Edited by: Oded Berger-Tal, Ben-Gurion University of the Negev, Israel
Reviewed by: Sylvain Pincebourde, UMR7261 Institut de Recherche sur la Biologie de l'Insecte (IRBI), France; Yasuhiro Sato, University of Zurich, Switzerland
This article was submitted to Animal Conservation, a section of the journal Frontiers in Conservation Science
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
Migratory species are expected to demonstrate habitat selection that occurs at multiple spatial and temporal scales. Western monarch butterflies migrate seasonally to overwintering groves at geographically predictable locations along the coast of California. To date, overwintering habitat selection by western monarch butterflies has primarily been studied assuming the microclimate hypothesis. Specifically, that microclimate habitat selection occurs when monarchs form dense overwintering aggregations in overwintering groves. However, western monarch butterflies are migratory; thus, previous habitat selection studies could have commingled selection at different scales into a single local scale in the site of aggregation. Therefore, we explore monarch overwintering habitat selection to determine whether an explicit spatial framework is necessary. We studied nine groves on the coast of California, and at each we collected temperature, humidity, and light data from grove edges, grove interiors, and aggregation locations for several weeks during the overwintering season. We tested the hypothesis that monarchs aggregate in locations in groves that have a unique microclimate that is consistently selected across groves (the microclimate hypothesis). We find no evidence supporting the hypothesis that aggregation locations have a unique microclimate that differs significantly from that of other locations inside the grove or that aggregation locations are uniform in their microclimatic attributes across overwintering groves. Rather, we find that microclimatic attributes in aggregation locations vary spatially with latitude, and that aggregation conditions exist in a large portion of each grove. We conclude that it will be necessary to consider spatial effects when studying or managing western monarch butterfly overwintering habitats, and that interpretations of habitat selection to date likely commingle habitat selection on the local and geographical scales.
Many species demonstrate specific habitat associations and are known to select habitat components on different spatial and temporal scales (Johnson,
When a species is shown to express specific habitat affinities and those affinities or preferences are not contextualized in environmental space, geographic space, and time, then the scale of the affinity has not been considered. Selection on different scales would be confounded into a single conceptual or analytical spatially/temporally non-explicit scale such that the interaction of selection and scale would be obscured (Mayor et al.,
One example of species with specific habitat affinities is the migratory North American monarch butterfly (
The overwintering habitat of western monarch butterflies has been studied extensively and leads to acceptance of the “microclimate hypothesis.” In part, this is because habitat use by overwintering monarch shows habitat selection on various temporal scales, ranging from use of individual trees for a month or more (Anderson and Brower,
In the narrower literature on western monarchs, Leong (
Therefore, we present an analysis that tests for the existence of spatial non-hierarchical variation in groves (wherein selection may be expressed as relational or gradients of use) and hierarchical spatial variation between groves (wherein selection could be predicated on location). Our analysis partitions between the two scales. The microclimatic attributes we analyze are the ones thought to be under selection by monarch butterflies in coastal California's overwintering groves (Leong,
Our initial goal was to select 25 groves from the top 50 (according to Pelton et al.,
Nine groves sampled along the California coast in Ventura, Santa Barbara, and San Luis Obispo counties that met study design criteria (see text). The groves from south to north are: Arundell Barranca (V), Harbor Blvd (V), Tecolote Canyon (SB), Hollister Ranch (SB), Spring Canyon Vandenberg Air Force Base (SB), Black Lake (SLO), Oceano Campground (SLO), Pismo Beach State Park (SLO), and Morro Bay Golf Course (SLO).
The start and stop dates in each grove varied (
Start and end dates for data collection for each grove in the study. The start dates were delayed awaiting 1,000 or more monarchs per grove and varied as a function of monarch presence and counts and by access availability once the monarchs were present. The end dates were defined by monarchs departing entirely from a grove. The shortest sampling period was 30 days, and the longest was 61 days (mean = 43 days).
In each of the nine groves, data was collected in five locations, herein referred to as “groves” containing “sample locations” (
Sampling design relative to the aggregation's location in the groves. The first sample location was placed in the location of an aggregation (Aggregation). Two more sample locations were placed on the SE and NW edges of each grove relative to the aggregation's location to capture morning light and prevailing wind (SE edge and NW edge, respectively). Two interior sample locations were placed halfway between the aggregation's location and the grove's edge in the NE and SW directions (NE interior and SW interior, respectively).
We built small weather stations to collect climatic data and placed one in each of the five sample locations (
Weather stations were hung in an aggregation's location at a height of and within 2 m of the aggregation. All five stations in each grove were hung at the same height (±1 m). Thus, sensor heights represented selected habitat height and could vary between groves but not within. Each station was assembled in the following manner. In brief, a tree with a sturdy trunk was selected to which we attached a lock box (with screws or steel cable; Pro Strand; 1/8” diameter, part no: 21005100), a telescoping pole (Unger 30 Foot Pole, item #: U-TF900) base was inserted into the lockbox and attached, a second tree with an accessible branch higher than a monarch aggregation was selected, a steel cable (Pro Strand; 1/8” diameter, part no: 21005100) was put over this branch, and this cable, once attached to the tip of the telescoping pole, was used to lift and guide the pole into place. The sensors hung as a single array, had uniform directional settings, were drawn up to the tip of the telescoping pole via a cord (threaded through the pole), and could be lowered for data download.
More specifically, the following approach was used. Each wind meter (supporting all the sensors) was inverted, allowing for upward attachment of the meter's PVC support to the end of the telescoping pole. The wind meter's PVC was coupled with a custom-built directional attachment (the “insert”). The insert fit into a custom-built directional sleeve hanging down vertically from the end of each telescoping aluminum pole. The sleeve's direction relative to north was adjusted for each sample location. The insert, thus, rotated and locked with the sleeve, resulting in directionality for all the sensors. We then threaded a paracord through the base of the pole, the directional sleeve, and the insert, and tied it off at the tip of the insert. Thus, the insert and a weather station could be separated by gravity from the sleeve of the aluminum pole by allowing the station's weight to pull the paracord and drop or lower the station. Gravity would pull the paracord through the pole as we fed the cord into the pole, thus dropping a weather station to ground level while the pole remained deployed. After downloading data, we pulled the paracord, which pulled the insert (and sensors) back up into the directional sleeve, relocking the insert directionally. The paracord was then coiled and placed into the lock box and secured with a keyed padlock. The aluminum lock box (approx. 3” × 3” × 9”) was custom-built and secured by wood screws or a cable to a base tree. The extended telescoping pole was placed into a hinged aluminum socket in each lock box and secured with lock nut and bolt. The pole was then extended into the air to appropriate length and lowered, and its tip was secured to the cable. Finally, we raised the pole into position by pulling the cable over the branch. Once the pole was in position, the supporting cable was anchored by threading it through the lock box and securing it with crimp locks.
Microhabitat data were collected once in each grove, giving an attribute snapshot of late February. We quantified the amount of vegetative cover in the emergent layer, canopy, understory, and shrub layers, and ground cover layer by image analysis. Different lenses were used to capture images from different layers (details below). Habitat data were used to explore correlations between habitat attributes and microclimate attributes under habitat selection (if any) given that correlated habitat attributes might serve as tools for in-grove restoration or management.
Standing below weather stations, each station was placed at the center of a fisheye lens (Shuttermoon, 198°) image, viewed through a camera (iPhone 8). This lens captures a circular image, encompassing 198° out of a possible 360° (top 55% of a sphere with the observer at the center). Each lens was held 1.83 m above ground, so the resulting image represents vegetation from 1.83 m up, capturing emergent, canopy, and upper understory layers. Functionally, these vegetative layers could contribute a vertical component of light and wind abatement or obstruction.
Standing directly beneath each weather station and using a 0.63 × wide lens with a 74° field of view in portrait format, a photo was taken in the NW, SW, NE, and SE directions, resulting in a 360° view (minus a 16° gap between images). A 3-m extension pole was used so the images could capture understory, shrub, and ground cover layers and topographical hillside obstructions. Functionally, these layers and topographical features could contribute a horizontal component of light and wind abatement or obstruction.
A camera with a 0.63 × wide lens and facing downward was placed atop a 13-m pole directly below each weather station, resulting in an image covering a ground area of 4.5 × 4.5 m and capturing the ground cover layer.
In each station location, we collected five random samples of litter depth by creating a 4.5 × 4.5 m grid centered under each sensor array and using a random number generator to get two values (range 1–450 cm) that were used as x and y coordinates to determine where to collect a litter depth sample. Litter depth was measured using a meter stick placed vertically with its end touching bare ground.
The distance from below each sample location to the nearest nectar source was determined using a range finder (Leica LRF 800 Lazer Rangemaster) and by measuring the horizontal distance from the observer to the nectar source. Distances were readily observable, since most nectar was in stands or beds of low flowering plants and flowering shrubs rather than in the canopy. Each value was corrected for the height of the weather station to determine the straight-line distance.
We defined microclimate conditions in individual overwintering groves as local and part of environmental space as defined by light, temperature, and relative humidity. In this way, local variation was also non-hierarchical. We defined microclimate conditions that are spatial as those which are geographically variable because of their location in a geographic space. Latitude and longitude served as a proxy for functional attributes (i.e., distance from coastline and regional topography, etc.), which are covariates associated with space. If local microclimates were predictable in part by a location's geographic variables, then the microclimates would demonstrate hierarchical variation. All statistical analyses and figures were conducted in R version 4.0.5. The external packages used were tidyverse, nlme, stats, and multcomp.
The microclimate hypothesis posits that monarch aggregation locations have distinct climatic attributes (Leong et al.,
The microclimate hypothesis generally posits that unique microclimatic attribute values are created by specific microhabitat features (Leong et al.,
We conducted an ANOVA using the five sample locations as the fixed effect, the vertical and horizontal habitat obstructions as response variables, and grove as a random effect. For the type of nectar source available, we pooled the sample locations across groves to increase our sample size, since we had one record for each sample location in each grove. For type of nectar, we conducted a Chi-squared test using sample location as the fixed effect and nectar type as the response variable.
Habitat selection did not appear to happen at the level of individual sample location. Therefore, we tested a more general microclimate hypothesis that posited that monarch butterflies overwinter inside groves because the interior of groves contains suitable attributes that differ from those of the exterior of groves. From this more general hypothesis, we predicted that climatic attributes in three sample locations inside the grove would differ from climatic attributes in sample locations on two edges of groves. To test the prediction “interior locations,” aggregation, SW interior, and NE interior (
Tests of aggregation location effect.
MiDT | 0.9624 | |||||
MaDT | 0.0009 | a | a | a | a | b (>) |
ADT | 0.0306 | |||||
VDT | <0.0001 | a | a | a | a | b (>) |
MiDH | 0.2694 | |||||
MaDH | 0.9829 | |||||
ADH | 0.854 | |||||
VDH | 0.0093 | a | ab | ab | ab | b (>) |
MaDL | 0.0002 | a | a | a | a | b (>) |
ADL | <0.0001 | a | a | a | a | b (>) |
VDL | <0.0001 | a | a | a | a | b (>) |
To test the hypothesis (Leong et al.,
In order to directly test for a hierarchical geographic scale, we hypothesized that if there was a hierarchical or between-grove spatial scale, then there would be spatial autocorrelation in daily values of climatic data. Therefore, we predicted a spatial correlation with latitude for both temperature and light because of the correlation between latitude and day length. We focused exclusive on aggregation location data (defined as selected habitat) and conducted a Durbin-Watson test for spatial autocorrelation of each climatic variable for latitude, and thereby tested for a latitudinally correlated climatic niche (hierarchically or geographically correlated) rather than a uniform climatic niche (non-hierarchical or geography independent test above).
We hypothesized that some climatic variables would be correlated. We predicted that light and temperature would have a positive correlation, and that humidity and temperature would have a negative correlation. We also predicted that daily minimums, averages, and maximums in the same variable would be highly correlated. We used a linear correlation matrix to identify pairwise significant correlations (
The microclimate hypothesis posits that habitat selection occurs in an aggregation site and that microclimate attributes are selected
Magnitude and direction of significant pairwise differences (
We found some significant differences in temperature between aggregation locations and the two interior and two exterior locations of the groves (
On this fine-grained scale, aggregation location did not differ from any of the other sample locations for any of the humidity variables. There was no evidence for difference in minimum daily humidity (MiDH), average daily humidity (ADH), or maximum daily humidity (MaDH) across the sample locations. Variance in daily humidity (VDH) across the sample locations did differ (
On this spatial scale, we found a few significant differences in light between aggregation locations and the four other grove locations. There was a difference in maximum daily light (MaDL) across the sample locations (
Overall, aggregation location did show instances of being different from other locations for certain variables, but only as a member of a group that collectively showed a difference. In these cases, aggregation location grouped with either all the interior sites (NE, A, and SW) or all the interior sites plus the NW edge. These results do not provide evidence to support the within-grove microclimate hypothesis, meaning monarch aggregations did not seem to form in locations with a specific microclimate. If anything, on this fine-grained scale, temperature, humidity, and solar radiation effectively set the SE location apart from all the other locations combined, and a large portion of the grove appears to support a suitable microclimate.
Given we did not find strong evidence that aggregations formed where there was evidence for habitat selection, we proposed and tested a more general version of the microclimate hypothesis. We proposed the hypothesis that monarchs overwinter in groves because the interior, rather than just aggregation location, represented suitable climatic attributes. We tested this more generalized hypothesis by pooling aggregation location with two interior locations (which were not significantly different collectively or pairwise,
Tests of grove effect.
MiDT | 0.4106 | |||
MaDT | 0.0012 | < | < | |
ADT | 0.0211 | > | ||
VDT | 0.0002 | > | < | |
MiDH | 0.0168 | |||
MaDH | 0.8541 | |||
ADH | 0.4683 | |||
VDH | 0.0158 | > | ||
MaDL | <0.0001 | < | < | |
ADL | <0.0001 | > | < | |
VDL | <0.0001 | < | > | < |
Climate data comparing the interior, NW edge, and SE edge of the groves in December and January of the 2018–2019 overwintering season for
There were some differences in temperature between the interior of the grove and the edges (
There was no evidence that ADH or MaDH differed between the edges and interior of the groves. MiDH did differ between grove interiors and edges (
There were some significant differences in light between grove interior and edges (
Collectively, we found qualified support for the hypothesis tested that monarchs aggregated in groves because the interior and exterior of overwintering groves were significantly different in terms of temperature, humidity, and solar radiation (
Magnitude and direction of significant pairwise differences (
Relationship between MaDL, ADL, and MaDT. The two interior locations were pooled with the aggregation location, since these three were not significantly different (
The hypothesis that monarchs cluster in parts of groves that have unique microclimate conditions is based on the hypothesis that conditions are created by unique microhabitat attributes. We tested the microhabitat hypothesis by testing the prediction that aggregation locations would have microhabitat attributes that are different from those of all other (interior and edge) locations. We also considered the interior locations collectively, and thereby considered the microhabitat hypothesis as it relates to the more general microclimate hypothesis. Our focus was on vegetative (and other) obstructions, as these are expected to greatly impact wind (Leong,
In contrast to the predictions, we found no evidence that percent canopy cover (overhead obstruction component) differs across the five sample locations when accounting for grove differences. Likewise, there was no evidence of horizontal obstruction differences in the SW and NE directions across the five sample locations. In contrast, differences in horizontal obstruction were found in the NW and SE directions for some locations. The aggregation locations had less obstruction to the SE than to the NW location (
Percent of live ground cover was different across the sample locations (
Overall, we found no evidence supporting the hypothesis that aggregation location was significantly different in ground cover from all other sampling locations in the overwintering groves. With regard to three categories of ground cover, we only found that the aggregation location was significantly different from NW and NE in terms of percent of live and percent of dead ground covers. This result suggests that the microhabitat hypothesis is not correct, or that at best, it is only correct relative to parts of the grove, implying that the more generalized microhabitat hypothesis is more likely correct.
There was no difference in litter depth across the sample locations. There was no difference in distance to nectar or type of nectar across the sample locations.
Collectively, the results suggest that the microhabitat hypothesis is not at all correct, or at least not at a scale that includes multiple locations in the interior of the grove, and specifically not correct in aggregation location.
On the next larger scale (the portion of overwintering range sampled), we directly tested the potential for hierarchical effects on habitat selection. Specifically, we tested the hypothesis that there are no hierarchical effects, and that, therefore, monarchs overwinter in a uniform set of climatic conditions (i.e., within grove overwintering climatic niche). This was conducted by testing the prediction that climatic attributes in aggregation location would not be significantly different across overwintering groves (no impact of geographic space). Instead, they should be uniform (only a function of a selected environmental space). In complete contrast to the prediction (
Climate data for the interior of the groves in December and January of the 2018–2019 overwintering season for
Climate data for each sample location across the groves in December and January of the 2018–2019 overwintering season for
As an alternative to the hypothesis that microclimate selection is non-hierarchical and only based on a set of local environmental conditions that are repeatedly selected in aggregation locations across overwintering groves, we tested whether across the nine overwintering groves the aggregation locations showed a hierarchical latitudinally correlated climatic niche. For average values, there was a latitudinal correlation with ADL in aggregation locations (
We predicted correlations among variables and discovered correlations consistent with the predictions. There was a positive linear relationship between ADL and ADT (
Overall, we do not find support for the prediction that aggregation locations have distinct microclimates relative to the other sample locations in overwintering groves. The results do support the hypothesis that temperature, humidity, and light are microclimate attributes under selection by overwintering monarch butterflies. Specifically, we find that the aggregation locations are associated with differences in ADL, MaDL, VDL, MaDT, VDT, and MiDH. However, we find that the non-aggregation locations share the same microclimate attributes. Generally, the aggregation locations group with the other interior grove locations and one edge location. Likewise, we do not find support for the prediction that aggregation locations represent a variation constrained, unique, or uniform set of climatic attributes across overwintering groves. Instead, we find that microclimate attributes in the aggregation locations (except for MiDH) show a significant correlation with latitude. Therefore, it appears that monarchs select groves, that aggregation locations are not entirely conditioned on local microclimate attributes, that a large proportion of each overwintering grove represents suitable aggregation locations, and that what is selected as aggregation microclimate varies geographically. We conclude that if a climatic niche does exist, it is a geographically variable realized niche, thus broader than what can be discovered in a single grove, and broader than what has been recognized to date.
Generalizing, ecologically we find that aggregation locations become brighter, hotter, and not necessarily drier in a southern progression. Conceptually, we find that the conditions selected by overwintering aggregations of monarch butterflies are described by both local environmental attributes and spatial or geographically variable environmental attributes. Although it was difficult to find support for the microclimate hypothesis expressed as habitat selection in aggregation locations, it was possible to support a microclimate hypothesis on the non-hierarchical scale of grove interior. In stark contrast, for aggregation locations, it was quite easy to find support for latitude-correlated hierarchical habitat attributes. This means that it will not be possible to understand the overwintering ecology of monarch butterflies in California without explicitly considering scale. Local environmental attributes are nested in and interact with larger geographic variations.
Tracking aggregations over time proves to be more problematic than anticipated and presents a challenge to any study on habitat selection on the shortest time scale. Sensor groups needed to be moved to new locations as aggregations moved, but those locations blinked on or off in time, creating a dynamic system. We think that valid fine-grained (hourly) comparisons on the time axis will require real-time tracking of aggregations, which we were not able to accomplish. Instead, our results address habitat selection on a daily scale. Thus, we acknowledge habitat selection by overwintering monarch on a previously recognized temporal scale, including for months at a time (Anderson and Brower,
We acknowledge that we likely missed chronicling specific microclimate changes that triggered movement episodes and that at best we capture microclimate changes on a daily scale. However, we would argue that sorts of fine-scale studies that show movement can happen over hours (e.g., Leong et al.,
A significant and predictable complication was that attributes are correlated with each other. This suggests a potential limitation to separating climatic attributes statistically, although these conditions are ecologically correlated as well. In other words, our conceptual interpretation might be off if only one of these attributes is the driver and the others only seem important because they correlate with the driver, or, alternatively, they may all be important but to variable degrees or under different conditions. The data are, thus, inherently difficult to analyze. In particular, we suspect that a model selection approach would end up selecting a single attribute if the analysis incorporated an attribute disqualification step based on correlation. We also suspect that the collection or real-time data across multiple groves will be necessary in order to understand which attributes are the most important under what conditions and on what scale.
The microclimatic data from interior grove locations (including the aggregation locations) show that selection across the continuum of possible values is for high maximum daily light, low average daily light, and low maximum daily temperature (
Specifically, overwintering monarch butterflies exposed to energy fluxes from light can passively raise their body temperature, avoid energetically expensive shivering, and exceed the body temperature required for flight while also minimizing energy expenditure (Masters et al.,
The potential for the latitudinal dependent pattern we are suggesting seems to have been recognized by Chaplin and Wells (
The correlation between increase in day length and decrease in latitude (Hooker et al.,
The exploration of beneficial grove attributes based on grove locations rather than intrinsic (local internal environment) grove characteristics seems particularly ripe for exploration or modeling. For example, Fisher et al. (
Given climate change has effects on butterflies (Halsch et al.,
Finally, our finding that the aggregation location does not have a unique set of attributes suggests that we should question why monarchs aggregate. If it is not to concentrate itself into a limited and unique microclimate in the aggregation location, then maybe there is a fitness benefit derived from the aggregation behavior itself. It has been shown in monarchs and other insects (Brower et al.,
There does not appear to be a unique set of microclimate conditions across all the overwintering aggregation locations. Therefore, it is not surprising that habitat selection is not uniform across the overwintering range. Instead, we find variable selection for ADL, MaDL, VDL, MaDT, VDT, and MiDH. We also find that these attributes show a significant correlation with latitude, again, making it not surprising that habitat selection is not uniform across the range. In addition, overwintering habitat is not distributed continuously over the landscape, and instead it occurs in distinct groves. This makes it plausible that at least some of the variations among the groves are hierarchical (i.e., habitat selection predicated on geographic position). These hierarchical attributes depend on the pre-existence of another attribute (Kristan,
A variable regarded as relevant to habitat selection by overwintering western monarch butterflies is wind. We think it is important to speculate that wind has the potential to be significantly confounded if spatial and temporal hierarchical variations are underappreciated. In the long term (weeks to months), the prevailing winds in coastal California are from the west or northwest [Greeley et al.,
We consider some possible sources of a type II error (a potential to not find differences when they actually exist).
It was difficult to place sensors directly onto the aggregation of monarch butterflies given we wanted to avoid disturbing the aggregation or impacting clustering behavior. Therefore, sensor groups were placed approximately 2 m from the actual aggregation (but at the same height). Thus, if attribute(s) under selection varied over a horizontal distance of <2 m, we would not have had the resolution to measure them.
The historically low overwintering monarch population (Pelton et al.,
Our microhabitat data were collected at one time and toward the end of the overwintering season. The microhabitat data were a snapshot of mostly fixed attributes. However, for variable attributes (i.e., distance to and type of nectar source and percent live cover) and to the degree that they vary, it is possible we could have gotten different results had we taken the data at a different time.
It is possible that there are important characteristics that we did not monitor. Wind may be an attribute that defines suitable microclimate for monarchs given the conclusion that monarch clustering behavior may be heavily dependent on avoiding wind (Leong,
We cannot find evidence that groves need management for aggregation locations
We cannot define one set of suitable overwintering microhabitat attributes that apply to all aggregation locations in our suite of groves. This is likely because we found no singular suite of microhabitat attributes that correlated with microclimate. Individual attributes are considered below.
We find that aggregation locations were significantly different from NW locations in terms of having less vegetative obstruction in the SE direction. This pattern could be consistent with exposure to short-duration bright light and the associated energetically favorable habitat. Managing the interior of overwintering groves so some locations could have exposure to SE seems appropriate.
Groves are currently managed for tree covers as a means to dampen wind and light. We did not measure wind in the sampling locations within groves. Instead, we measured its proxy: vegetative obstruction. We were surprised that the aggregation locations differed by having less vegetative obstruction in the SE direction. Less obstruction to SE makes sense in terms of thermal ecology (see above) but suggests that aggregation sites are vulnerable to SE storm winds. Would we recommend using trees to dampen SE winds? No, dampening SE winds seems at odds with maintaining sufficient light from SE. An additional difference we found was that the NE sample locations (interior) have more obstruction to SE than expected. Thus, the NE sample locations would provide a more significant wind break in the SE direction (storm winds) than an average portion of the grove. Taken together, this suggests the following management strategy: focus on using trees for wind abatement should be in the NE section of overwintering groves, while some SE sun exposure should be maintained in other interior portions by not overplanting to abate SE winds.
We were likewise surprised that the aggregation locations did not differ from the other locations in some additional components of obstruction. Logically, it seems that if there is no difference in tree cover among locations in the groves we sampled, then there should also be no difference in wind. Therefore, we suggest more critical theoretical and empirical assessments of the relationship among vegetative obstruction, obstruction positioning, landscape obstruction, and wind in aggregation locations, and the contribution of hierarchical and non-hierarchical attributes at geographic and local grove scales.
Humidity does vary across overwintering groves, but it does not vary significantly with latitude. Minimum daily humidity (MiDH) and variance in daily humidity (VDH) both differ across some sample locations, but collectively, the interior locations are different from the edges. Thus, if humidity is a plausible management target, it would be at the level of grove interior and not at the level of aggregation. The aggregation locations did have a more live ground cover. Alonso-Mejia and Arellano-Guillermo (
To characterize a species' environmental associations, habitat preference and habitat selection implicitly require assumptions on appropriate spatial scale (Mertes et al.,
We find that non-hierarchical local environmental conditions and hierarchical latitudinally variable attributes are relevant to characterizing western monarch butterfly overwintering microclimate. We predict that for the overwintering distribution of monarch butterflies, as has been found by Siegloch et al. (
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
FV conceptualized the theoretical context and experimental design and contributed significantly to the writing. KS contributed to the theoretical context and experimental design, conducted the analysis, and contributed significantly to the writing. All authors contributed to the article and approved the submitted version.
United States Fish and Wildlife Service grants number 48733 and 52035: support for equipment and personnel. Larkin Group: donation to the Western Monarch Butterfly Fund, College of Science and Mathematics, California Polytechnic State University, San Luis Obispo: support for supervision of student research.
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
We thank Dr. Clinton Francis and Dr. Andrew Schaffner for their collaboration in the study design and analysis. We thank Charis van der Heide and Makenna Glisson for their assistance in equipment deployment and data collection. We thank Rob Brewster and Doug Brewster for their support in building the equipment for the study. This study arose at United States Fish and Wildlife Service Expert Elicitation that addressed the conservation status of western monarch butterflies and was principally supported by a grant from the Department of Interior, U.S.F.W.S, Coastal Division. We thank the organizers and participants of the elicitation for their collective guidance.