Thermally moderated firefly activity is delayed by precipitation extremes

Abstract

The timing of events in the life history of temperate insects is most typically primarily cued by one of two drivers: photoperiod or temperature accumulation over the growing season. However, an insect’s phenology can also be moderated by other drivers like rainfall or the phenology of its host plants. When multiple drivers of phenology interact, there is greater potential for phenological asynchronies to arise between an organism and those with which it interacts. We examined the phenological patterns of a highly seasonal group of fireflies (Photinus spp., predominantly P. pyralis) over a 12-year period (2004-2015) across 10 plant communities to determine whether interacting drivers could explain the variability observed in the adult flight activity density (i.e. mating season) of this species. We found that temperature accumulation was the primary driver of phenology, with activity peaks usually occurring at a temperature accumulation of approximately 800 degree days (base 10°C); however, our model found this peak varied by nearly 180 degree-day units among years. This variation could be explained by a quadratic relationship with the accumulation of precipitation in the growing season; in years with either high or low precipitation extremes at our study site, flight activity was delayed. More fireflies were captured in general in herbaceous plant communities with minimal soil disturbance (alfalfa and no-till field crop rotations), but only weak interactions occurred between within-season responses to climatic variables and plant community. The interaction we observed between temperature and precipitation accumulation suggests that, although climate warming has the potential to disrupt phenology of many organisms, changes to regional precipitation patterns can magnify these disruptions.

Full text

rsos.royalsocietypublishing.org
Research
Cite this article: HermannSL,XueS,RoweL,
Davidson-Lowe E, Myers A, Eshchanov B,
Bahlai CA. 2016 Thermally moderated rey
activity is delayed by precipitation extremes.
R. Soc. open sci. 3: 160712.
http://dx.doi.org/10.1098/rsos.160712
Received: 19 September 2016
Accepted: 17 November 2016
Subject Category:
Biology (whole organism)
Subject Areas:
ecology/environmental science
Keywords:
lightning bug, Lampyridae, phenology,
ecoinformatics, long-term ecological research
Author for correspondence:
Christie A. Bahlai
e-mail: cbahlai@msu.edu
Thermally moderated rey
activity is delayed by
precipitation extremes
Sara L. Hermann1,2,5,SaisiXue
3,6,LoganRowe
1,
Elizabeth Davidson-Lowe1,2, Andrew Myers1,5,
Bahodir Eshchanov1and Christie A. Bahlai4,7
1Department of Entomology, 2Ecology, Evolutionary Biology and Behavior Program,
3Biomass Conversion Research Laboratory, Depar tment of ChemicalEngineering, and
4Department of Integrative Biology, Michigan State University, East Lansing, MI
48824, USA
5Department of Entomology, Pennsylvania State University, State College,
PA 16803, USA
6DOE Great Lakes Bioenergy Research Center, East Lansing, MI 48824, USA
7Mozilla Science Laboratory, Mozilla, Mountain View, CA 94041, USA
CAB, 0000-0002-8937-8709
The timing of events in the life history of temperate insects
is most typically primarily cued by one of two drivers:
photoperiod or temperature accumulation over the growing
season. However, an insect’s phenology can also be moderated
by other drivers like rainfall or the phenology of its host plants.
When multiple drivers of phenology interact, there is greater
potential for phenological asynchronies to arise between an
organism and those with which it interacts. We examined the
phenological patterns of a highly seasonal group of fireflies
(Photinus spp., predominantly P. pyralis) over a 12-year period
(2004–2015) across 10 plant communities to determine whether
interacting drivers could explain the variability observed in the
adult flight activity density (i.e. mating season) of this species.
We found that temperature accumulation was the primary
driver of phenology, with activity peaks usually occurring
at a temperature accumulation of approximately 800 degree
days (base 10°C); however, our model found this peak varied
by nearly 180 degree-day units among years. This variation
could be explained by a quadratic relationship with the
accumulation of precipitation in the growing season; in years
with either high or low precipitation extremes at our study
site, flight activity was delayed. More fireflies were captured
in general in herbaceous plant communities with minimal soil
disturbance (alfalfa and no-till field crop rotations), but only
weak interactions occurred between within-season responses
to climatic variables and plant community. The interaction we
observed between temperature and precipitation accumulation
2016 The Authors. Published by the Royal Society under the terms of the Creative Commons
Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted
use, provided the original author and source are credited.

2
rsos.royalsocietypublishing.org R. Soc. open sci. 3: 160712
................................................
suggests that, although climate warming has the potential to disrupt phenology of many organisms,
changes to regional precipitation patterns can magnify these disruptions.
1. Introduction
Much can be learned about biological systems by observation alone [1], and observational data are often
captured incidentally as a result of human activity [2]. Incidental data can range from the very informal
and uncontrolled (e.g. comments on a topic in a Web forum) to highly controlled and meticulously
collected (e.g. unused data from scientific experiments). Indeed, research activities can produce systemic
observational data of very high quality; for instance, insect trapping systems seldom only capture
target taxa. This ‘by-catch’ can provide data that support investigations into entirely uninvestigated
phenomena. In this study, we examine one such ‘by-catch’ dataset: a 12-year time series of firefly
observations in southwestern Michigan, for their responses to environmental and habitat conditions.
Over 2000 species of firefly (Coleoptera: Lampyridae) have been identified across various temperate
and tropical environments around the world [3]. As larvae, species within the family Lampyridae
spend much of their time living underground feeding on earthworms, molluscs and other subterranean
invertebrates [4]. As adults, most species abstain from feeding [5], with the exception of the species
Photuris pennsylvanica, of which the female is a voracious predator of both conspecifics as well as
other insects [5–7]. Few studies have been conducted on firefly conservation and broader-scale ecology
in relation to changing environments and land uses, and little is known about how environmental
parameters drive firefly life history. It has been demonstrated that the life history of at least one species
of firefly is temperature-dependent; researchers found that P. pennsylvanica adult emergence could be
artificially accelerated by exposing larvae to increased soil temperature [8]. However, much of the
primary research on fireflies has focused on the bioluminescent properties of the firefly [9–14], while
research describing basic population and community ecology of this important family is lacking.
In addition to the scientific importance of the Lampyridae for their bioluminescent properties and as
model organisms for evolutionary investigations, fireflies are also among the most widely recognized
and culturally valued insect families among non-scientists. Two US states have designated the firefly as
their ‘State Insect’ [15]. Notably, fireflies also feature prominently in Japanese culture, where they have
been designated as national natural treasures in many districts and have been used to generate support
for biodiversity conservation efforts in Japanese agricultural regions [16–18]. They have also been touted
as useful classroom tools for sparking student interest in biology [19]. Because of their popular appeal,
it is unsurprising that public concern has grown about apparent declines of firefly populations from
regions around the world where they occur [20].
Considering the paucity of ecological information about fireflies, their widespread popularity, ease
with which adults are observed and concerns about their population viability, fireflies represent an
ideal species for citizen science investigations. Citizen science efforts are currently underway seeking
to gain information about the status, geographical distribution and phenology of fireflies [21–23],
and peer-reviewed publications on fireflies have already been produced based on these volunteer-
generated data [24,25]. The popularity of fireflies gives them great potential as a flagship and umbrella
conservation species and potentially an indicator species of ecological degradation in agricultural regions
[26]. However, to our knowledge, no long-term systematic study of firefly phenology and responses to
environmental drivers has been published.
Phenology plays a significant role in regulating species abundance, distribution and biodiversity
[27,28]. The timing of phenological events in insect life histories is strongly linked to climatic conditions
[29–31], such as temperature and precipitation [27,32,33]. Changes in phenology can have community-
wide consequences, and differential responses among various species within a community can lead
to trophic mismatches [28,30]. For example, the timing of larval winter moth (Operophtera brumata)
emergence was formerly largely synchronized with oak (Quercus robur) bud burst. Caterpillars that
emerge too early lack a sufficient food source and will starve, while caterpillars that emerge too late will
be exposed to older, poor-quality leaves, leading to negative physiological implications [34]. Increased
spring temperature has resulted in changes in the timing of oak bud burst. However, the winter moth
has yet to adapt to changing temperatures, which has led to disrupted synchrony between these two
species [34]. Thus, phenological shifts can have both top-down and bottom-up consequences extending
throughout multiple trophic levels. Long-term observations are important for understanding ecological
trends and the merit of phenology as a predictor of ecological consequences. A long-term study on the

3
rsos.royalsocietypublishing.org R. Soc. open sci. 3: 160712
................................................
Genji firefly (Luciola cruciate) in Japan found that population patterns changed in response to rainfall,
potentially leading to early larval emergence and reduced foraging [35]. However, the ways in which
climate change and other environmental events have impacted firefly species is less understood.
Developing a model for the emergence of adult fireflies is key to developing our understanding of
firefly phenology, which can then be used to expand firefly conservation efforts, educational outreach
and environmental research and to predict peak firefly display. In this study, we examine a ‘by-catch’
dataset documenting captures of fireflies at the Kellogg Biological Station over a 12-year period and
place it in the context of other available data to gain insights into the long-term dynamics and phenology
of this charismatic, but understudied taxon.
2. Material and methods
2.1. Data sources
Data were obtained through two publicly available datasets—a weather dataset that included daily
maximum and minimum temperature and precipitation, as well as a dataset that focuses on ladybeetle
observations, but also documents captures of the other insect species. Both datasets arise from Michigan
State University’s Kellogg Biological Stations (KBS), located in southwestern Michigan. The firefly
abundance data were collected as a part of the KBS Long Term Ecological Research (LTER) Site within the
Main Cropping System Experiment (MCSE) and forest sites starting in 2004. Fireflies were recorded to
family alone; however, from spot-checks of the collected data, it appears that fireflies collected belonged
to the genus Photinus,mainlyPhotinus pyralis, the big dipper firefly, although captures of other species
cannot be excluded.
Within the MCSE, seven plant community treatments were established in 1989, ranging from a three-
year rotation of annual field crops (maize, soyabean and wheat) under four levels of management
intensity (conventional, no-till, reduced input or biologically based), to perennial crops including alfalfa,
poplar and early successional vegetation (i.e. abandoned agricultural fields maintained in an early
successional state by yearly burnings; table 1 and figure 1). Each of these treatments is replicated six
times across the MCSE site with each replicate consisting of a 1 ha plot. We also included three forest sites
in our analysis; these sites were established in 1993 within 3 km of the MCSE site on KBS and represent
one of three plant community treatments: conifer forest plantations, late-successional deciduous forest
and successional forest arising on abandoned agricultural land (table 1 and figure 1). Forested treatment
plots are also 1 ha in size but are replicated three times for each treatment.
Observations were taken on a weekly basis throughout the sampling season at five sampling stations
within each replicate (both MCSE and forest sites). These insect abundance data are available publicly,
online at http://lter.kbs.msu.edu/datatables/67. Insect abundance monitoring was done using unbaited
two-sided yellow cardboard sticky cards (Pherocon, Zoecon, Palo Alto, CA, USA) suspended from a
metal post within each sampling station, 1.2 m above the ground. Cards were deployed each week for
a one-week exposure for the duration of the growing season. Sampling start and end dates varied each
year depending upon planting date of the various crops; however, the length of sampling was fairly
consistent (14 ±1 weeks, on average, per year).
In addition to plant community treatment management information, we also included weather as
an environmental factor to explain firefly abundance. These data were also obtained through a publicly
available dataset, online at http://lter.kbs.msu.edu/datatables/7.
2.2. Data preprocessing
All analyses performed are available as an R script at https://github.com/cbahlai/lampyrid/. Analyses
presented in this manuscript were run in R v. 3.3.1 ‘Bug in Your Hair’ (R Development Core Team 2016).
Firefly data were extracted from the database held at the KBS data archive and combined with relevant
agronomic data (which are encoded in plot and treatment numbers in the main database) and are hosted
at figshare at https://ndownloader.figshare.com/files/3686040.
Data were subjected to quality control manipulations to remove misspellings in variable names
that had occurred with data entry. Observations with missing values for firefly counts were excluded
from analysis. Because subsample data were zero-biased, we used reshape2 [36] to sum within date,
within plot observations, and created an additional variable to account for sampling effort (which
was usually consistent at five traps per plot per sampling period, but on occasion traps were lost or
damaged).

4
rsos.royalsocietypublishing.org R. Soc. open sci. 3: 160712
................................................
SF-2
SF-3
DF-2
DF-3
CF-3
CF-2
0 0.50 km0.25
SF = mid-successional
CF = coniferous forest
DF = deciduous forest
LTER main site
Kellogg Biological Station LTER
Main Cropping System Experiment
successional and forested sites
airphoto date: 12 August 2011
SF-1 DF-1
CF-1
T8
Figure 1. Map of sites at the Kellogg Biological Station LTER (reproduced and modied from http://lter.kbs.msu.edu/maps/images/
current-lter-forest-successional-sites.pdf). Red outlined areas indicate successional and forested sites, denoted by the key atthe bottom
right of the photo.O utlined in yellowis the MCSE (LTER main site) which houses seven treatments:conventional crop, no-till crop, reduced
input crop, biologically based crop, poplar, alfalfa and early successional. Each replicate consists of a randomized block of 1 ha plots for
each of the seven treatments, and the experiment is replicated six times in the MCSE.
Tabl e 1. Description of the plant community treatments at the Kellogg Biological Station Long Term Echnological Research Site, in which
insect sampling occurred.
crop type plant community treatments description
annual conventional rotated crop eld (corn–soyabean–wheat), with conventional chemical input
which is chisel ploughed
.............................................................................................................................................................................................
no-till rotated crop eld (corn–soyabean–wheat), with conventional chemical input,
with no tilling
.............................................................................................................................................................................................
reduced input biologically based rotated crop eld (corn–soyabean–wheat), with low-input
chemical control and a winter cover crop (leguminous). Plots are treated
with banded herbicide and starter nitrogen at planting
.............................................................................................................................................................................................
organic biologically based rotated crop eld (corn–soyabean–wheat), low input
chemical control with winter cover crop (leguminous). Certied organic
.........................................................................................................................................................................................................................
perennial poplar trees 10-year rotation cycle of a fast-growing Populus clone
.............................................................................................................................................................................................
alfalfa continuously grown alfalfa
.............................................................................................................................................................................................
early successional abandoned eld from 1989, left to grow into native successional plants which
are annually burned
.........................................................................................................................................................................................................................
forest successional 40–60-year-old successional forest, left from former agricultural elds
.............................................................................................................................................................................................
coniferous 40–60-year-old conifer plantations
.............................................................................................................................................................................................
deciduous late successional deciduous forest
.........................................................................................................................................................................................................................

5
rsos.royalsocietypublishing.org R. Soc. open sci. 3: 160712
................................................
Weather data (daily maximum and minimum temperatures were reported in degrees Celsius and
daily precipitation in millimetres) were downloaded directly from the Kellogg Biological Station
Data archive (http://lter.kbs.msu.edu/datatables/7.csv). To overcome errors in calculations requiring
accumulated annual weather data caused by rare missing data points (most often occurring during
winter, in periods of extreme cold leading to equipment malfunction), we created a function to replace
missing values in the temperature data with the value that was observed for that variable from the day
before the missing observation.
We created a dummy variable for ‘start day’ to enable the user to test the sensitivity of our
conclusions to varying our within-year start of accumulation of environmental conditions. We
empirically determined (by the Akaike information criterion (AIC)) that 1 March (start =60) provided the
best compromise between capturing early growing season weather variation and negating brief variation
in winter conditions; however, the selection of the precise day did not dramatically influence the overall
trends in the results unless changed by more than 15 days.
We then created a function to calculate daily degree-day accumulation and season-long degree-day
accumulation based on Allen’s [37] double sine function, using our daily maximum and minimum
temperature data. We created a dummy variable for our minimum development threshold to facilitate
sensitivity analysis, but set it to a default value of 10°C. We did not use a maximum development
threshold in our calculation, assuming that temperatures exceeding its hypothetical value (often more
than 30°C for temperate insects) were relatively rare. Accumulations were calculated from the start day
variable, as described above. We also created functions that calculated the accumulation of precipitation
over the sampling week, the accumulation of precipitation over the growing season, from the start date
and the number of rainy days in a sampling period. Weather data were merged with firefly data to
facilitate subsequent analyses.
2.3. Data analysis
We used ggplot2 [38] to visualize trends in captures of fireflies by plant community treatment over
years. We then conducted a multivariate analysis to determine whether firefly plant community use
patterns changed within or between years, and what environmental factors were associated with
plant community use patterns. To accomplish this, data were cast as a date-by-treatment matrix
at two resolutions (weekly observations and yearly observations), transformed using the Wisconsin
standardization, and Bray–Curtis differences were subjected to non-metric multi-dimensional scaling
(NMDS) in vegan [39]. Environmental parameters were fit to the NMDS plots using envfit to determine
whether patterns were influenced by weather.
To examine patterns in firefly captures over time, and the interactions of these captures with
environmental variables, we visualized trends in capture data by sampling week and degree-day
accumulation. Noting that degree-day accumulation was associated with the clearest patterns in firefly
captures (see Results), with some variation owing to plant community, we built a generalized linear
model (GLM) with a negative binomial structure to explain these patterns. The model included
the degree-day accumulation in linear and quadratic forms as continuous variables, year and plant
community treatment as factors, and trapping effort as an offset variable (to account for lost or
compromised traps). Model structure was determined empirically by AIC. After fitting the model, we
used the resultant regression parameters to generate predicted values, so we could visually compare the
performance of the model to the raw data.
Because the model found year-to-year variation in the activity peak that was not explained by
degree-day accumulation, we extracted the activity peaks from each year as predicted by the model
to a new data frame, and matched these data to other relevant environmental variables in the weather
matrix (week the peak occurred in, precipitation variables corresponding to that week). We visualized
the relationship between activity peak and other variables, and then constructed a generalized linear
model for a quadratic relationship between the activity peak, by degree days, and the precipitation
accumulation at the activity peak.
For all frequentist analyses, a significance level of α=0.05 was used.
3. Results
Over the 12-year study, 17 084 fireflies were captured in the trapping network. Visualizations of firefly
capture data by treatment and time period revealed several patterns. Numbers of fireflies captured in
each trap varied by plant community type and across samples (figure 2), but in general, more fireflies

6
rsos.royalsocietypublishing.org R. Soc. open sci. 3: 160712
................................................
0
1
2
3
alfalfa
coniferous
conventional
deciduous
early successional
no till
organic
poplar trees
reduced input
successional
treatment
adults per trap
Figure 2. Box plot of average rey captures, 2014–2015, by plant community treatment. Yearly average number of adult reies
capturedonweeklysampledyellowsticky cards across10 plant community treatments at Kellogg Biological Station. Median rey density
in each treatment is represented by the bold line, and upper and lower margins of each box represent the upper and lower quartiles in
that treatment, respectively.
0
1
2
3
2004 2007 2010 2013
year
adults per trap
treatment
alfalfa
coniferous
conventional
deciduous
early successional
no till
organic
poplar trees
reduced input
successional
Figure 3. Average rey captures, 2004–2015, by plant community treatment, by year. Yearly average number of adult reies captured
on weekly sampled yellow sticky cardsacross 10 plant community treatments at Kellogg Biological Station. Loess smoother lines represent
smoothed captures within a given treatment and are used to illustrate general trends in the population across treatments.
were captured in alfalfa and no-till row crop treatments. Average numbers of fireflies captured per
trap also demonstrated variation by year independent of plant community treatment. Overlaid plots
of average captures for all treatments against year (figure 3) suggest a 6–7-year firefly population cycle
that appears uncorrelated with environmental variables.
NMDS revealed only weak trends in patterns of capture between plant community treatments at both
the yearly and weekly resolutions. At the yearly resolution (figure 4a), plant community treatment use
varied slightly with the number of rainy days in the growing season (R2=0.16, p=0.006, 2D NMDS
stress =0.14) with herbaceous habitat use generally associated with greater amounts of rainfall. At the
weekly resolution (figure 4b), 2D NMDS stress was higher (0.19), but a general trend away from forest
plots was observed with increasing degree-day accumulation (R2=0.15, p=0.001) and week (R2=0.15,
p=0.001).
When plotting firefly abundance by week of capture, the timings of the peaks in firefly emergence
show asynchrony among years (figure 5a), indicating that week of year (and, by proxy, day length)

7
rsos.royalsocietypublishing.org R. Soc. open sci. 3: 160712
................................................
−0.8 −0.6 −0.4 −0.2 0 0.2
−0.3
−0.1
0
0.1
0.2
0.3
NMDS2
rain. days
alfalfa
coniferous
conventional
deciduous
early successional no till
organic
poplar trees
reduced input
successional
year
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
−2 −1 0 1 2
−1.0
−0.5
0
0.5
1.0
NMDS1
NMDS2
week
ddacc
alfalfa
coniferous
conventi onal
deciduous
early successional
no till
organic
poplar trees
reduced input
successional
(a)
(b)
Figure 4. Two-dimensional NMDS and environmental tting of plant community treatment plot use by reies over time. (a)Atthe
yearly resolution, a 2D NMDS stress of 0.14 was observed. (b) At the weekly resolution, a 2D NMDS stress of 0.19 was observed.
is not a strong driver of firefly emergence. However, plotting firefly numbers instead against degree-
day accumulation dramatically reduced the asynchrony of emergence peaks and indicated that a single
activity peak occurred in each year (figure 5b). Thus, degree-day accumulation appears to be a better
predictor of firefly populations than week of year or associated variables.
Our model for firefly activity incorporating degree-day accumulation, plant community treatment
and year performed well at predicting the timing of the activity peaks (figure 6), accounting for more
than 40% of the variation in the raw data. However, model selection favoured the inclusion of a year
term as a factor, suggesting that another factor in addition to degree-day accumulation was varying from
year to year and impacting firefly activity. Activity peaks varied from year to year by nearly 180 degree-
day units, varying from 720 ±38 DD in 2004 to 898 ±55 DD in 2012 (figure 7). However, we found the
year-to-year variation was well explained by precipitation accumulation: a quadratic relationship occurs
between degree days at peak emergence and precipitation accumulation (pseudo-R2=0.456, p=0.026;
figure 8).
4. Discussion
The greatest proportion of fireflies was captured in alfalfa and no-till plant communities (figure 2),
indicating that areas with moderate soil disturbance and primarily herbaceous plant communities
favoured firefly emergence. This result was unexpected; because fireflies spend much of their life
cycle in the soil, it might be expected that plots with little soil disturbance (coniferous, deciduous and

8
rsos.royalsocietypublishing.org R. Soc. open sci. 3: 160712
................................................
0
1
2
3
4
20 25 30 35
week
adults per trap
0
1
2
3
4
500 1000 1500
degree-day accumulation
year
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
(a)(b)
Figure 5. Average number of adult reies per trap across all sampled treatments at Kellogg Biological Station plotted by year. Samples
were taken weekly over the growing season from 2004 to 2015, and plotted by (a) week of capture and (b) degree-day accumulation
at capture. Loess lines represent smoothed capture trends for a given year and were used to assess consistency of response to a given
variable between years.
0
10
20
30
40
50
0
10
20
30
40
50
predicted
observed
0 2000 4000 6000
observation number
no. adults captured
Figure 6. Number of rey adults captured, as predicted by GLM and, as observed, by observation number. Predicted values were
generated using GLM accounting for variability owing to plant community treatment degree-day accumulation, and year as a factor
variable. Details of GLM can be found in the data analysis section in Material and methods.
successional forests) would foster the greatest populations of fireflies. However, these plots produced
capture rates similar to those observed in the intensively managed and tilled conventional row crop
plots. Our result contrasted with observations of another genus of fireflies in Malaysia (Pteroptyx), where
researchers found that plant canopy structure was the most important determinant of abundance [40].
Also surprising was the relatively low capture rate in early successional plots, which are primarily
herbaceous, with a no-till management regime. Thus, the yearly burnings may play a role in suppressing
firefly populations in these plots. An alternative explanation for these variations in captures could be

9
rsos.royalsocietypublishing.org R. Soc. open sci. 3: 160712
................................................
0
250
500
750
1000
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
year
DD at peak emergence
Figure 7. Degree-day accumulation at peak rey activity by year. Degree-day accumulation (±s.e.m.) at peak emergence of rey
adults varied by sample year. Activity peaks were extracted from regression coecients from GLM.
700
800
900
300 400 500 600
precipitation accumulation (mm)
DD at peak emergence
year
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
Figure 8. Firey activity peaks by precipitation accumulation. Firey activity per degree-day accumulation had a quadratic relationship
with precipitation accumulation (pseudo-R2=0.456, p=0.026).
differences in trapping efficiencies between plant communities. However, if this were the case, we would
expect trapping efficiencies in the three other row crop treatments (conventional, organic and reduced
input management) not to differ appreciably from that of the no-till row crop plant community.
When plotted over sample years (figure 3), captures of fireflies by treatment seem to suggest an
intriguing cyclical dynamic, with alternating peaks and troughs in captures on an approximately 6-year
cycle. Our time series only spans 12 years, meaning more data will be required to elucidate this pattern
and its drivers. Similarly, analysis of plant community use patterns was inconclusive (figure 4). At the
weekly resolution there was a trend away from woody treatments over the growing season (figure 4b;
i.e. with both increasing week and increasing degree-day accumulation). Although this pattern was not
strong, it could result from fireflies overwintering in forest habitats and then moving to lower-canopy
herbaceous habitats for mating displays. We observed very similar performance of both degree day
and week, probably owing to autocorrelation between the two variables that cannot be resolved at the
sampling resolution used over the course of the study.
The degree-day model GLM suggested that activity peaks occurred at a degree-day accumulation of
approximately 800 DD, accumulated from March 1. A model for Photinus carolinus in the Great Smokey
mountains found peak display occurred at approximately 1100 degree days (using a base of 10°C and the
same start date as our model) [25]. The difference between heat units required for peak activity observed
between this and our study may be a result of species or locality differences (i.e. more southern firefly

10
rsos.royalsocietypublishing.org R. Soc. open sci. 3: 160712
................................................
populations are probably adapted to warmer spring conditions). Similarly, differences in methodology
for calculating degree days may account for some of these differences. However, both studies support
the observation that degree-day accumulation is the dominant cue governing the activity patterns of
temperate fireflies.
Although both photoperiod and degree-day accumulation can both play a role in the phenology
of insects, our results suggest that degree-day accumulation is the dominant driver of firefly flight
activity. The model was unable to account for between-trap variation within a single sampling
day (figure 6), though it was able to capture the overall trends in activity quite well, using only
degree-day accumulation, plant community treatment and year as predictors. Nevertheless, degree-
day accumulation was not the sole driver in within-season variability. Our model found year-to-year
variability in activity peaks that could not be explained by degree-day accumulation alone. We found
that this variation in activity peak by degree-day accumulation had a quadratic relationship with
precipitation, indicating that both drought and heavy rainfall in the time period leading up to their
activity peak can delay the peak (figure 8). Assuming an approximate 20°C daily average temperature
at this site in late June and early July, this could translate to a 10-or-more day change in activity
peak owing to precipitation extremes in any given year. Yet, there are several alternate explanations
for the pattern we observed, and some patterns detected may have been driven by statistical outliers.
For example, very high rainfall in 2015 at the site strongly influences our conclusion that a quadratic
relationship exists between degree-day accumulation and precipitation accumulation in explaining
firefly activity peaks. Indeed, if observations from 2015 had not been included in our analysis, we
would probably have concluded that degree-day accumulation had a negative, linear relationship with
precipitation accumulation; that is, increasing rainfall would cause fireflies to emerge earlier, given
constant temperatures. This result would align with previous work showing that firefly abundance
in Japan is generally negatively correlated with rainfall [35]. However, considered within the context
of firefly biology, it seems unlikely that the relationship between these parameters would be linear
throughout the range of possible precipitation values, as soil-dwelling larvae of non-aquatic firefly
species and/or their prey are probably adversely affected by abnormally waterlogged soils.
As the sampling at our study site continues, we will watch rainy years with particular interest to
determine whether population data collected in these years support or refute this pattern, or if an
alternate driver can explain more of the variation. Indeed, firefly activity may have been driven by factors
not considered in this study. Although using a start date of 1 March was favoured in our analysis (i.e. the
AIC of the model using this start date was minimized), when the start day was changed in a sensitivity
analysis, the relationship between degree-day accumulation and precipitation in firefly activity changed
or disappeared. This result could suggest that alternate drivers not accounted for in this study may be
driving aspects of firefly activity. Factors such as winter snow cover and variations in winter temperature
are known to affect the phenology of temperate insects [41], and thus these factors should be considered
in subsequent work.
In this study, we have clearly demonstrated a taxon whose phenology varies in response to multiple
drivers. Species with phenological responses to multiple drivers are not rare [42]. Yet ecological
interactions among species with multiple drivers of phenology may be complex and unpredictable
[43,44], potentially leading to dire consequences in a changing environment [45]. Our study examined
the phenological responses to environmental conditions of adult fireflies; however, data on larvae or sex
of the adults were unavailable. Adult Photinus fireflies are non-feeding [14], so shifts in their activity are
unlikely to have direct consequences through phenological asynchronies. Shifts in adult activity probably
correspond to shifts in development or activity among larvae, potentially leading to asynchronies
between larvae–prey populations at this critical development time period. Resources acquired during
the predaceous larval stage are important in determining mating success among adult fireflies: males
provide an energetically costly nuptial gift to the female in the form of a spermatophore [46]. If sex
differences in phenological responses to environmental conditions exist, asynchronies between males
and females may additionally reduce mating success and fecundity [47]. Male fireflies were always
observed earlier than females in Elkmont, TN, USA. In fact, in that system, females were often found
during or after peak emergence of males and thus this should be an area of emphasis in future study
[25]. Additionally, phenological shifts in fireflies may lead to consequences at other trophic levels.
For example, generalist ground-dwelling predators like firefly larvae and other predaceous beetles are
known to have dramatic effects on the establishment of agricultural pests early in the growing season
[48]. Similarly, although distasteful and avoided by many predators, some birds, lizards and frogs are
known to feed on adult fireflies [49], thus shifts in firefly activity may have dietary consequences for
animals at higher trophic levels.

11
rsos.royalsocietypublishing.org R. Soc. open sci. 3: 160712
................................................
5. Conclusion
Fireflies are a charismatic and important taxon with ties to trophic function, economic importance
and culture. Although empirical evidence of specific declines of Photinus fireflies has not been clearly
demonstrated in longitudinal studies, naturalists and citizen scientists perceive a decline in their number
[21], leading to interest in their conservation. Our study has offered new insight to support conservation
efforts and to direct future research. Photinus pyralis appears to thrive in habitats with moderate soil
disturbance. Thus, efforts to foster no-till and perennial agricultural systems [50,51] will probably benefit
the species. Climate warming may advance the activity of fireflies to progressively earlier in the growing
season, but other extremes of climate in the form of precipitation may introduce unpredictable elements
to this, and add the possibility of inducing asynchrony with other systems.
The availability of long-term observational data, made freely accessible to the public, was an essential
factor in the discoveries made in this study. Although the study that provided these data was not initiated
with this purpose in mind, we were able to empirically demonstrate and disentangle the effect of multiple
drivers on firefly phenology simply because we had the statistical power to do so. Although species
that respond to multiple, interacting environmental drivers are relatively common, data supporting
investigations of this kind are rare [52]. We therefore encourage all practising ecologists to curate
their species observation data and make them publicly available, to foster long-term, broad-scale
investigations in the future [53–55].
Data accessibility. All data and analysis code produced by this study are publicly available. Lampyrid abundance data
are available at https://ndownloader.figshare.com/files/3686040. Weather station data are available at http://lter.
kbs.msu.edu/datatables/7.csv. These data are automatically downloaded when the R script file at https://github.
com/cbahlai/lampyrid/ is run.
Authors’ contributions. S.L.H. contributed to conception, led study design and wrote major portions of the manuscript.
S.X. conceived of analyses and wrote portions of the manuscript. L.R. led exploratory analysis and wrote portions of
the manuscript. E.D.L., A.M. and B.E. provided critical commentary on the design and analysis of the study and wrote
portions of the manuscript. C.A.B. conceived of the study, led the analysis, supervised the drafting of the manuscript
and critically revised its content. All authors approved the final content of the manuscript.
Competing interests. We declare no competing interests.
Funding. Data used in this study were produced with funding from the National Science Foundation Long Term
Ecological Research program grant no. 1027253. C.A.B. was funded by a fellowship from the Mozilla Foundation
and the Leona M. and Harry B. Helmsley Charitable Trust.
Acknowledgements. This paper was written as part of a course in Open Science and Reproducible Research offered
through the Department of Entomology at Michigan State University. The authors would like to thank the Mozilla
Science Laboratory and the rest of the open science community for their support in the drafting of this manuscript,
and Stuart Gage, Manuel Colunga-Garcia and Douglas Landis for design and maintenance of the study that produced
the data used herein. Monica Granados helpfully provided review of our analysis R code. Lynn Faust provided keen
insight and invaluable commentary on an earlier version of this manuscript. We would like to thank Tyson Wepprich
and one additional anonymous reviewer for helpful comments that helped us question our assumptions and improve
the clarity of this manuscript.
References
1. Rosenheim JA et al. 2011 Ecoinformatics for
integrated pest management: expanding the
applied insect ecologist’s tool-kit.J. Econ. Entomol.
104, 331–342. (doi:10.1603/EC10380)
2. Bahlai CA, Landis DA. 2016 Predicting plant
attractiveness to pollinators with passive
crowdsourcing.R. Soc. open sci. 3, 15067 7.
(doi:10.1098/rsos.150677)
3. WangY, Fu X, Lei C, Jeng M-L, Nobuyoshi O. 2007
Biological characteristics of the terrestrial rey
Pyrocoelia pectoralis (Coleoptera:Lampyridae).
Coleopterists Bull.61, 85–93. (doi:10.1649/907.1)
4. WingSR.1989Energeticcostsofmatingina
ightless female rey,Photinus collustrans
(Coleoptera: Lampyridae).J. Insec t Behav.2,
841–847. (doi:10.1007/BF01049405)
5. Hess WN. 1920 Noteson the biology of some
common Lampyridæ. Biol. Bull.38,39–76.
(doi:10.2307/1536232)
6. Williams FX. 1917 Notes on the life-histor y of some
NorthAmericanLampyridæ.J. NY Entomol. Soc. 25,
11–33.
7. Eisner T, Goetz MA, Hill DE, Smedley SR, Meinwald
J. 1997 Firey ‘femmes fatales’ acquiredefensive
steroids (lucibufagins) from their rey prey.Proc.
NatlAcad.Sci.USA94,9723–9728.(doi:10.1073/
pnas.94.18.9723)
8. KeiperRR, Solomon LM. 1972 Ecology and yearly
cycle of the rey Photuris pennsylvanica
(Coleoptera: Lampyridae).J. NY Entomol. Soc. 80,
43–47.
9. BranchiniBR, Murtiashaw MH, Magyar RA, Portier
NC, Ruggiero MC, Stroh JG. 2002 Yellow-greenand
red rey biol uminescence fro m
5,5-dimethyloxyluciferin. J. Am. Chem. Soc. 124,
2112–2113. (doi:10.1021/ja017400m)
10. Schaap AP, Gagnon SD. 1982 Chemiluminescence
from a phenoxide-substituted 1,2-dioxetane: a
model for rey bioluminescence. J. Am. Chem. Soc.
104, 3504–3506. (doi:10.1021/ja00376a044)
11. Seliger HH, Buck JB, Fastie WG, McElroyWD. 1964
The spectral distribution of rey light. J. Gen.
Physiol. 48,95–104.(doi:10.1085/jgp.48.1.95)
12. Branham MA, Wenzel JW. 2003 The origin of photic
behavior and the evolution of sexual
communication in reies (Coleoptera:
Lampyridae). Cladistics 19,1–22.
(doi:10.1111/j.1096-0031.2003.tb00404.x)
13. Lall AB, Seliger HH, Biggley WH, Lloyd JE. 1980
Ecology of colors of rey bioluminescence.
Science 210, 560–562. (doi:10.1126/science.210.
4469.560)
14. Lloyd J. 1997 Firey mating ecology,selection and
evolution. In The evolution of mating systemsin
insects and Arachnids (eds JC Choe, BJ Crespi), pp.
184–192. Cambridge, UK: Cambridge University
Press.

12
rsos.royalsocietypublishing.org R. Soc. open sci. 3: 160712
................................................
15. NationalMuseumofNaturalHistory2013State
Bugs. Encyclopedia Smithsonian.Seehttps://www.
si.edu/Encyclopedia_SI/nmnh/buginfo/statebugs.
htm (accessed 2 June 2016).
16. Kobori H, Primack RB. 2003 Participatory
conservation approaches for Satoyama,the
traditional forest and agricultural landscape of
Japan. Ambio 32, 307–311. (doi:10.1579/0044-
7447-32.4.307)
17. TakedaM,AmanoT,KatohK,HiguchiH.2006The
habitat requirement of the Genji-reyLu ciola
cruciata (Coleoptera: Lampyridae), a representative
endemic species of Japanese rural landscapes.
Biodivers. Conserv.15, 191–203. (doi:10.1007/s10531-
004-6903-y)
18. Takada K. 2011 Popularityof dierent lampyrid
species in Japanese culture as measured by Google
search volume. Insects 2, 336–342. (doi:10.3390/
insects2030336)
19. Faust LF. 2004 Fireiesas a catalyst for science
education. Integr.Comp. Biol. 44, 264–265.
(doi:10.1093/icb/44.3.264)
20. Gardiner T. 2009 Glowing, glowing, gone? A history
of glow-worm recordingin the UK. Corby, UK: British
Nationalists’ Association. (http://www.bna-
naturalists.org/gworm.pdf) (accessed 2 June 2016).
21. Chow AT, Chong J-H, Cook M, White D. 2014
Vanishing reies: a citizen-scienceproject
promoting scientic inquiry and environmental
stewardship.Sci. Educ. Civic Engag. 6, 23–31.
22. Firey Watch.2016 Museum of science. See https://
legacy.mos.org/reywatch/(acce ssed 2 June
2016).
23. Scagell, R. 2016 The UK glow worm survey.See
http://www.glowworms.org.uk/(acces sed 2 June
2016).
24. Picchi MS, Avolio L, Azzani L, Brombin O, Camerini
G. 2013 Fireies and land use in an urban landscape:
thecaseofLuciolaitalica L. (Coleoptera:
Lampyridae) in the city of Turin. J.Insec t Conserv.17,
797–805. (doi:10.1007/s10841-013-9562-z)
25. FaustLF, Weston PA. 2009 Degree-day prediction of
adult emergence of Photinus carolinus(Coleoptera:
Lampyridae). Environ. Entomol. 38, 1505–1512.
(doi:10.1603/022.038.0519)
26. Guiney MS, Oberhauser KS. 2009 Insects as agship
conservation species. Terr. Ar thropodRev. 1, 111–123.
(doi:10.1163/187498308X414733)
27. Diez JM, Ibáñez I, Miller-Rushing AJ, Mazer SJ,
Crimmins TM, Crimmins MA, Bertelsen CD, Inouye
DW. 2012 Forecasting phenology: fromspecies
variability to community patterns. Ecol. Lett.
15, 545–553. (doi:10.1111/j.1461-0248.2012.
01765.x)
28. Miller-Rushing AJ, Høye TT, Inouye DW, Post E. 2010
The eects of phenological mismatches on
demography.Phil.Trans.R.Soc.B365,317 7–3186.
(doi:10.1098/rstb.2010.0148)
29. Root TL, MacMynowski DP, Mastrandrea MD,
Schneider SH. 2005 Human-modied temperatures
induce species changes: joint attribution. Proc. Natl
Acad.Sci.USA102, 7465–7469.(doi:10.1073/pnas.
0502286102)
30. Parmesan C. 2006 Ecological and evolutionary
responses to recent climatechange. Annu.Rev.Ecol.
Evol. Syst. 37,637–669.(doi:10.1146/annurev.
ecolsys.37.091305.110100)
31. Rosenzweig C et al. 2008 Attributing physical and
biological impacts to anthropogenic climate
change. Nature 453, 353–357. (doi:10.1038/
nature06937)
32. Crimmins TM, Crimmins MA, Bertelsen D, Balmat J.
2008 Relationships between alpha diversity of plant
species in bloom and climatic variables across an
elevation gradient. Int.J. Biometeorol. 52, 353–366.
(doi:10.1007/s00484-007-0130-7)
33. Crimmins TM, Crimmins MA, David Bertelsen C.
2010 Complex responses to climatedrivers in onset
of spring owering across a semi-arid elevation
gradient. J. Ecol.98,1042–1051.
(doi:10.1111/j.1365-2745.2010.01696.x)
34. Visser ME, Holleman LJM. 2001 Warmer springs
disrupt the synchrony of oak and winter moth
phenology. Proc.R.Soc.Lond.B268, 289–294.
(doi:10.1098/rspb.2000.1363)
35. YumaM. 2007 Eect of rainfall on the long-term
population dynamics of the aquatic rey Luciola
cruciata.Entomol. Sci. 10, 237–244.
(doi:10.1111/j.1479-8298.2007.00219.x)
36. Wickham H. 2016 resh ape2: Flexibly reshape data: a
reboot of the reshape package, version 1.4.2. See
https://cran.r-project.org/web/packages/
reshape2/index.html.
37. Allen JC. 1976 A modied sine wave method for
calculating degree days. Environ. Entomol. 5,
388–396. (doi:10.1093/ee/5.3.388)
38. Wickham H, Chang W. 2016 ggplot2: an
implementation of the grammar of graphics,
version 2.1.0. See https://cran.r-project.org/web/
packages/ggplot2/index.html.
39. Oksanen J et al. 2016 vegan: communit y ecology
package, version 2.4-1. See https://cran.r-project.
org/web/packages/vegan/index.html.
40. Jusoh WFAW,Hashim NR, Ibrahim ZZ. 2010 Firey
distribution and abundance on mangrove
vegetation assemblages in Sepetang Estuary,
Peninsular Malaysia. WetlandsEcol. Manag. 18,
367–373. (doi:10.1007/s11273-009-9172-4)
41. Forrest JRK, Thomson JD. 2011 An examination of
synchrony between insect emergence and
owering in Rocky Mountain meadows. Ecol.
Monogr.81,469–491.(doi:10.1890/10-1885.1)
42. Wolkovich EM et al. 2012 Warming experiments
underpredict plant phenological responses to
climate change. Nature 485, 494–497.
(doi:10.1038/nature11014)
43. Hua F et al. 2016 Community-wide changes in
intertaxonomic temporal co-occurrence resulting
from phenological shifts. Glob.Change Biol. 22,
1746–1754. (doi:10.1111/gcb.13199)
44. Robinet C, Roques A. 2010 Direct impacts of recent
climate warming on insect populations. Integr.
Zool. 5,132–142.(doi:10.1111/j.1749-4877.2010.
00196.x)
45. McLaughlin JF, Hellmann JJ, Boggs CL, Ehrlich PR.
2002 Climate change hastens population
extinctions. Proc. Natl Acad.Sci. USA 99, 6070–6074.
(doi:10.1073/pnas.052131199)
46. Lewis SM, Cratsley CK, Rooney JA. 2004 Nuptial gifts
and sexual selection in Photinus reies. Integr.
Comp. Biol. 44, 234–237. (doi:10.1093/icb/44.
3.234)
47. CalabreseJM, R ies L, MatterSF, D ebinski DM,
Auckland JN, Roland J, Fagan WF. 2008
Reproductive asynchrony in natural buttery
populations and its consequences for female
matelessness. J. Anim. Ecol.77, 746–756.
(doi:10.1111/j.1365-2656.2008.01385.x)
48. Safarzoda S, Bahlai CA, Fox AF, Landis DA. 2014 The
role of natural enemy foraging guilds in controlling
cereal aphids in Michigan wheat. PLoS ONE 9,
e114230. (10.1371/journal.pone.0114230)
49. Lloyd JE. 1973 Firey parasites and predators.
Coleopterists Bull.27, 91–106.
50. Werling BP, Meehan TD, Gratton C, Landis DA. 2011
Inuence of habitat and landscape perenniality on
insect natural enemies in three candidate biofuel
crops. Biol. Control59,304–312.(doi:10.1016/
j.biocontrol.2011.06.014)
51. Werling BP et al. 2014 Perennialgrasslands enhance
biodiversity and multiple ecosystem services in
bioenergy landscapes. Proc.Natl Acad. Sci. USA 111,
1652–1657. (doi:10.1073/pnas.1309492111)
52. StrayerDL, Eviner VT, Jeschke JM, Pace ML. 2006
Understanding the long-term eects of species
invasions. Trend s Ecol. Evo l. 21, 645–651.
(doi:10.1016/j.tree.2006.07.007)
53. Michener WK, Jones MB. 2012 Ecoinformatics:
supporting ecology as a data-intensive science.
Tren ds Ecol. Ev ol. 27, 85–93. (doi:10.1016/j.tree.
2011.11.016)
54. Hampton SE et al. 2015 The Tao of open science for
ecology. Ecosphere 6,art120.(doi:10.1890/ES14-
00402.1)
55. Hampton SE, Strasser CA, Tewksbury JJ, Gram WK,
Budden AE, Batcheller AL, Duke CS, Porter JH. 2013
Bigdataandthefutureofecology.Front. Ecol.
Environ. 11, 156–162. (doi:10.1890/120103)