change on nonfederal rangeland over the last 30 yr and question if milkweed population and species dynamics have been consistent.Since similarity index estimates for Kansas, Nebraska, Oklahoma, Texas, South Dakota, North Dakota, andMontana between 1982 and1992 were 58% compared with 49% for the 2004-2018 NRI sample, we surmise that rangeland plant community composition on US non-federal rangeland has been somewhat consistent with expected annual temporal environmental fluctuations.We do recognize that v www.esajournals.orginvasive species are becoming more of a problem in Great Plains grasslands (USDA-NRCS 2018b); however, it is now possible to monitor community composition quantitatively.No NRI data exist that explicitly address milkweed populations from 1982 to 2004; however, given that two common milkweed species on rangeland [A.syriaca (10.9% constancy) and A. speciosa (9.8% constancy)] seem to be resilient to disturbances such as invasive plants and shifts in HPC functional plant groups, it is likely milkweed populations have remained functionally static.We propose a hypothesis: Since the similarity index has dropped by 9% over the last 30 yr, some milkweed species associated with higher regimes of disturbance may actually have increased-notably A. syriaca, a dominant species in the Midwestern US.
There are several different dynamics, which need to be sorted out with more research: (1) long-term data on the advancing monarch population from Mexico into Texas and Oklahoma, and subsequent generation cycles throughout the Midwest and eastern US; (2) the role and effect of predators and parasites in monarch populations; (3) the quantitative effects of insecticides (especially neonicotinoids) in suppressing monarch populations; and (4) the abundance and presence of flowering nectar plants during the autumn migration through Texas (Agrawal 2019).Agrawal (2017) mentions that the southern part of the autumn migration through southern Texas is not well understood and little is known about the last 25-30% of the monarch journey of one thousand kilometers, where monarchs need to accumulate lipid reserves (converting floral nectar to fat).Milkweed plants are typically not in bloom, and senescence is well advanced in Texas during the fall migration.Monarchs are not dependent on milkweed in fall and early winter; therefore, other plant species are relied upon for nectar reserves (USDA-NRCS 2016, Agrawal 2019).Agrawal (2019) summed up the situation concerning the decline of the Monarchs: "saving an iconic butterfly is important and would help us sustain beauty, wonder, and majesty in nature. . . the concern is much larger than a single species.The warning sign I see is that the health of our continent may be at risk.Monarchs are sentinels, traversing the continent and tasting their way as they move.Thus, to understand what is happening to environmental health … may be a source of information for both our own population and biodiversity." In conclusion, we document the previously unknown extent of milkweed availability throughout non-federal rangelands in the western US.Our findings show that milkweed is widespread, especially in mid-latitude Great Plains states.We also documented changes and trends in ecosystem health throughout the western half of the contiguous US.We also describe some broad environmental gradients associated with milkweed populations: Milkweed species seem to be largely associated with mollisols, generally well-drained, loamy and sandy loam soil textures, neutral pH, and non-saline soils.Finally, we described environmental relationships related to rangeland health for the seven most abundant milkweed species.It is our hope that this broad assessment of milkweed will (1) provide an invaluable reference for future assessments of pollinator habitat, (2) aid in recovery efforts for the monarch, and (3) serve as a model for robust plant-environment interaction assessments at large spatial scales.

Fig. 1 .
Fig. 1.Schematic map of monarch spring migration depicting migration patterns with locations of common Asclepias species.After completing development in southern Mexico, the first-generation monarchs move north, east, and southeast (note: some monarchs stay in Mexico).One population moves to the southwest US (dashed line).Some butterflies go west and join the California population (adapted from Agrawal 2017).

Fig. 3 .
Fig. 3. Graphs showing (a) estimated density (bars) of Asclepias plants (all species, midpoint value estimate) by US latitude, secondary y-axis is estimated hectares of Asclepias presence; and (b) estimated Asclepias plants (all species) by US longitude, secondary y-axis is estimated hectares (dashed line) of Asclepias presence.

Table 4 .
Ranking of lat-long cells in order of magnitude with milkweed (Asclepias spp.) densities >200 million plants (seven lat-long cell numbers correspond to Fig.5numbers) and ranking of within-cell associated Omernik level III ecoregions according to density of milkweed plants.
For example, in lat-long cell N35-40, W-95-100, ER level III Central Great Plains had the highest density of milkweed plants and High Plains had the lowest density of milkweed plants."No data" indicates ER III level classification occurs in cell, but no milkweed plants were recorded.B = billion, M = million.
) similarity index calculations used in this study do not include

Fig. 6 .
Fig. 6.Non-metric multidimensional scaling ordination of rangeland health indicators for seven dominant Asclepias species with joint plot arrows signifying strength of coefficient of correlations (Pearson's r) and four flexible beta cluster groups.See Table5for an explanation of the variable abbreviations and the Pearson r values.

Fig. 7 .
Fig. 7. Three basic grassland types in the US (shortgrass, mixed-grass, and tallgrass prairies) superimposed over lat-long cells that represent the ranking of Asclepias species density.

Table 1 .
Rangeland health indicators used to assess three ecosystem attributes: soil and site stability (SSS), hydrologic function (HF), and biotic integrity (BI).
, PC-ORD for Windows, version 7.07; McCune and Mefford 2018) was used to ordinate mean values for each of the 17 rangeland health indicators for the seven most dominant milkweed species (see Table 2 for list).Non-metric multidimensional scaling is well suited for the analysis of floristic data because it avoids the assumption of linear relationships between variables (McCune and Grace 2002).

Table 3 .
Estimated milkweed (Asclepias spp.) frequency of occurrence (constancy) in the NRI sample; estimated number of milkweed plants for low-, midpoint-, and high-density classes (as defined in NRI sampling protocol); and total estimated hectares of non-federal rangeland with milkweed presence for entire US, central and western US regions, and US state.

Table 5 .
Corresponding environmental variables displayed in Fig.6with coefficient of correlation (r) for both axes 1 and 2.

Table 6 .
Percent constancy of sites with milkweed (Asclepias spp.) and entire data set for similarity index classes and apparent trend (HPC = historic plant community).

Table 7 .
Soil and environmental quality data (% constancy of NRI points with milkweed [Asclepias spp.] presence, and % constancy of total rangeland NRI sample [with and without milkweed]).