Hummingbird foraging preferences during extreme heat events

Abstract

Climate change is projected to increase global mean annual temperatures as well as the frequency and intensity of extreme heat events. These changes are anticipated to alter the behavior of animals as they seek to thermoregulate in extreme heat. An important area of research is understanding how mutualistic interactions between animals and plants, such as pollination, will be affected by the cascading effects of extreme heat on animal foraging behavior. In this study, we used an experimental and observational approach to quantify the effects of extreme heat on hummingbird foraging preferences for nectar sources in shady versus sunny microsites. We also quantified pollen deposition using artificial stigmas at these sites to quantify potential cascading effects on plant reproduction. We hypothesized that hummingbirds would respond to extreme heat by preferentially foraging in shady microsites, and that this would reduce pollen deposition in sunny microsites on hot days. We found little support for this hypothesis, instead hummingbirds preferred to forage in sunny microsites regardless of ambient temperature. We also found weak evidence for higher pollen deposition in sunny microsites on hot days.

Full text

Ecology and Evolution. 2023;13:e10053.  
|
1 of 15
https://doi.org/10.1002/ece3.10053
www.ecolevol.org
1 | INTRODUCTION
The continuation of pollination services to plants in a warming cli-
mate is critical to sustaining plant biodiversity and ecosystem func-
tion. A met a- analysis of plant dependence on vertebrate pollinators
found that when birds were excluded from pollinating, fruit or seed
production was reduced by 46% (Ratto et al., 2018). Changes in pol-
linator behavior can have cascading effects on plant populations
(Anderson et al., 2011), as pollinator visitation rates have a pos-
itive effect on pollen receipt (Engel & Ir win, 2003). An estimated
87.5% of flowering plant species rely on animal pollination (Ollerton
et al., 2011). While some research has investigated the effect s of
higher ambient temperatures on insec t pollinators such as bumble-
bees, vertebrate pollinators have received relatively little attention.
Hummingbirds (Aves: Trochilidae) are a critical group of vertebrate
pollinators in the western hemisphere, and visit over 130 0 spe-
cies of plant s from 100 different families in the Americas (del Coro
Arizmendi & Rodríguez- Flores, 2012).
Hummingbirds are highly reliant on daily nectar from plant mu-
tualist s due to their high metabolic rates (Cronk & Ojeda, 2008;
González- Gómez et al., 2011; Shankar et al., 2 019), though they
do also eat insects as a source of amino acids that are absent from
Received:20January2023
|
Revised:4A pril20 23
|
Accepted :17April2023
DOI: 10.1002 /ece3.10 053
RESEARCH ARTICLE
Hummingbird foraging preferences during extreme heat events
Sabina Lucke Lawrence | Jenny Hazlehurst
This is an op en access ar ticle under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium,
provide d the original work is properly cited.
© 2023 The Authors. Ecol ogy and Evolution publishe d by John Wiley & S ons Ltd.
Depar tment of Biological Sciences,
Califo rnia State University E ast Bay,
Hayward, California, USA
Correspondence
Jenny Hazlehurst, Department of
Biologi cal Sciences, Califo rnia State
University East Bay, 25800 Carlos Bee
Blvd, Hay ward, CA 945 42, USA.
Email: jenny.hazlehurst@csueastbay.edu
Abstract
Climate change is projected to increase global mean annual temperatures as well as
the frequency and intensity of extreme heat events. These changes are anticipated
to alter the behavior of animals as they seek to thermoregulate in extreme heat. An
important area of research is understanding how mutualistic interactions between
animals and plants, such as pollination, will be affected by the cascading effects of
extreme heat on animal foraging behavior. In this study, we used an experimental
and observational approach to quantify the effects of extreme heat on hummingbird
foraging preferences for nectar sources in shady versus sunny microsites. We also
quantified pollen deposition using artificial stigmas at these sites to quantify potential
cascading effects on plant reproduction. We hypothesized that hummingbirds would
respond to extreme heat by preferentially foraging in shady microsites, and that this
would reduce pollen deposition in sunny microsites on hot days. We found little sup-
port for this hypothesis, instead hummingbirds preferred to forage in sunny microsites
regardless of ambient temperature. We also found weak evidence for higher pollen
deposition in sunny microsites on hot days.
KEYWORDS
extreme heat, foraging behavior, heat waves, hummingbirds, microsite preference
TAXONOMY CLASSIFICATION
Behavioural ecology, Global change ecology, Urban ecology, Zoology

2 of 15
|
LAWRENCE and HAZLEHURST
nectar (Battey, 2019; Clark & Russell, 2012). Hummingbirds have
a low energ y storage capacity and high fixed metabolic costs, and
thus are sensitive to daily fluctuations in metabolic costs and en-
ergy availability (González- Gómez et al., 20 11; Shankar et al., 2019 ;
Shankar et al., 2020). On extremely hot days, some birds will alter
their movement ecology to preferentially spend time in the shade,
or they may spend more time engaging in heat dissipating behav-
iors such as panting or spreading their wings. These behavioral
shifts may come at an opportunity cost. For example, male Southern
yellow- billed hornbills foraging on hot days panted more frequently
and spent more time in thermal refugia, resulting in decreased for-
aging success and body mass losses (Van de Ven et al., 2019). Du
Plessis et al. (2012) found that the foraging ef fort of Southern pied
babblers was not affec ted by ambient temperature, but foraging ef-
ficiency was negatively affected.
Hummingbirds may use behavioral thermoregulation techniques
to cope with high daytime temperatures (Powers et al., 2017).
Hummingbirds in southeastern Arizona were found to use heat dissi-
pation areas (HDAs) around the eyes, shoulders, and feet for passive
cooling when a thermal gradient existed between the ambient tem-
perature and the HDA surface temperature. Passive cooling is only
effective if ambient temperature is below HDA temperature. Powers
et al. (2017) found that the thermal gradient driving passive heat dis-
sipation in five species of North American hummingbird disappears
between an ambient temperature of 36 and 40°C. Powers (1992)
found that C. anna approaches hyperthermia at temperatures of
37°C. When temperatures exceed 35°C, hummingbirds may need to
use behavioral thermoregulation, for example by retreating to shady
microsite s, to balance thei r daily energ y budgets (Powers e t al., 2017).
All five hummingbird species studied experienced long daytime pe-
riods when their thermal gradient disappeared. Even in the absence
of a thermal gradient from their HDAs, broad- billed hummingbirds
maintained their mean body surface temperature near 38°C (Powers
et al., 2017). This suggests some alternate cooling mechanism such
as behavioral thermoregulation. Hummingbirds of some species
decrease certain activities, like territorial defense behavior, past a
threshold ambient temperature of 19.9°C (González- Gómez et al.,
2011). An organism's operative temperature includes the combined
effects of radiative, convective, and conductive heat flux specific to
that organism (Van de Ven et al., 2 019). While ambient temperature
is immutable until the weather changes, the operative temperature
of an animal can be changed in several ways, including by seeking
out increased or decreased insolation, increased or decreased wind
exposure, and shelter from or exposure to precipitation. The differ-
ence between ambient and operative temperatures can be physio-
logically significant. For small avian species, operative temperature
can var y by up to 15– 20°C between shaded and sunny microsites
within the same habitat, dramatic ally changing the physiological cost
of thermoregulation (van de Ven et al., 2019). Due to hummingbirds'
high surface area to volume ratio, they experience a greater increase
in operative temperature than larger avian species when they move
from a shaded microsite into a sunny one (Abdu et al., 2018). If hum-
mingbirds use behavioral thermoregulation in a way that alters their
foraging behavior, it could have potential for cascading effects on
plant reproduction.
California is both a global floral biodiversity hotspot with many
unique ornithophilous plants (Myers, 2000) and is expected to ex-
perience more extreme heat waves and increased temperatures in
the future. Northern California has experienced some of the great-
est temperature anomalies in daily maximum temperatures (Luo
et al., 2017). Between 1950 and 2005, the San Francisco Bay Area
average annual maximum temperature increased by 0.95°C, and
is likely to see additional annual mean warming of 1.8°C (Ackerly
et al., 2018) making it an ideal location to study how heat waves may
alter hummingbird foraging behavior and have cascading effect s on
pollination. In California there is only one year- round resident hum-
mingbird throughout the state, the Anna's hummingbird (Calypte
anna). The Anna's hummingbird has experienced an expansion of its
geographic range northwards, which may be related to supplemen-
tal feeding at backyard feeders, introduced floral species such as
Eucalyptus, and climate change (Battey, 2019; Clark & Russell, 2012;
Greig et al., 2017). The adaptability of C. an na to land use change
and climate change make it an ideal species in which to study the
potential for cascading effects of behavioral changes due to cli-
mate change on pollination, as it is likely to be an abundant visitor
of California's unique and threatened native plants in the future. In
ornithophilous plants that Anna's hummingbirds might visit, pollen
is usually deposited by the anthers onto the forehead or throat of
the bird during a visit to a flower. That pollen is then transferred to
another flower's stigma during pollination.
In this study, we test the hypothesis that Anna's hummingbirds
(Calypte anna) would preferentially forage at experimental feeders
in shady microsites versus sunny microsites on hot days (operative
temperature in the sun >35°C) when compared to average days
over the summer in California. We also test for differences in pollen
deposition on artificial stigmas at feeders bet ween sunny and shady
microsites on hot days compared to average days. We hypothesized
that pollen deposition would be greater at shady versus sunny mi-
crosites on hot days as compared to average days.
2 | MATERIALS AND METHODS
2.1 | Study organism
While most North American hummingbird diversity is concen-
trated in Mexico, approximately 24 species occur in the US and
Canada. Eight hummingbird species occur in California as perma-
nent or seasonal residents, with an additional six species recorded
as rare vagrants. Anna's hummingbird (Calypte anna; Figure 1) are
the only year- round resident found throughout the state, and is the
most commonly observed hummingbird in California. Anna's hum-
mingbirds are on the larger side for hummingbirds in California,
with a leng th of 10 cm and a weight of 3–6 g. Calypte anna uses a
wide variety of habitats for foraging and nesting, including chapar-
ral, oak and riparian woodlands, savannah, coastal scrub, and urban

|
3 of 15
LAWRENCE and HAZLEHURST
and suburban gardens (Clark & Russell, 2012). Some of the most
frequently visited California native plant nectar sources include
gooseberries (Ribes), monkeyflower (Diplacus), sage (Salvia), keckiel-
las (Keckiella), penstemons (Penstemon), and willowherb (Epilobium)
(Clark & Russell, 2012). Most interactions between conspecifics are
agonistic, including chases, vocalizations, and occasional contacts
(Clark & Russell, 2012; González- Gómez et al., 2011). Males defend
breeding territories during the winter, and will attempt to mate with
any female who enters the territory by courting her with dramatic
display dives and advertising flights (Clark & Russell, 2012; Stiles,
1971). After the breeding season concludes in May, banding studies
have revealed that many individuals migrate short distances to new
feeding grounds either at a higher elevation or inland to the south
oreast,of tenreturningtotheiroriginalareaafter2–3 months(Clark
& Russell, 2012). However, because there are multiple patterns
of local migration for C . anna, a site may have a year- round hum-
mingbird presence but substantial turnover of individuals (Clark &
Russell, 2012).
2.2 | Field site
This study took place in a seminatural environment on the California
State University East Bay (CSUEB) campus in Hayward, California
in the US (Figure 2). The study site consists of a mix of parking lots,
buildings, paved walkways, and ornamental landscaping. The cli-
mate in the region is Mediterranean, with cool, mild winters and hot,
dry summers, and the heat and dryness generally extend well into
November. In the City of Hayward over the period 1991– 2020, the
summer average temperature was 18.9°C, summer high tempera-
ture was 23.9°C, and summer minimum temperature was 13.8°C,
and the average humidity was 63%. This study took place from June
11, 2021– November 19, 2021, encompassing the hottest time of
the year, and there was no precipitation during this time (Figure 3).
Hot trials or hot observation sessions were categorized as when the
maximum temperature in the sun was greater than 35°C.
FIGURE 1 Anna'shummingbirds(Calypte anna) visiting a
hummingbird feeder hanging on a porch in summer.
FIGURE 2 Locationsoffeedertrials(Feeder/SiteB)andseminaturalbehavioralobser vations(SitesA,B,C,D,E,F,G,H)ontheCalifornia
State University East Bay (CSUEB) campus in Hayward, California. Inset shows Feeder Trial set up at Site B.

4 of 15
|
LAWRENCE and HAZLEHURST
2.3 | Feeder trials
Glass feeders with five feeding por ts and perches (Perky Pet, Inc.)
containing a 30% sucrose solution were placed under a tent canopy
withdimensionsof3.05 m × 3.05 m × 2.44 m(length × width × height;
Figure 2). Nectar concentrations in flowers visited by hummingbirds
exhibit great natural variation, but average 25% sucrose (Pacini et al.,
2003). Based on preliminary data from a low- volume sucrose refrac-
tometer conducted during the period of this study, field sucrose con-
centrations at the study site ranged from 15 to 55%, and thus 30%
is on the high end of average based on the literature, however, well
within what a hummingbird might encounter at our field site. The
sunny microsite treatment had the frame of the canopy, but not the
shade cloth. The shady treatment had the shade cloth on and the
feeder was within the shaded area for the duration of observations.
The shade cloth of the canopy was made with 300D Polyester with a
Polyurethane lining capable of withstanding 99% of ultraviolet rays.
Asingleshadeandasinglesuntreatmentwereplaced10 mapartin
an open lawn to present foraging hummingbirds with a choice be-
tween the two microhabitats, and treatments were shuffled in posi-
tion randomly during each session to avoid any spatial bias. Feeders
were obser ved once pe r day between 1 2:30 p.m. and 5:30 p .m. to
capture the hottest part of the day, after an acclimatization period of
1 h,andeachobservation session averaged135 min(min = 111 min,
max = 213 min,SD = 22 min).Totracktemperatureineachmicrosite,
iButton temperature sensors and dataloggers (Maxim Integrated,
Inc.) were at tached to the feeders using modeling clay. For each ses-
sion the following environmental data were recorded: cloud cover
(sunny, partly cloudy, overcast), observation start and end times, the
numberofpeoplethatwalkedwithin25 mofthefeedersduringthe
observations (low: <10, moderate: 10– 20, and high: >20), and the
estimatednumberofopenfloralinflorescenceswithin25 m of the
feeders. We quantified hummingbird preference by number of visits,
defined as any time a bird entered the area under the canopy, forag-
ing visits, defined as any time a hummingbird inserted its bill into the
feeder, and visit duration, defined as the amount of time the bird
spent in the canopy area.
2.4 | Pollen deposition
We measured pollen deposition during feeder trials by placing
artificial stigmas above the flowers on the feeding ports of the
hummingbird feeders. Artificial stigmas were made by placing a
1 cm × 1 cm × 1 cm cube of fuchsin-stained pollen collecting gelatin
(Kearns & Inouye, 1993) inside a metal gemstone setting and attach-
ingittoalengthofwire5 mminlengthtosimulatethestigmalength
and position of California fuchsia (Epilobium canum), a locally abun-
dant hummingbird- pollinated California native plant. At the end of
a feeder trial, artificial stigmas were collected and mounted on glass
slides for analysis. Slides were obser ved at a magnification of 100×
using a digital microscope, and any pollen grains present were photo-
graphed. Pollen grains deposited per feeder trial were counted manu-
ally in slide photos using the program ImageJ (Schneider et al., 2012).
2.5 | Seminatural behavioral observations
In order to provide context to our feeder experiments, we also con-
ducted observations of free- foraging birds in the campus landscape
on extremely hot and average days during the same time period (11
June 2021– 19 November 2021) as the feeder trials to determine if
they preferred shady microsites to sunny microsites on extremely
hot versus average days. Seminatural observations and feeder trials
were both spread out evenly during the study period. Birds were
observed at 8 different locations on the CSUEB campus (Figure 2)
that had blooming flowers and shady and sunny microsites. Each
locationwasobser vedfor45 minbetween12:30and5:30 p.m.We
conducted a total of 20 sessions across 8 loc ations on the C SUEB
campus(samplesizesforlocationA = 3sessions,B = 2,C = 4,D = 3,
E = 2, F = 3, G = 2, H = 1) of the se, 8 were catego rized as hot ses-
sions, and 12 were categorized as average. Sampling was uneven
between sites because some sites stopped flowering while others
began to flower. During sessions, we used scan sampling to re-
cord the behavior and microsite (sunny vs. shady) of every visible
humming bird within a 25 m ra dius at 5 min inter vals. The follow-
ing behavior categories were recorded: perching, preening, gaping,
vocalizing, aggression, flying through, and foraging (fly- catching
or nectaring; Table 1). If a bird exhibited a combination of simulta-
neous behaviors, all simultaneously occurring behaviors were re-
corded (for example, perching and vocalizing). If a bird exhibited
multiple sequential behaviors, only the behavior(s) that occurred
at first sighting were recorded. During each session, the following
environmental variables were recorded: cloud cover (sunny, partly
cloudy, overcast), observation start and end times, the number of
peoplethatwalkedwithin25 mofthefeedersduringtheobserva-
tions (low: < 10,moderate:10–20,andhigh:>20), and the estimated
numberofopenfloralinflorescenceswithin25 mofthelocation.
FIGURE 3 Averagedailytemperaturesinsunatfieldsiteduring
months of feeder trials and seminatural observations. Error bars are
standard deviations (SD).

|
5 of 15
LAWRENCE and HAZLEHURST
2.6 | Data analysis
Feeder sessions and seminatural observations were categorized as
hot or average based on whether the maximum temperature in the
sun exceeded 35°C for over 50% of the trial. All statistical analy-
ses were conducted in the software program R (R Core Team 2022).
The effect of temperature on number of visits, visits per hour, and
visit duration to each category of microsite was analyzed using gen-
eral linear mixed models (GLMM) using the R package “lme4” (Bates
et al., 2015). For feeder trials, candidate GLMMs were constructed
with the tot al number of visits, average number of visits per hour
(rounded to the nearest integer), or visit duration as response vari-
ables. Microsite and temperature were included as interacting fixed
effects, and separate models were construc ted using categorical
(hot/average) versus maximum temperature in the sun as the tem-
perature term. Julian date, session ID, and human presence were in-
cluded as random effects. Session duration varied slightly between
trials, so it was included as an offset term in all models. Total number
of visits and visitation rate (visits/h) were modeled using a Poisson
distribution. Visit duration data was heavily skewed to the right, and
thus was centered and scaled by standard deviation before running
models (Bolker et al., 2009; Meehan et al., 2020; Schielzeth, 2010).
Initial data exploration showed the pollen count data were overdis-
persed, so an observation- level random effect was added to all pol-
len deposition models to correct for overdispersion. Model selection
was done for each test (response of total visits, visit rate, or visit
duration and fixed effect of categorical or continuous temperature)
by constructing models with every possible combination of fixed
and random effect s. The package ‘MuMIn’ was then used to selec t
thebestmodel(Bartoń,2022). Candidate models were checked for
normal distribution of residuals based on Q– Q and Shapiro– Wilk
tests. Models were selected by the second order Akaike Information
Criterion (AICc) (Bolker et al., 2009). Post- hoc pairwise comparison
for significant interaction terms was conducted using the package
‘emmeans’ with Bonferroni corrections (Lenth, 2022).
For seminatural obser vations, we constructed GLMMs with a
Poisson distribution using the cumulative number of foraging vis-
its to a flower as the response variable, the interaction of microsite
(whether individual was observed visiting a flower in the shade or
the sun) and ambient temperature (categorical or maximum tem-
perature in the sun, as in feeder trial models) as the fixed effect, and
a unique session ID, Julian date, human presence (using same sc ale
as in feeder trials), location, and floral abundance as random effects.
Observation session duration was included as an of fset term. Model
selection, quality control, and post- hoc pairwise comparisons were
conducted following the same process as in feeder trial models.
3 | RESULTS
3.1 | Feeder trials
We conducted a total of 30 feeder trials, during which we recorded
529 foraging visits by Anna's hummingbirds; 31% (N = 164) of all
visits were in the shady microsite, while 69% (N = 365)were inthe
sunny microsite. Feeder trials were approximately balanced between
hot (N = 14)andaverage(N = 16)ambienttemperatures.
The best model for cumulative number of visits included the
interaction of microsite and maximum sun temperature as fixed
effects, trial ID as the random ef fect, and an offset term for trial
duration (Figure 4; Tables 2 and A1). We found a significant effect of
microsite, but not ambient temperature or the interaction of ambi-
ent temperature and microsite. Sunny microsites had a significantly
higher number of visits than shady microsites regardless of ambient
temperature(Coef = 1.50, z = 2.83,p < .01). A model with the same
random effects but with temperature as a categorical variable found
a similar result (Tables 2 and A1), with sunny microsites receiving
significantly more visits than shady microsites regardless of ambient
temperature(Coef = 0.82,z = 4.85,p < .001).
When considering visitation rate as the dependent variable, the
averagevisitationrate(visits/hour)was2.67 ± 1.65(SD)inshadeand
5.75 ± 3.00(SD)insunduringhottrials,and2.02 ± 1.08(SD)inshade
TABLE 1 Descriptionsofbehaviorsrecordedduringseminatural
behavioral observation sessions.
Behavior Description
Perching Sitting on a perch
Foraging - nectar Making physical contact with a flower
Foraging - insect Fly- catching or gleaning arthropods
Flying Actively flying through the field of view
but not hover- feeding.
Preening Moving own feathers with bill or foot
Vocalizing Making a vocalization of any kind
Aggression Chasing or direc ted physical contact with
another bird
FIGURE 4 Predictedeffectsplotshowingcumulativenumber
of visits by Anna's hummingbirds per feeder trial in shade and
sun microsites when maximum sun temperature is used as a fixed
effect. Error bars are standard errors (SE).

6 of 15
|
LAWRENCE and HAZLEHURST
and4.91 ± 2.07(SD)insunduringaveragetrials.Thebestmodelfor
visitation rate included the interac tion of microsite and maximum
sun temperature as the fixed effect, trial ID as a random effect,
and trial duration as an offset term (Figure 5, Tables 3 and A2). The
only significant fixed effect was microsite, with sunny microsites re-
ceiving significantly more visits per hour than shady microsites re-
gardles s of ambient tempe rature (Coef = 1.76, z = 2.17,p < . 05). An
identical model with temperature as a categorical variable was the
second best model during model selection, and had a similar result,
with sunny microsites receiving significantly higher visitation rates
regardless of temperature category (Tables 3 and A2;C oef = 0.96,
z = 3.68,p < .001).
When considering visit duration during feeder trials, the av-
erage visit duration in the shade was 41.41 s ± 25.34 (SD) and
52.85 s ± 12.79(SD)inthesunduringhottrials.Theaveragevisitdu-
rationintheshadewas28.81 s ± 18.23(SD)and45.61 s ± 14.93(SD)
in the sun during average days. The best model for the average visit
duration included the interaction of microsite and categorical tem-
perature as fixed effects, Julian day as a random ef fect, and trial du-
ration as an offset term (Figure 6, Tables 4 and A3). No other models
had a ΔAICc < 2.Thefinalmodelfoundasignificanteffectof both
temperature categor y and microsite, with higher visit duration on
hottrials(Coef = 0.50,t-value = 1.48,p < .05)andlowervisitduration
intheshadethaninthesun(Coef = −0.67,t-value = −2.22,p < .01).
3.2 | Seminatural behavioral observations
We conducte d a total of 20 sessions across 8 locations on the CSUEB
campus(samplesizeperlocation:siteA = 3sessions,siteB = 2,C = 4,
D = 3,E = 2,F = 3, G = 2, H = 1);of these,8 werecategorized ashot
sessions, and 12 were categorized as average. Out of a total of 409
individual observations of behavior across all sessions, 107 were of
foragin g. Session dur ation averaged 47 min ± 1. 85 (SD), for a total
ofapproximately17 h ofobservations.Hummingbirdtimebudgets
were apparently different on hot and average days (Figure 7) with
greater incidences of foraging, aggressive interac tions, and flying
on hot sessions as compared to average sessions. Birds spent less
time perching, vocalizing (for example calling), and preening on hot
sessions. Hummingbirds also apparently used microsites differently
(Figure 8). Birds spent more time foraging, flying, and in aggressive
interactions in sunny microsites, and more time perching, preening,
and vocalizing in shady microsites. Birds were observed feeding at 9
different flowering ornamental plants during sessions (Table 5), two
TABLE 2 Finalmodelforcumulativenumberofforagingvisitsinfeedertrialswithambienttemperatureasthemaximumtemperaturein
the sun during the session and temperature as a categorical variable where “hot” is categorized as an ambient maximum temperature in the
sun >35˚Cand“average”is≤35˚C.
Variable Coefficient SE Lower 95% CI Upper 95% CI z- value p- value
Model:Num.Visits ~ microsite*max_sun_temp + (1|session _id) + offset(session_duration)
Intercept −1.2 0.63 −2 .4 3 0.03 −1. 92 .06
Microsite (sun) 1.5 0.53 0.46 2.25 2.83 <.01* *
Max.suntemp(˚C) 0.02 0.02 −0.02 0.05 0.89 .37
Microsite(sun)*Max.sun
temp
−0.02 0.01 −0.05 0.01 −1. 34 .18
Model:Num.Visits ~ microsite*temp_cat + (1|session_id) + offset(session_duration)
Intercept −0.74 0.2 −1 .1 2 −0.35 −3.75 <.01*
Microsite (sun) 0.82 0.17 0.49 1.15 4.85 <. 001* *
Temp category (hot) 0.14 0. 24 −0.34 0. 61 0.57 .53
Microsite(sun)*Temp
category (hot)
−0.02 0.2 −0.42 0.37 −0.11 .91
Note: Variables with significant ef fects are shown in bold.
*p < .05;**p < .01;***p < .001.
FIGURE 5 Effectsplotshowingnumberofvisitsperhourby
Anna's hummingbirds per feeder trial in shade and sun microsites
when maximum sun temperature is used as a fixed effect. Error
bars are standard errors (SE).

|
7 of 15
LAWRENCE and HAZLEHURST
of which are native to California, of which Mexican sage (Lamiaceae,
Salvia leucantha) and strawberry tree (Ericaceae, Arbutus unedo)
were the most visited (Table A4).
The best model for the cumulative number of nectaring obser-
vations per session included the interaction of microsite and cate-
gorical temperature (hot vs. average) as the fixed effect and session
ID and floral abundance as random effects (Figure 9, Tables 6 and
A5).Theinteractiontermwassignificant(Coef = 1.15,z-value = 2.87,
p < .01),andpost-hoc tests revealedthat this was driven bydiffer-
ences in microsite use on hot sessions, where hot sessions had more
foragin g in the sun than in t he shade (Rat io = 0.44, z- r a t i o = −2.97,
p < .05),apatternwhichdif fered fromtheirpreferences onnormal
temperature days (Table A6).
3.3 | Pollen deposition
When considering pollen deposition during the feeder trials,
the average number of pollen grains deposited in the shade was
81.29 ± 157.80(SD)and593.02 ± 950.29(SD)inthesun during hot
trials. The average number of pollen grains deposited in the shade
was404.63 ± 747.27(SD)and275.38 ± 747.27inthe sunduringav-
erage days. The average pollen load (total pollen grains deposited/
totalvisits)intheshadewas14.90 ± 26.02(SD)and58.68 ± 109.22
(SD) in the sun during hot trials. The average pollen load in the shade
was 80. 37 ± 170.30 (SD) a nd 20.03 ± 29.25 (SD) in the su n during
average days. The best model for average pollen deposition in-
cluded the interaction of microsite and categorical temperature as
fixed effects, Julian day and observation effect as random effects,
and session duration as an offset term ( Tables 6 and A6). We found
the inter action term was ve ry close to signif icant (Coef = 1.52, z-
value = 1.94,p = .05),butneithermicrositenorcategoricaltempera-
ture was significant individually (Table 7). On hot trials the sunny
microsite received much more pollen than the shady microsite,
though the confidence interval is considerably wider for the sunny
microsite than the shady microsite (Figure 10).
4 | DISCUSSION
The hypothesis that Calypte anna will forage preferentially in shady
microsites on hot days was not supported by the feeder trials or
seminatural behavioral observations. There were some apparent
differences in how hummingbirds spent their time between differ-
ent behaviors on hot sessions. The most frequently observed behav-
ior in the sunny microsites during hot sessions was foraging (4 4%,
Figure 5a), while the most frequently observed behavior in the sunny
microsite during average sessions was aggression (29%; Figure 6b).
TABLE 3 Finalmodelforvisitationrate(visits/hour)infeedertrialswithambienttemperatureasthemaximumtemperatureinthesun
during the session and ambient temperature as a categorical variable where “hot” is categorized as an ambient maximum temperature in the
sun >35˚Cand“average”is≤35˚C.
Variable Coefficient SE Lower 95% CI Upper 95% CI z- value p- value
Model:Visitrate ~ microsite*max_sun_temp + (1|session_id) + offset(session_duration)
Intercept −2.19 0.83 −3.81 −0.06 −2. 64 <. 01**
Microsite (sun) 1.76 0.81 0.17 3.35 0.85 <.05*
Max.suntemp(˚C) 0.02 0.02 −0.02 0.06 0.85 .40
Microsite(sun)*Max.sun
temp
−0.03 0.02 −0.07 0.02 −1. 17 .24
Model:Visitrate ~ microsite*temp_cat + (1|session _id) + offset(session_duration)
Intercept −1.66 0.26 −2.18 −1.15 −6.35 <.01**
Microsite (sun) 0.96 0.26 0.45 1.47 3.68 <.0 01***
Temp category (hot) 0.25 0.32 −0.37 0.87 0.8 .427
Microsite(sun)*Temp
category (hot)
−0.18 0 .31 −0.78 0.43 − 0.57 .57
Note: Variables with significant ef fects are shown in bold.
*p < .05;**p < .01;***p < .001.
FIGURE 6 Categoricalinteractionplotshowingaveragevisit
duration from feeder trials in sun and shade microsites on hot
session and average sessions when temperature is a categorical
variable (average vs. hot). Error bars are st andard errors (SE).

8 of 15
|
LAWRENCE and HAZLEHURST
Perching was the most common behavior in shady microsites dur-
ing both hot and average temperature days (Figures 5 and 6), likely
due to the inherent presence of perching sticks in shade. It may be
that extremely hot ambient temperatures (the highest temperature
recorded in the sun during our study was 48.57°C on July 9, 2021
around 1:37 p.m .) have some effec t on the overal l time budget of
hummingbirds, but that this does not change their foraging microsite
preferences due to the ways in which they detect food sources.
Previous research on microsite occupancy in other avian
taxa has found that birds will preferentially forage in the shade
during hot periods (Abdu et al., 2018; Cunningham et al., 2015;
Lee et al., 2 017; Van de Ven et al., 2019). In many taxa, birds will
forage in the shade above certain operative temperatures even
when there is a fitness cost to doing so (Cunningham et al., 2015;
Van de Ven et al., 2019). In hummingbirds, there is evidence for
context- dependent and contrasting responses to ambient tem-
perature. In territorial hummingbirds like C. anna, increased ther-
moregulatory costs can lead to either decreased foraging activity
to minimize energy loss, or conversely, increased foraging activity
to maximize energy gain (Powers et al., 2017; Shankar et al., 2019).
In our study, it is possible that the increased thermoregulatory
costs of high temperatures could be driving increased foraging,
TABLE 4 Finalmodelforvisitdurationinfeedertrialswithtemperatureasacategoricalvariablewhere“hot”iscategorizedasanambient
maximum temperature in the sun >35˚Cand“average”is≤35˚C.
Model: Visit duration ~ microsite*temp_cat + (1|session_id) + of fset (session_duration)
Variable Coefficient SE Lower 95% CI Upper 95% CI t- value p- value
Intercept −2.34 0.27 −2.83 −1 .8 1 −8.69 <. 01**
Microsite (shade) −0.67 0.3 −1 .2 6 −0.08 −2. 22 <. 01**
Temp category (hot) 0.5 0.34 − 0.16 1.16 1.48 <.05
Microsite(shade)*Temp
category (hot)
0.29 0.38 −0.46 1.03 0.75 .45
Note: Variables with significant ef fects are shown in bold.
*p < .05;**p < .01;***p < .001.
FIGURE 7 Timebudget sofAnna'shummingbirdsduringhotseminaturalbehavioralobservationsessionsbasedonaveragepercentof
observations falling into each behavior category per session in (a) sunny and (b) shady microsites.
FIGURE 8 Timebudget sofAnna'shummingbirdsduringaverageseminaturalbehavioralobservationsessionsbasedonaveragepercent
of observations falling into each behavior category per session in (a) sunny and (b) shady microsites.

|
9 of 15
LAWRENCE and HAZLEHURST
in combination with a preference for sunny microsites. However,
Shankar et al. (2019) found that thermoregulation made up a small
part of energy budgets when compared to activity costs, and
Powers et al. (2017) point out that flight generates large quantities
of extra heat, so this will need further study. The preference for
sunny microsites may be based on the reliance of hummingbirds
on visual cues for detecting both nectar sources and threats from
predators (though cover could also hide them from predators).
There is some evidence that birds may feel less able to leave an ar-
tificially shady microsite because of a shade cloth, leading to more
vigilance and lower preference for these sites (Abdu et al., 2018).
Hummingbird dependence on visual cues to locate flowers is well
established, and flowers in the sun may look different to them
than flowers in the shade due to their ability to see ultraviolet
light (Stoddard et al., 2020).
Hummingbirds do not typically drink water because their water
needs are generally met through the nectar in their diet (Clark &
Russell, 2012; Nicolson & Fleming, 2003), which is usually somewhat
dilute in ornithophilous flowers (Nicolson & Fleming, 2003). However,
cooling through evaporative water loss is an important physiological
thermoregulation mechanism for hummingbirds. At an average tem-
perature of 24°C, over 30% of a hummingbird's required daily water
volume is lost through evaporation, while at 40°C this can be up to
50% (Powers 1992). Gaping behavior allows birds to use evapora-
tive cooling on hot days, and results in additional water loss. It is
possible that hummingbirds need to forage for nectar more on hot
days to keep up with their water needs. This increased demand for
nectar sources on extremely hot days, combined with a preference
for the visual stimuli presented by flowers in the sun, could explain
the patterns we observed. While this study did not specifically quan-
tify the frequency of gaping behavior, we did observe it incidentally
throughout the study during hot trials and sessions, and future stud-
ies in this area should quantify this.
Our results demonstrate that contrary to our hypothesis, flow-
ers in sunny microsites may experience increased pollen deposition
on extremely hot days due to increased frequency and duration
of hummingbird visits to flowers in sunny microsites on hot days.
Similar to the trend observed in the seminatural observations, the
apparent preference for foraging in sunny microsites was amplified
on hot days, with the sunny microsite receiving much more pollen
than the shady site on hot days or either site on average days. Plants
in sunny microsites may actually see increased hummingbird polli-
nation ser vices both during extreme heat events and in the future
under a warming climate regime; however, future studies should
seek to directly quantify pollen deposition, pollen tube formation,
seed set, and seed germination rates as a result of pollination during
heat waves. Pollen deposition was lowest in the shady microsite on
hot days, suggesting that hummingbird- pollinated plants in shady
microsites may be more likely to experience pollen limitation during
extreme heat events. The study of how nectar properties may shift
as a result of warming temperatures is an emerging area of research.
For example, Shrestha et al. (2018) found that above a certain tem-
perature, bees shift their preferences towards flower shapes and
colors that stay cooler, and Russell and McFrederick (2022) found
that higher temperatures can change nectar microbiomes, sugars,
and bee foraging preferences. Pollinator behavior has major implica-
tions for plant conservation, particularly in plants that are specialists
TABLE 5 Finalmodelfromseminaturalobservationsforcumulativenumberofvisits(totalnumberofbirdsobservedforagingpersession
per microsite), with temperature as a categorical variable where “hot” is categorized as an ambient maximum temperature in the sun >35˚C
and“average”is≤35˚C.
Model: Semi- natural observations total visits ~ microsite*temp_cat + (1|session_id) + offset (session_duration)
Variable Coefficient SE Lower 95% CI Upper 95% CI z- value p- value
(Intercept) 1.05 0. 24 0.58 1. 51 4.39 <.001***
Microsite (sun) −0.32 0.29 −0.89 0.24 −1.1 2 .26
Temp category (hot) −0.22 0.37 − 0.94 0.5 −0.6 .55
Microsite(sun)*Temp
category (hot)
1.15 0.4 0.36 1.94 2.87 <.01**
Note: Variables with significant ef fects are shown in bold.
*p < .05;**p < .01;***p < .001.
FIGURE 9 Categoricalinteractionplotshowingthecumulative
number of visits per seminatural behavioral observation session in
sun and shade microsites on hot session and average sessions when
temperature is a categorical variable (average vs. hot). Error bars
are standard errors (SE).

10 of 15
|
LAWRENCE and HAZLEHURST
for hummingbirds and may not receive supplemental visits from in-
vertebrate pollinators or other vertebrate pollinators.
It is thus not surprising that pollen deposition followed similar
trends to visitation rate and visit duration, with sunny microsites on
hot days receiving higher visitation and longer visit durations, and
thus slightly higher pollen deposition. It is interesting that pollen
deposition varied so much in sunny microsites on hot days; this is
likely due to the fact that not all visitors had pollen already on their
bodies when they visited the feeder. Thus, because there is a pref-
erence for sunny microsites on hot days, we see that there is greater
variation in pollen deposition, as more birds are visiting overall.
Morgan et al. (2014) found evidence of individual foraging
preferences in wild rufous hummingbirds (Selasphorus rufus).
Future studies should control for individual hummingbirds by
using color markings to distinguish between individuals if possible,
though not all researchers are able to do this given welfare con-
cerns to birds and variation in local regulations on auxiliary color
markings (Tell et al. 2021). In addition, the sex and age of hum-
mingbirds may affect their foraging decisions, as male, female, and
juvenile hummingbirds in California have distinct diets (Hazlehurst
et al., 2021). Another important factor that could have affected
Plant species Plant family
Num.
Visits
shade
Num.
Visits sun
Tot al
num.
Visits
Proportion
of visits in
sun
Salvia leucantha* Lamiaceae 13 35 48 0.73
Arbutus unedo Ericaceae 24 731 0.23
Gambelia
speciosa*
Plantaginaceae 2 9 11 0.82
Phormium
cookianum
Asphodelaceae 3 5 8 0.63
Erythrina
crista- galli
Fabaceae 0 2 2 1
Lantana camara Verbenaceae 0 1 1 1
Melaleuca
viminalis
Myrtaceae 0 1 1 1
Eucalyptus sp. Myrtaceae 0 1 1 1
Agapanthus
praecox
Amaryliidaceae 1 0 1 0
Note:Anas terisk(*)indicatesthatthespeciesisnativetoCalifornia.
TABLE 6 Ornamentalplantspecies
visited by hummingbirds during
seminatural foraging observations from
N = 20observationsessionsat8locations
on the California State University East Bay
(CSUEB) campus field sites.
TABLE 7 Finalmodelforpollendepositiononthefalsestigmasduringfeedertrialswithtemperatureasacategoricalvariablewhere“hot”
is categorized as an ambient maximum temperature in the sun >35˚Cand“average”is≤35˚C.
Model: pollen count ~ microsite*temp_cat + (1|julian_day) + (1|obs_effect) + offset (session_duration)
Variable Coefficient SE Lower 95% CI Upper 95% CI z- value p- value
Intercept 1.83 0.43 0.99 2.67 4.28 <.0 01***
Microsite (sun) 0. 51 0.54 −0.54 1. 55 0.95 .34
Temp category (hot) −0.78 0.63 −2 .0 1 0.45 −1. 24 .21
Microsite(sun)*Temp
category (hot)
1.52 0.78 −0.02 3.06 1.94 .05
Note: Variables with significant ef fects are shown in bold.
*p < .05,**p < .01,***p < .0 01.
FIGURE 10 Categoricalinteractionplotshowingaverage
number of pollen grains deposited per feeder trial in sun and shade
microsites on hot session and average sessions when temperature
is a categorical variable (average vs. hot). Error bars are standard
errors (SE).

|
11 of 15
LAWRENCE and HAZLEHURST
our results is aggressive territorial behavior, as being chased away
from a feeder could artificially shorten visit duration or cause
hesitancy to return. Previous research suggests that frequency
of aggression declines at both low and high temperatures due to
increased thermoregulatory costs (González- Gómez et al., 2011).
Future studies should also consider foraging behavior over the en-
tire daytime period rather than just focusing on the hottest part
of the day. It is possible that hummingbirds shift as much of their
foraging behavior as possible to early morning to avoid afternoon
heat, and thus there may be additional factors that could influence
pollen deposition patterns on hot and average days. However,
there is mixed evidence for this strategy in hummingbirds, perhaps
due to their need to feed frequently throughout the day, unlike in
many other avian taxa. Powers et al. (2017 ) found only one of five
hummingbird species studied became inactive during hot periods
when it lacked a thermal gradient for passive cooling, and overall
temporal patterns of foraging activity seem to vary by humming-
bird species, time of year, and location (Clark & Russell, 2012). It
is not currently known if wild hummingbirds are facing declines
due to increased temperatures, and more research is needed on
this subject.
There has been extremely limited research in this area to date,
but future studies should further explore how nectar properties and
avian pollinator behavior may change in a warming climate. For ex-
ample, future research could determine if hummingbird behavioral
responses during extreme heat events vary depending on the nec-
tar properties (such as sucrose concentrations) of plants. While this
study was conducted in a seminatural environment that consisted
primarily of cultivated plants, future studies should consider polli-
nation in more natural environments, as patterns of preference and
pollen deposition may dif fer in those habitats. This is especially rel-
evant considering the potential effects of heat stress and increased
evaporation during periods of extremely high temperatures on floral
trait expression (Carroll et al., 2001).
AUTHOR CONTRIBUTIONS
Sabina Lawrence: Formal analysis (lead); investigation (lead); meth-
odology (equal); project administration (equal); validation (equal);
visualization (equal); writing – original draft (equal); writing – review
and editing (equal). Jenny Hazlehurst: Conceptualization (lead);
formal analysis (equal); investigation (supporting); methodology
(supporting); project administration (supporting); resources (lead);
software (equal); supervision (lead); validation (equal); visualization
(equal); writing – original draft (equal); writing – review and editing
(equal).
ACKNOWLEDGEMENT
The authors would like to thank the undergraduate research as-
sistant s who helped collect this data, especially Henry Odufalu.
We also thank all of the members of the Pollination Ecology &
Conser vation lab, and members of the graduate thesis committee
at CSUEB, including Dr. Erica Wildy and Dr. Nazz y Pakpour, whose
early feedback helped shape this research.
FUNDING INFORMATION
None.
CONFLICT OF INTEREST STATEMENT
The authors have no conflic t of interest to declare.
DATA AVAIL ABILI TY STATEMENT
The data that support the findings of this study are openly available
in "Dryad" at https://doi.org/10.5061/dryad.x95x6 9ppd
ORCID
Jenny Hazlehurst https://orcid.org/0000-0002-3520-9846
REFERENCES
Abdu, S., McKechnie, A . E., Lee, A. T. K., & Cunningham, S. J. (2018). Can
providing shade at water points help Kalahari birds beat the heat?
Journal of Arid Environments, 152, 21– 27.
Ackerly, D., Jones, A., Stacey, M., & Riordan, B. (2018). San Francisco Bay
Area summary repor t. California's fourth climate change assessment.
University of California, Berkeley, California, USA.
Anderson, S. H., Kelly, D., Ladley, J. J., Molloy, S., & Terry, J. (2011).
Cascading effec ts of bird functional extinction reduce pollination
and plant density. Science, 331, 1068– 1071.
Barto ń, K. (2022). Mu MIn: Multi-Model Inferenc e. R Package Versio n
1.46.0.
Bates, D., Maechler, M., Bolker, B., & Walker, S. (2015). Fitting linear mixed-
effec t models using lme4. Journal of Statistical Software, 67, 1– 4 8.
Battey, C. J. (2019). Ecological release of the Anna's hummingbird
during a northern range expansion. The American Naturalist, 194 (3),
306– 315.
Bolker, B. M., Brooks, M. E., Clark, C. J., Geagne, S. W., Poulsen, J. R .,
Stevens, M. H. H., & White, J. S. S. (2009). Generalized linear mixed
models: A practical guide for ecology and evolution. Trends in
Ecology and Evolution, 24(3), 127– 135.
Carroll, A. B., Pallardy, S. G., & Galen, C. (20 01). Drought s tress, plant
water status, and floral trait expression in f ireweed, Epilobium
angustifolium (Onagraceae). American Journal of Botany, 88(3),
438– 446.
Clark, C. J., & Russell, S. M. (2012). Anna's hummingbird (Calypte anna).
In A. Poole (Ed.), The birds of North America # 226. Cornell L ab of
Ornithology.
Cronk, Q., & Ojeda, I. (2008). Bird- pollinated flowers in an evolutionary
and molecular context. Journal of Experimental Botany, 59(4), 715–
727. https://doi.org/10.1093/jxb/ern009
Cunningham, S. J., Mar tin, R. O., & Hockey, P. A. R. (2015). Can behaviour
buffer the impacts of climate change on an arid- zone bird? Ostrich,
86(2), 119– 126. https://doi.org/10.2989/00306 525.2015.1016469
Del Coro A rizmendi, M., & Ro dríguez- Flores, C. I . (2012). How many plant
species do hummingbirds visit? Ornitologia Neotropical, 23, 7 1– 75 .
Du Plessis, K. L., Martin, R. O., Hockey, P. A. R., Cunningham, S. J., &
Ridley, A. R . (2012). The cos ts of keeping cool in a warming world:
Implications of high temperatures for foraging, thermoregulation
and body condition of an aridzone bird. Global Change Biology, 18,
3063– 3070. http s://doi. org /10.1111/j.1365 - 2486 .2012.0 2778 .x
Engel, E. C ., & Irwin, R. E. (2003). Linking pollinator visitation rate and
pollen receipt. American Journal of Botany, 90(11), 1612– 1618 .
González- Gómez, P. L., Ricote- Martinez, N., Razeto- Barry, P., Cotorás, I.
S., & Bozinovic, F. (2011). Thermoregulatory cost affects territorial
behavior in hummingbirds: a model and its applic ation. Behavioral
Ecology and Sociobiology, 65, 2141– 2148 .
Greig, E . I., Wood, E. M., & Bonter, D. N. (2017). Winter range ex-
pansion of a hummingbird is associated with urbanization and

12 of 15
|
LAWRENCE and HAZLEHURST
supplementary feeding. Proceedings of the Royal Societ y B: Biological
Sciences, 284(1852), 20170256.
Hazlehurst, J. A., Rankin, D. T., Clark, C. J., McFrederick, Q. S., & Wilson-
Rankin, E. E. (2021). Macroecologic al patterns of resource use in
resident and migratory hummingbirds. Basic and A pplied Ecology, 51,
71– 82.
Kearns, C . A., & Inouye, D. W. (1993). Techniques for pollination biologists.
University Press of Colorado.
Lee, A. T. K., Wright, D., & Barnard, P. (2017). Hot bird drinking patterns:
Drivers of water visitation in a fynbos bird community. African
Journal of Ecology, 55, 5 41– 5 53 .
Lenth, R . V. (2022). emmeans: Estimated marginal means, aka least-
squares means. R Package Version 1.7.4- 1. https://github.com/
rvlen th/emmeans
Luo, L., Apps, D., Arc and, S., Xu, H., Pan, M., & Hoerling, M. (2017).
Contribution of temperature and precipitation anomalies to the
California drought during 2012– 2015. Geophysical Research Letters,
44(7), 318 4– 3192.
Meehan, T. D., Dale, K., Lebaron, G. S., Rowden, J., Michel, N. L.,
Wilsey, C. B., & Langham, G. M. (2020). Interacting with hum-
mingbirds at home: Associations with supplemental feeding,
plant diversity, plant origin, and landscape setting. Landscape
and Urban Planning, 197, 103774. https://doi.org/10.1016/j.landu
rbplan.2020.103774
Morgan , K. V., Hurly, T. A., & Healy, S. D. (2014). Individual differences in
decision making by foraging hummingbirds. Behavioural Processes,
109, 195– 200.
Myers, N., Mittermeier, R. A., Mittermeier, C. G., Da Fonsec a, G. A., &
Kent, J. (20 00). Biodiversity hotspots for conservation priorities.
Nature, 403(6772), 853– 858.
Nicolson, S. W., & Fleming, P. A. (2003). Nectar as food for birds: The
physiological consequences of drinking dilute sugar solutions. Plant
Systematics and Evolution, 238 (1/4), 139– 153. https://www.jstor.
org/stabl e/23644984
Ollerton, J., Winfree, R., & Tarrant , S. (2011). How many flowering
plants are pollinated by animals? Oikos, 120, 321– 326. h t t p s :// do i .
org /10.1111/j.1600- 0706.2010.18644 .x
Pacini, E . N. M. V. J., Nepi, M., & Vesprini, J. L. (2003). Nectar biodiver-
sity: a short review. Plant Systematics and Evolution, 238 (1- 4), 7– 21 .
Powers, D. R . (1992). Effect of temperature and humidity on evapora-
tive water loss in Anna's hummingbird (Calypte anna). Journal of
Comparative Physiology B, 16 2(1), 74– 84.
Powers, D., L angland, K., Wethington, S., Powers, S., Graham, C., &
Tobalske, B. (2017). Hovering in the heat: Effec ts of environmen-
tal temperature on heat regulation in foraging hummingbirds. Royal
Society Open Science, 4(12), 17105 6. ht tps://doi.o rg/10.10 98/
rso s.171056
Ratto, F., Simmons, B. I., Spake, R., Zamora- Gutierrez, V., MacDonald,
M. A., Merriman, J. C., Tremlett, C. J., Poppy, G. M., Peh, K. S. H., &
Dicks, L . V. (2018). Global importance of vertebrate pollinator s for
plant reproductive success: A meta- analysis. Frontiers i n Ecology and
the Environment, 16(2), 82– 90. https://doi.o rg/10 .1002 /fee.1763
Russell, K. A., & McFrederick, Q. S. (2022). Elevated temperature may
affect nectar microbes, nectar sugars, and bumble bee foraging
preference. Microbial Ecology, 84(2), 473– 482 .
Schielzeth, H. (2010). Simple means to improve the interpretability of
regression coefficients. Methods in Ecology and Evolution, 1(2),
10 3– 113 .
Schneider, C. A., Rasband, W. S., & Eliceiri, K. W. (2012). NIH Image to
ImageJ: 25 yearsofimageanalysis.Nature Methods, 9(7), 671– 675.
https://doi.org/10.1038/nmeth.2089
Shankar, A., Graham, C. H., Canepa, J. R., Wethington, S. M., & Powers,
D. R. (2019). Hummingbirds budget energy flexibly in response to
changing resources. Functional Ecology, 33(20 ), 1904– 1916. h t tp s : //
doi .org/10.1111/1365- 24 35.13 404
Shankar, A., Powers, D. R., Davalos, L . M., & Graham, C. H. (2020). The
allomet ry of daily energy expenditure in hummingbirds: An energy
budget approach. Journal of Animal Ecology, 89(5), 1254– 1261.
https://doi.org /10.1111/1365- 2656.1318 5
Shrestha, M., Garcia, J. E., Bukovac, Z., Dorin, A., & Dyer, A. G. (2018).
Pollination in a new climate: Assessing the potential influence of
flower temperature variation on insect pollinator behaviour. PLoS
One, 13(8), e020 0549.
Stiles, F. G. (1971). Time, energy, and territorialit y of the Anna
Hummingbird (Calypte anna). Science, 173(3999), 818– 821.
Stoddard, M. C., Eyster, H. N., Hogan, B. G., Morris, D. H., Soucy, E. R ., &
Inouye, D. W. (2020). Wild hummingbirds discriminate nonspectral
colors. Proceedings of the National Academy of Sciences of the United
States of America, 117 (26), 15112– 15122. ht tps://doi.o rg/10.1073/
pna s.19193 77117
Tell, L. A., Hazlehurs t, J. A., Bandivadekar, R. R., & Brown, J. C. (2021).
Hummingbird (Family Trochilidae) Research: Welfare- conscious Study
Techniques for Live Hummingbirds and Processing of Hummingbird
Specimens. Museum of Texas Tech University.
Van de Ven, T. M. F. N., McKechnie, A. E., & Cunningham, S. J. (2019). The
costs of keeping cool: Behavioural trade- offs between foraging and
thermoregulation are associated with significant mass losses in an
arid zone bird. Oecologia, 191, 205– 215. https://doi.org/10.1007/
s0 04 4 2- 019- 04 48 6 - x
How to cite this article: Lawrence, S. L., & Hazlehurst, J.
(2023). Hummingbird foraging preferences during extreme
heat events. Ecology and Evolution, 13, e10053. h t t p s :// do i .
org /10.1002/ece3.10053

|
13 of 15
LAWRENCE and HAZLEHURST
APPENDIX A
TABLE A1 Toprankedmodelsfromfeedertrialsforthecumulativenumberofvisitsbyhummingbirds.
Rank Intercept
Fixed effects
Fixed effects
Random
effects Offset AICc ΔAICcMicrosite (sun)
Max sun
temp
Temp
category
(hot)
Microsite
(sun) * max
sun temp
Microsite
(sun) * temp
category (hot)
1−1. 20
SE = 0.63
p = .06
1.50
SE = 0.53
p < .01
0.02
SE = 0.02
p = .37
na −0.02
SE = 0.01
p = .18
na Microsite * max sun temp Session id Session
duration
373.5 0
2−1. 28
SE = 0.62
p < .05
1.45
SE = 0.53
p < .01
0.02
SE = 0.02
p = .27
na −0.02
SE = 0.01
p = .18
na Microsite*maximumsun
temperature
Session id,
human
presence
Session
duration
374.4 1.43
3−0.73
SE = 0.20
p < .001
0.82
SE = 0.17
p < .001
na 0.14
SE = 0.24
p = .57
na −0.02
SE = 0.20
p = .91
Microsite*temperature
category
Session id Session
duration
374.9 1.44
4−0.75
SE = 0. 24
p < .01
0.82
SE = 0.17
p < .001
na 0.24
SE = 0.24
p = .33
na −0.02
SE = 0.20
p = .91
Microsite*temperature
category
Session id,
human
presence
Session
duration
375.1 2.13
Note: If “na” is present, that means a fixed effect was not present in that model. The model in bold was selected and is presented in the results. Only models that successfully converged and were not
overdispersed are shown.
*p < .05.
TABLE A2 Toprankedmodelsfromfeedertrialsforthevisitationrate(visits/hour)byhummingbirds.
Rank Intercept
Fixed effects
Fixed effects
Random
effects Offset AICc ΔAICcMicrosite (sun) Ma x sun temp
Temp category
(hot)
Microsite (sun)
* max sun
temp
Microsite
(sun) * temp
category (hot)
1−2.19
SE = 0.83
p < .01
1.76
SE = 0.81
p < .05
0.02
SE = 0.02
p = .40
na −0.02
SE = 0.02
p = .24
na Microsite * Max
sun temp
Session id Session
duration
275.5 0
2−1. 66
SE = 0.26
p < .001
0.96
SE = 0.26
p < .001
na 0.25
SE = 0.32
p = .43
na −0.18
SE = 0.31
p = .57
Microsite*Temp
category
Session id Se ssion
duration
276.2 0.74
3−2. 26
SE = 0.82
p < .01
1.74
SE = 0.80
p = .03
0.02
SE = 0.02
p = .31
na −0.02
SE = 0.02
p = .25
na Microsite*Max
sun temp
Session id,
Human
presence
Session
duration
276.5 1.55
4−1. 69
SE = 0.30
p < .001
0.96
SE = 0.26
p < .001
na 0.36
SE = 0.31
p = .25
na −0.18
SE = 0.31
p = .57
Microsite*Temp
category
Session id,
Human
presence
Session
duration
276.6 1.59
Note: If “na” is present, that means a fixed effect was not present in that model. The model in bold was selected and is presented in the results. Only models that successfully converged and were not
overdispersed are shown.
*p < .05.

14 of 15
|
LAWRENCE and HAZLEHURST
TABLE A3 Toprankedmodelsfromfeedertrialsforthevisitdurationbyhummingbirds.
Rank Intercept
Fixed effects
Fixed effects
Random
effects Offset AICc ΔAICcMicrosite (sun) Max sun temp
Temp
category
(hot)
Microsite (sun)
* max sun temp
Microsite (shade)
* temp category
(hot)
1−2. 34
SE = 0.27
p < .01
−0. 67
SE = 0.30
p < .01
na 0.50
SE = 0.34
p = .02
na 0.29
SE = 0.38
p = 0.45
Microsite * Temp
category
Julian day Session
duration
152.7 0
2−5.01
SE = 0.80
1.28
SE = 0.94
p < .01
0.07
SE = 0.02
p < .001
na −0.02
SE = 0.03
p = .39
na Microsite*Maxsuntemp Julian day Session
duration
158.8 6.18
Note: If “na” is present, that means a fixed effect was not present in that model. The model in bold was selected and is presented in the results. Only models that successfully converged and were not
overdispersed areshown.
*p < .05.
TABLE A4 Toprankedmodelsfromseminaturalobservationsforthecumulativenumberofvisitsbyhummingbirds.
Rank Intercept
Fixed effects
Fixed effects Random effects Offset AICc ΔAICcMicrosite (sun) Temp category (hot)
Microsite (sun) *
temp category (hot)
11.05
SE = 0.24
p < .001
−0.32
SE = 0.29
p = .26
−0.22
SE = 0.37
p = .55
1.15
SE = 0.40
p < .01
Microsite * Temp
category
Session id, Floral abundance na 165.1 0
21.13
SE = 0.29
p < .001
−0.32
SE = 0.29
p = .27
−0.35
SE = 0.44
p = .43
1.14
SE = 0.40
p < .01
Microsite*Temp
category
Session id, Floral abundance,
Julian day
na 166.7 2 .75
Note: If “na” is present, that means a fixed effect was not present in that model. The model in bold was selected and is presented in the results. Only models that successfully converged and were not
overdispersed are shown.
*p < .05.

|
15 of 15
LAWRENCE and HAZLEHURST
Contrast Estimate SE t ratio p value
Shade average/sun
average
−0. 67 0.3 −2. 22 .21
Shade average/shade hot −0.79 0.34 −2.33 .15
Shade average/sun hot −1. 17 0.34 −3.47 <.01**
Sun average/shade hot −0.17 0.34 −0.34 1
Sun average/sun hot −0.5 0.34 −1.4 8 .87
Shade hot/sun hot −0.39 0.23 −1.6 8 .63
Note: Significant ef fects are shown in bold.
Note:*p < .05,**p < .01,***p < .001.
TABLE A5 Post-hocpairwise
comparisons for significant interaction
term from best model for seminatural
observations cumulative visits by
hummingbirds.
TABLE A6 Toprankedmodelsfrompollendepositionduringfeedertrials.
Rank Intercept
Fixed effects
Fixed effects Random effects Offset AICc ΔAICc
Microsite
(sun)
Temp
category (hot)
Microsite
(sun) * temp
category (hot)
11.83
SE = 0.43
p < .001
0. 51
SE = 0.54
p = .34
−0.78
SE = 0.63
p = .21
1.52
SE = 0.78
p = .05
Microsite * Temp
category
Julian day,
Observation
effect
Session
duration
766 0
11.83
SE = 0.43
p < .001
0. 51
SE = 0.54
p = .34
−0.78
SE = 0.63
p = .21
1.52
SE = 0.78
p = .05
Microsite * Temp
category
Session id,
Observation
effect
Session
duration
766 0
24.12
SE = 0.43
p < .001
0.50
SE = 0.53
p = .35
−0.80
SE = 0.63
p = .21
1.52
SE = 0.78
p = .05
Microsite*Temp
category
Julian day,
Observation
effect
na 766 .4 0.4
24.11
SE = 0.43
p < .001
0.50
SE = 0.53
p = .35
−0.80
SE = 0.63
p = .21
1.52
SE = 0.78
p = .05
Microsite*Temp
category
Session id,
Observation
effect
na 766 .4 0.4
31.83
SE = 0.43
p < .001
0. 51
SE = 0.60
p = .40
−0.78
SE = 0.63
p = .21
1.52
SE = 0.88
p = .09
Microsite*Temp
category
Floral
abundance,
Observation
effect
Session
duration
767. 3 1.37
Note: If “na” is present, that means a fixed effect was not present in that model. The model in bold was selected and is presented in the results. Only
models that successfully converged and were not overdispersed are shown.
*p < .05.