Butterflies are important bioindicators that can be used to monitor the effects of climate change, particularly in montane environments. Changes in butterfly population size over time, reflective of indicator life stages, can signal changes that have occurred or are occurring in their environment indicating ecosystem health. From the perspective of understanding butterflies as bioindicators in these systems, it is essential to identify influential environmental variables at each life stage that have the greatest effect on population dynamics. Life stage hypothesis modeling was used to assess the effects of multiple temperature and precipitation metrics on the population growth rate of a Parnassius clodius butterfly population from 2009 to 2018. Extreme maximum temperatures during the larval-pupal life stages were identified to have a significant negative effect on population growth rate. We speculate that higher temperatures during the spring ephemeral host plant’s flowering, and P. clodius' larval stage, may lead to earlier plant senescence and lower P. clodius growth. Because Parnassius butterflies are well studied from a global perspective, results may aid in understanding the potential indicator life stages of other insect species in montane environments to climatic changes. Findings from this study demonstrate the value in assessing a butterfly species’ response to short-term weather variation or long-term climatic changes at each life stage in order to protect and conserve insects and their interactions with other organisms.

Butterfly Mark-Recapture

Mark-recapture methods were used to study a population of P. clodius at Pilgrim Creek in Grand Teton National Park, Wyoming, USA across annual flight seasons between 2009 and 2018 during June and July. Surveys were not carried out in 2012 and 2013. Six 50m x 50m plots a minimum of 100m apart, were located using GPS units, flagged prior to the flight season of P. clodius, and surveyed each year. Survey plots were initially established in 2000 in an effort to balance increasing the area sampled, decreasing the number of recaptures, and maintaining independent sampling within a single meadow (Auckland et al. 2004). Mark-recapture surveys began a few days after the beginning of the flight season and continued until only one or two butterflies per plot were caught during a survey period. Plots were monitored daily if weather permitted throughout each flight season. Surveys were conducted when temperatures were above 21°C, wind was <16kmh^-1, and clouds were not obscuring the sun. If weather prevented all six plots from being sampled in one day, the rotation was completed the following day. The order in which plots were surveyed rotated throughout the flight season to vary the time of day surveys were conducted for each plot.

Two people surveyed each plot for 20 minutes between 10:00 and 17:00 hours. All P. clodius butterflies in a plot were netted by hand and held in glassine envelopes until the survey ended. The time of capture and activity of the butterfly when captured (i.e., nectaring, chasing, stationary) was noted on each envelope. Following each survey, captured butterflies were marked with a permanent marker on each hindwing with an identification number indicating the plot it was caught in and the butterfly’s individual number (Figure 1). The number corresponded to the consecutive butterfly count throughout a season. The following metrics were recorded for each captured butterfly: date, plot number, identification number, time of capture, sex, and whether the butterfly was a recapture. After data were collected for each butterfly and butterfly markings were added, all butterflies were released in the center of each plot by placing butterflies on surrounding plants, rather than being placed on the ground where they could be stung by ants.

Weather Variable Selection

A list of meaningful weather variables expected to affect populations of P. clodius butterflies was formulated based upon previous research that identified influential weather variables affecting the population dynamics of other Parnassius species (Matter et al. 2011, Roland and Matter 2013, 2016, Matter and Roland 2017) as well as our observations at Pilgrim Creek over time. The list of explanatory variables included: date of snowmelt, maximum snow water equivalent (cm), mean snow water equivalent (cm), mean temperature (C), mean maximum temperature (C), extreme maximum temperature (C), mean minimum temperature (C), extreme minimum temperature (C), accumulated rainfall (cm) from 2008 to 2018 (Table 1). All temperature data were sourced from the USDA Base Camp SNOTEL site located at 43.93333°N, -110.45000°W at 2151 m in elevation and 10.7 km from Pilgrim Creek meadow. All snow water equivalent and precipitation data were sourced from the USDA Snake River Station SNOTEL site located at 44.13333°N, -110.66667°W at 2109 m in elevation and 25.1 km from Pilgrim Creek meadow. 

Analysis

The goal of these analyses was to identify which weather variables, known to affect other butterfly species, explain variation in P. clodius population growth through each life stage. Estimates of population change were calculated for the Pilgrim Creek, WY population of P. clodius for the duration of the study (i.e., 2009-2011, 2014-2018).  Calculations of population change (R_t) were estimated as the ratio of the number of caught butterflies (N_t) for one flight season after accounting for effort, relative to the number of caught butterflies of the previous flight season after accounting for effort.  Effort was measured as the number of two-person 20-minute plot surveys conducted during each flight season. R_t was calculated as log₁₀((N_t+1/effort_t+1)/(N_t/effort_t)) (Roland and Matter 2016). This simplified approach to estimating population size used in previous research (Roland and Matter 2016) uses information regarding known observed butterflies in combination with standardized sampling effort allowing researchers to produce tangible values.  Our estimate of population size was included in each model to evaluate density dependence as an explanatory variable on population growth. Population size was estimated by taking the log of the number of caught butterflies divided by the number of surveys conducted for each flight season (i.e., log₁₀(N_t/effort_t)).

A life stage hypothesis modeling approach was used to evaluate the effect of each chosen weather variable during each life stage on the population growth (i.e., R_t) of P. clodius butterflies. The observed values for each weather variable were partitioned by life stage to evaluate each variable’s effect during each stage of the P. clodius lifecycle. Dates of each life stage were estimated from researcher observations (L. Crees, personal communication, October 2022) of P. clodius over time to capture the estimated beginning and end dates of each life stage.  We categorized the egg stage as July 21 - April 30, the larva-pupa stage as May 1 - May 31, and the adult stage June 1 - July 20. Each weather variable was tabulated for the egg stage, larva-pupa stage, and the adult butterfly stage (Table 1). The larva and pupa life stages were combined due to the uncertainty of the date of when one stage ended and the other began.

Correlation tests were conducted between all explanatory weather variables to prevent the inclusion of highly correlated variables in models, which could lead to multicollinearity and unreliable estimates. Uncorrelated precipitation and temperature explanatory variables and logN_t (i.e., density dependence) were included in each generalized linear model. Then, generalized linear models using the glm function in R were fit to evaluate which explanatory weather variables were the most descriptive in explaining population change (R_t). Additionally, generalized additive models using the gam function in R were fit to evaluate quadratic relationships between population change (R_t) and explanatory weather variables. AIC comparison between models was conducted to determine the most informative explanatory variables. Following generalized linear models that included logN_t and interactions between the most informative temperature and precipitation explanatory variables were used to evaluate potential interactive effects between weather variables on population change. Final AIC comparison was conducted to determine the best model to describe population change for each P. clodius life stage.