The cost of movement: assessing energy expenditure in a long-distant ectothermic migrant under climate change Free

Animal migration, a large-scale ecological and physiological phenomenon, occurs across vast heterogeneous landscapes where individuals are subjected to a suite of different climatic conditions, including temperature, precipitation and relative humidity, in a relatively short amount of time (Dingle, 2014). This phenomenon occurs in species ranging from whales to butterflies (Dingle, 2014), and across multiple modes of movement including swimming, flying or walking, all of which vary in their cost of transport (Schmidt-Nielsen, 1972; Dingle, 2014). The preparation required for these long voyages consists of changes to behavioral (e.g. hyperphagia; Odum, 1960), physiological (e.g. fat accumulation; Bairlein, 2003), morphological (e.g. wing aspect ratio; Satterfield and Davis, 2015), reproductive (e.g. diapause; Reppert and de Roode, 2018) and navigational (e.g. compass calibration; Guerra and Reppert, 2013) phenotypes to ensure success. Given the vast distances traveled, many migratory animals must stop and refuel to meet the high metabolic demands associated with the long-distance movement in tandem with homeostatic maintenance (Sapir et al., 2011; Ferretti et al., 2019). However, unfavorable migratory conditions, such as extreme temperature, storms or high winds, may force individuals to stop or refuel more often, thus impacting their migratory progression (Sapir et al., 2011; Clipp et al., 2020).

For ectothermic migrants such as lepidopterans (Gao et al., 2020; Guerra, 2020), there may be limitations on where and when to stop and refuel due to habitat degradation or fragmentation of resources (Hobson et al., 2020; Chowdhury et al., 2021). The ambient conditions of the landscape that ectothermic species migrate through can influence their physiological performance norms (Angilletta, 2009), thus affecting movement propensity and the ability to orient and navigate to their destination (Rappole and Warner, 1976; Dingle, 2014). For a single-generation migrant that experiences a single maiden journey, having enough energy reserves during earlier life history stages (e.g. larva) to emerge as an adult with the morphological (e.g. reproductive diapause, wing morphology) traits and physiological (e.g. fat accumulation) capabilities to migrate are equally as important as the gene expression required for orientation and navigation (Hahn and Denlinger, 2011).

For smaller migratory animals such as butterflies and moths, there is difficulty in estimating field energy expenditure during migration because some tracking methods (e.g. biologging technology) are still too heavy. Although lepidopterans can generate 25% more lift than traditional fliers through the use of the clap and fling mechanism (Marden, 1987), the energetic implications of increased weight loading have not been thoroughly investigated, as greater mass necessitates a greater power requirement for flight and speed (Tennekes, 2009). As such, the general rule when tracking and monitoring flying migrants is that devices should be less than 5% of the total body mass as a conservative estimate, but most importantly these devices cannot impede the flight capability or performance of the animal (Kissling et al., 2014). A current alternative method to estimate energy expenditure for smaller flying migrants is to use detailed mathematical modelling (e.g. Tsuji et al., 1986) and computer simulations with micro-meteorological conditions from a vast array of weather stations (Shamoun-Baranes et al., 2010).

We used monarch butterflies (Danaus plexippus), a multi-generational migratory species, as a model organism to simulate the migratory propensity and estimated daily energy expenditure (DEE) of migratory individuals, from tag–recapture data and local weather conditions associated with these data during autumn migration. These animals exhibit a migratory syndrome within a single generation during the autumn, with many individuals traveling over 4000 km to their overwintering sites without ever having been to the location themselves (Urquhart, 1960). For instance, eastern North American monarchs, those butterflies that live east of the Rocky Mountains, will leave their northern breeding ranges in Southern Canada and the northern tier of the USA during the autumn, to migrate to specific overwintering locations in Central Mexico (Guerra, 2020). For monarchs, ideal weather conditions are necessary for fine-scale movement, including tolerable temperatures, no precipitation, moderate relative humidity and low wind speeds (Urquhart, 1960; Rowley et al., 1968; Kammer, 1970; Masters et al., 1988; Garland and Davis, 2002; Guo et al., 2020), although the precise range of conditions remains unclear. The monarch orientation response is robust and well documented for the eastern North American population, showing a southwest orientation during the autumn migration where individuals primarily use a time-compensated sun compass and a backup magnetic inclination compass (Guerra, 2020) to successfully migrate to their overwintering sites. For the monarch, migration is a fundamental component of the life cycle that integrates physiological, behavioral and ecological traits into a key life history migratory syndrome (Guerra, 2020; Chowdbury et al., 2021). However, under climate change, weather conditions will become more extreme and likely affect migratory movement as a result of hostile conditions (Halsch et al., 2021), such as higher maximum daily temperatures (Román-Palacios and Wiens, 2020).

Our goal was to identify the additional cost of temperature-dependent homeostatic maintenance with auxiliary functions for a migratory ectothermic invertebrate, and then determine what the effect would be on total energy reserves of a theoretical migratory butterfly. For this study, the benefit of tag–recapture data is that there are two geographic points at which the butterfly was observed, and this permits the generation of a straight-line path that is representative of the migratory path of that individual. Further, to estimate location-specific energy expenditure, we evaluated the migratory pace to determine time-specific thermal conditions. We tested the hypotheses that temperature-dependent physiological processes reduce energy reserves during migration faster than previously estimated and that increasing climatic temperatures will intensify baseline DEE. We first assessed the mass conversion from a caterpillar to an adult along with standard metabolic rate (SMR) across temperatures as a baseline for the estimation of potential adult fat stores for energy utilization. Next, using tag–recapture data, we determined the migratory pace of monarchs based on time-specific micro-meteorological conditions that would permit flight. Then, based on those local micro-meteorological conditions, we estimated the energy expenditure of migratory flight and compared it with previous estimates of monarch energy expenditure. Finally, we estimated the energy expenditure during the migratory months (August–October) across a time period from 1961 until 2018, and forecast potential future energy expenditure based on these trends.

Identifying the energy expenditure in a small ectothermic migrant requires an understanding of changes in life history patterns that may influence energy reserves for migration (Hahn and Denlinger, 2011). For our model, we had to consider the following aspects: (1) the variation in energy use throughout the monarch's life cycle (Fig. 1A), (2) the energy expenditure under different thermal conditions (Fig. 1A), and (3) the estimated speed of the monarch in response to real meteorological conditions (Fig. 1B). With these estimates, we can then look at changes in thermal conditions in parallel to forecast potential for changes in energy use (Fig. 1C), as well as ways to improve model simulation. Migratory monarchs need initial energy reserves that facilitate long-distance flight, and potential changes in energy use throughout the life history could contribute to the final eclosed size of the individual as a migratory adult. As metabolic rate is tightly correlated with temperature (Angilletta, 2009; Clarke, 2017), changes in ambient thermal conditions can alter the energy demand of migratory individuals. Finally, the migratory pace determines how real meteorological conditions experienced by the monarch facilitate migratory movement and how much energy is consumed. Below, we outline how we assessed each component, where we determined parameters or used literature values, and then how we integrated all components into the current working simulation. We used literature values in the model where possible and validated parameters, and elaborate on the comparison with Gibo and Pallett (1979) in the Supplementary Materials and Methods. Throughout the study, we used a hypothetical 500 mg adult monarch butterfly with 115 mg of fat stores. This represents a monarch in the early stages of migration, similar to the ratio of powered flight to gliding reported by Gibo and Pallett (1979). However, it is important to note that throughout the migration, monarchs can gain up to 500% of the lipid contents when migrating through Texas and northern Mexico (Brower et al., 2009) and this could change the ratios of powered to gliding flight.

Adult migratory monarchs, Danaus plexippus (Linnaeus 1758), were wild caught at the University of Cincinnati Center for Field Studies (UCCFS; N39.28, W84.74) during the 2021 autumn migration between 1 September and 15 October. Each captured butterfly was kept in a glassine envelope and housed in an environmentally controlled incubator (Percival model I-36LL, Perry, IA, USA) with a 12 h:12 h light:dark (L:D) cycle and a corresponding 21–12°C L:D temperature cycle, and butterflies were fed a 25% honey solution every other day.

We also reared monarchs from eggs (n=60; from Jaap de Roode, Emory University) under a 12 h:12 h L:D cycle with cooler night-time temperatures using a 21–12°C L:D temperature cycle to mimic autumn-like conditions. These eggs were from monarchs maintained under autumn-like conditions through the winter and then brought to summer temperatures to break diapause, similar to the natural progression of conditions during their breeding season. Animals were housed in a translucent plastic box (35.56×20.32×12.39 cm) with a maximum of 10 individuals per box and fed common milkweed (Asclepias syriaca) leaves (5 leaves per box) picked from wild plants located at the UCCFS. Milkweed leaves were changed daily, and we checked a subset of leaves for any Ophryocystis elektroscirrha (OE) spores, a harmful protozoan parasite of monarchs (Altizer and Oberhauser, 1999), under a microscope. We also checked eclosed adults for OE, by applying a small amount of scotch tape on the abdomen of the monarch and gently pulling back prior to viewing on a glass slide under a microscope. Individuals were considered infected if we found a single spore, and were not included in experiments. Prior to pupation, caterpillars were weighed the caterpillars and individuals were then allowed to form a chrysalis in a plastic cup with a unique ID. Once the chrysalis was formed and had had at minimum of 36 h to dry, we weighed the chrysalis and hung it back up in the cup using a metal clip to hold the silk button. We then monitored chrysalises until the adults emerged (20–23 days later) and obtained the final mass of the butterfly from which we then calculated the percentage change from caterpillar to adult. We determined whether female monarchs (n=21) were in reproductive diapause 1 month after eclosion, a characteristic trait of the migratory syndrome, by splaying the abdomen open and checking for reproductive status (Fig. S1A). We counted females as having broken diapause if we found a single mature oocyte.

We quantified SMR in laboratory-reared caterpillars and wild-caught adult butterflies using an oxygen sensor (O₂ Sensor, Vernier, Beaverton, OR, USA with KE-25F3 O₂ gas sensor, Figaro Engineering Inc., Mino, Osaka, Japan) in an airtight 50 ml Falcon tube. In each tube, we added a small amount of soda lime (100 mg; Jorgensen Laboratories, Loveland, CO, USA) to prevent CO₂ build up. Soda lime was placed underneath the sensor and above the animal in a custom-made wire mesh (Fig. S1B). We measured the relative humidity and pressure before the trial using a BME680 temperature, humidity, pressure sensor (Seeed Technology Co., Ltd, Shenzhen, China), and temperature was monitored during the trial using a copper-constantan thermocouple (HH806AU, Omega Inc., Stamford, CT, USA). Room and environmental chamber relative humidity were low (<20%), but we used the pre-trial measurement to correct gas volumes for subsequent calculations (eqn 1 of Bartholomew and Casey, 1978; Lighton, 2008; see Supplementary Materials and Methods and Fig. S1 for apparatus setup and sensor specifications). The Vernier oxygen sensor recorded the oxygen percentage using an electrochemical fuel cell, and from these data we determined the slope (% O₂ s⁻¹) and we then estimated metabolic rate (ml O₂ g⁻¹ h⁻¹). Caterpillars and adult monarchs were recorded for 10 min, as this produced a discernible decrease in oxygen within the airtight Falcon tube. Caterpillars (n=10) and chrysalises (n=10) were tested at 10, 20 and 30°C (n=10 per temperature) to provide an ecologically relevant range of temperatures. Given that lepidopteran pupae have extremely low rates of oxygen consumption (e.g. Hyalophora cecropia; Schneiderman and Williams, 1953), we measured the CO₂ production via flow-through respirometry using a LiCor gas analyzer (LiCor 6400XT, LiCor Biosciences, Lincoln, NE, USA) with the insect respirometry program (6400-89 Insect Respiration Chamber configuration, see Supplementary Materials and Methods and Fig. S1 for details). Metabolic rates were converted to microwatts by assuming joule equivalents of 20.7 J ml⁻¹ (O₂) or 24.7 J ml⁻¹ (CO₂) (Chown et al., 2007) for comparing between stages and temperatures. For the adult monarchs, previous work has estimated the metabolic rate of migratory individuals from 5 to 20°C (Chaplin and Wells, 1982), so we sought to validate these previous measurements (i.e. 5–20°C) and expanded the range to examine higher temperatures that animals may experience during migration (i.e. 25–35°C at 5°C increments). All animals were acclimated to the test temperature for at least 1 h prior to testing and were weighed before and after testing using a digital scale (MXX-412, Denver Instruments, Arvada, CO, USA; ±0.01).

For the model, we based the initial energy reserves (mg fat) on the final average eclosed mass of the laboratory-reared adult monarchs. While it is possible for monarchs to refuel on the migratory route, we did not include the ability to stop and refuel in our model because of limitations in accessible data during the tag–recapture dates.

We had a total of 108 tag–recapture locations of migratory monarchs between 1956 and 2018. However, of these 108 tag–recaptures, only 25 had complete meteorological conditions without missing information for simulation. Given our data source, we focused on the eastern North American monarch population, as weather station information was more reliable and consistent. We only used tag–recapture locations that had complete weather station datasets and had a displacement greater than 50 km. For the modelling of migratory movement, the locations varied between 67 and 1984 km in displacement and were 1–38 days between tagging and recapture. The tag–recapture data for simulations were from previous studies (Urquhart, 1960, n=18; Garland and Davis, 2002, n=6) and from Monarch Watch (www.monarchwatch.org, n=1). For estimating energy expenditure over time, we assessed all 108 tagging locations based on the month of capture to understand potential changes in energy use over time.

Based on the tag–recapture location data, we generated a straight-line path between the capture and recapture location to create new coordinates every 18 km and extracted local meteorological conditions from the nearest weather station (Fig. 1B). These data were taken during the start and end dates of the tag–recapture using the National Oceanic and Atmospheric Administration (NOAA) Local Climatological Data (LCD) (https://www.ncdc.noaa.gov/cdo-web/datatools/lcd, accessed 15 Oct 2020). These stations served as reference points that the computer simulation would index for temporally relevant climatological information to then initiate or inhibit movement between locations. To determine the meteorological conditions that could have been experienced by the monarch during migration, we tested several different proposed monarch butterfly speeds based on historical and contemporary studies that had estimates including 3.9 km h⁻¹ (Davis et al., 2012), 7.0 km h⁻¹ (Davis and Howard, 2005), 14.0 km h⁻¹ (Garland and Davis, 2002) and 18.0 km h⁻¹ (Urquhart, 1960), assuming the monarchs were flying near the ground (Brower, 1985). These speeds permit accessing time-dependent meteorological conditions that were used to estimate energy expenditure in the model. Given the variable speeds documented, we tested each to determine the best average speed to use for this model simulation. While it is possible for monarchs to travel slower or faster than the proposed speeds (Howard and Davis, 2015), we focused on established speeds in the literature given the hourly resolution of the weather station data.

We based the conditions to model monarch movement on several assumptions delineated from natural history observations and laboratory experiments. These conditions included the temperature range permissible for flight (12.7–33.2°C; Masters et al., 1988; Kammer, 1970), hourly mean wind speeds that were not too strong (<25.0 km h⁻¹, moderate winds; Garland and Davis, 2002), no precipitation present (Urquhart, 1960) and tolerable relative humidity for flight (30.0–82.5%; Rowley et al., 1968; Guo et al., 2020) (Fig. 1B). While there is a significant role of wind direction and speed for improving or impeding flight through tail or head winds (Gibo and Pallett, 1979; Gibo, 1986), respectively, the coarse resolution (hour intervals) of the weather station data precluded inclusion in the model simulation. We assumed that monarch butterfly flight took place between 10:00 and 19:00 h during migration (Knight et al., 2019) and limited the computer simulation to those hours for flight. While the upper thermal limit for monarchs is 38°C (Nail et al., 2015; Thogmartin et al., 2017), the coarse temporal resolution of the micro-meteorological conditions precluded allowing the model to search for cooler microclimates.

For estimating monarch migratory movement, the model simulation would index several conditions prior to allowing movement to the subsequent location (Fig. 1B). The sequence for indexing parameters included photoperiod, temperature, relative humidity, wind speed and then precipitation. If any of the conditions were not met based on the aforementioned thresholds, the simulated monarch then remained at its location. If all conditions were met, then the simulated monarch moved to the subsequent location based on the aforementioned speeds. All simulations started at midnight on the day of capture and indexed data from the nearest weather station.

In ectotherms, energy expenditure is tightly correlated with temperature (Angilletta, 2009). For monarchs, however, there is a lack of information on energy expenditure across the full range of ecologically relevant temperatures for SMR or flying metabolic rate. We generated a thermal performance curve using the ‘rTPC’ R package (Padfield et al., 2021; see Table S1 for model comparisons) to estimate the metabolic rate of monarchs under ecologically relevant conditions. We used the thermal performance curve from respirometry measurements from this study (5–35°C) to estimate metabolic rate from ambient temperature for monarch butterflies based on local weather station data.

While butterflies have been shown to maintain a maximum thoracic temperature higher than ambient (e.g. >5°C; Kammer, 1970; Bladon et al., 2020), we assumed that the ambient temperature is reflective of the thoracic temperature maintained during the hour of flight because of the flight strategy (i.e. soaring) primarily used by monarch butterflies during the autumn migration. Gibo and Pallett (1979) found that monarch butterflies used a flight strategy that switched between soaring flight (85% of the time during migration) and powered flight (15% of the time during migration). These time allocations for flight were used to estimate energy expenditure during the hour interval of flight for the computer simulation. Insect flight is one of the most energetically demanding mass-specific activities, where there can be over a 100-fold increase from resting metabolic rate (RMR) to flying metabolic rate in lepidopteran species (e.g. Zebe, 1954). For monarch butterflies, powered flight requires a significantly higher energy expenditure than RMR, with previous work showing between a 25-fold (32°C: Zhan et al., 2014) and 31-fold increase in metabolic rate during flight (22°C: Woods, 2005; Kammer, 1970; Chaplin and Wells, 1982; Masters et al., 1988). This degree of variability within a single species is also found in other lepidopteran species (Zebe, 1954).

For our model, we used the 31-fold increase during powered flight for the hypothetical migratory monarch and a temperature-dependent SMR during soaring flight. For each hour of predicted flight, we assumed that soaring flight accounted for 85% of the time and the remaining 15% of the time was allocated towards powered flight (Gibo and Pallett, 1979). Metabolic rates were then converted into fat used (mg fat) by assuming 1 ml O₂ was equivalent to 4.7 cal, and then using the conversion outlined above (0.11 mg cal⁻¹; Weiss-Foght, 1970) for a hypothetical 500 mg butterfly with 115 mg of fat stores. Powered and soaring energy expenditures were then summed for the total energy expenditure for each hour. If the butterfly was predicted not to move, then we only used the SMR estimate for the hourly energy expenditure (see Fig. 1).

We used the WorldClim2 monthly dataset (2.5 min, 21 km² spatial resolution) from 1961 to 2018 (Fick and Hijmans, 2017) to estimate the DEE (mg fat day⁻¹) of tag–recapture monarchs (n=108) throughout the eastern population of North America. To determine DEE, we first determined the baseline energy expenditure for a hypothetical 500 mg butterfly by using metabolic rate for the mean minimum and maximum temperature during the respective month of tagging for monarchs in August, September or October. Given that these data do not have variance or monthly averages, we used a general photoperiod for each month to estimate the night-time and daytime energy expenditure based on the mean minimum (night) and mean maximum (day) temperature for each month (August 13 h:11 h L:D; September 12.5 h:11.5 h L:D; October 11.5 h:12.5 h L:D) and then summed the estimates to obtain a DEE. This allowed us to compare the estimated DEE across a 57 year period for monarch tag–recapture locations. To forecast the DEE beyond 2018 until 2040, we used a simple moving average (SMA) to calculate the moving average from 1961 until 2018, and then forecast using the ‘smooth’ R package (https://CRAN.R-project.org/package=smooth). We also used the ‘StableClim’ package (Brown et al., 2020) to determine the increase in average temperature (°C year⁻¹) to compare with the month-specific forecasted DEE increase. The future regression was based on the RCP 2.6 to calculate the increase (°C year⁻¹) for North America (2.5 deg spatial resolution raster) and we extracted the rates specific to the tagging location of monarchs. We also used the ensemble mean estimates of monthly temperature (°C) for August, September and October from 1960 to 2040 to extract rates specific to the tagging location of monarchs and generate a SMA for point estimates from 1960 to 2040.

All data were checked for assumptions of normality prior to analysis, and non-parametric methods were used when data failed to meet the assumption of normality as described below. We first assessed the SMR at different test temperatures, life history stages, and the interaction from caterpillar to adult monarch to assess potential changes to initial energy reserves for a migratory monarch across ecologically relevant temperatures. To validate previously established literature values of SMR used in the model, we also compared monarch metabolic rate in 2021 with previous metabolic measurements in 1982 (Chaplin and Wells, 1982) with a two-way ANOVA. Estimating energy expenditure using our model simulation required site-specific meteorological conditions, so we tested the proposed speeds of 3.9, 7.0, 14.0 and 18.0 km h⁻¹. We applied each test speed for each tag–recapture sample, and then we selected the speed with the highest estimated precision to the actual recapture site as the representative speed for that monarch (Fig. 1B). We then analyzed the number of tag–recaptures that were most likely to occur based on the shortest distance from the recapture site by using an exact multinomial test to determine whether the migratory pace was biased towards any of the four proposed speeds. If the migratory pace was not randomly distributed among the four speeds, we used post hoc exact binomial tests to determine the probability of the tag–recapture matching one of the four proposed speeds (4 tests: Bonferroni-corrected α=0.0125).

For analysis of energy expenditure, we estimated the total amount of fat (mg) using the temperature-dependent SMR estimate from the non-linear least squares (NLS) regression, and the method used by Gibo and Pallett (1979). For our method, we calculated the hourly energy expenditure and then summed the results to obtain the fat used based on the number of hours between tagging and recapture. For estimating energy expenditure from the method of Gibo and Pallett (1979), we used their method and multiplied the number of hours from tag and recapture by 0.13 mg fat h⁻¹ to obtain the estimate of fat used under the assumption from that study. The estimated amount of fat reserves used during the migratory period for each individual butterfly was then compared using a paired Wilcoxon test.

Last, we used the individual tagging locations (n=108) to extract the estimated DEE (mg fat day⁻¹) from 1961 to 2018 for each month during the monarch migration (i.e. August–October, n=108). We do assume that these locations are representative of the monarch butterfly migration through time as they were all captured on different days and years. These data were then averaged for each month during migration each year and then regressed (ordinary least squares, OLS) over the 57 year period to determine whether there was a significant increase in the slope estimate (β) for DEE. We also compared DEE between August, September and October using a Kruskal–Wallis test for each independent year and a Dunn test for pairwise comparisons within each year.

All data were analyzed and simulated using program R (v3.6., http://www.R-project.org/) and code for generating the movement path, running the model monarch simulation, and estimating energy expenditure can be found in the supplemental code (doi:10.5281/zenodo.7771956).

Between 10 and 30°C, we found that metabolic rate (in microwatts) differed with test temperature [ANCOVA: F_2,81=46.18, P<0.001, partial eta squared (η_p²)=0.53] but not life history stage (F_2,81=1.59, P=0.20, η_p²=0.04) or their interaction (F_4,81=2.08, P=0.09, η_p²=0.09). From 5th instar caterpillar to adult, individuals had a 58.1% decrease in mass with an average 5th instar mass of 1.22±0.2 g and adult mass of 0.51±0.01 g. Female monarchs reared under laboratory conditions were found to be primarily in reproductive diapause (85%) up to 1 month after eclosion. Furthermore, comparison of adult monarch metabolic rates from previous work and our current study showed no difference between 1982 and 2021 (two-way ANOVA: F_1,46=0.01, P=0.91, η_p²=0.0003); the only differences in metabolic rate occurred between the temperatures tested (two-way ANOVA: F_3,46=8.92, P<0.001, η_p²=0.37) (Fig. 2).

Computer-simulated monarch butterflies had the highest precision (±s.e.) to their known recapture location when their migratory pace was set to 3.9 km h⁻¹ (n=17; average offset from recapture location 64.38±15.11 km), followed by 7.0 km h⁻¹ (n=5; average offset from recapture location 225.61±83.32 km), 14 km h⁻¹ (n=2; average offset from recapture location 116.26±21.26 km), then 18 km h⁻¹ (n=1; average offset from recapture location 540 km). We found that the speeds of these 25 butterfly simulations from tag–recapture data were non-randomly distributed across the four paces (exact test for multinomial, P<0.001). Using post hoc binomial tests (Bonferroni corrected α=0.0125), we found that there were significantly more butterflies that had a pace of 3.9 km h⁻¹ (n=17, P<0.001, Cohen's g=0.17) while the remaining butterflies were consistent with the number of expected individuals to move at those paces (7 km: n=5, P=0.83, Cohen's g=0.29; 14 km: n=2, P=0.06, Cohen's g=0.42; 18 km: n=1, P=0.02, Cohen's g=0.46). We found significantly higher estimates of energy expenditure using our method that included temperature-dependent SMR compared with the method used by Gibo and Pallett (1979) (paired Wilcox: V=251, P<0.01, effect size r=0.59; Fig. 3) for the tag–recapture data. Based on our estimate, a 500 mg butterfly would have around 750 h of flight without refueling while the constant consumption estimate is around 880 h (Fig. 3), indicating a significant decrease in flight time due to the additional costs of SMR from ambient temperature. We also found that by changing the ratio of powered to gliding flight from 15% and 85% (average 51.9±9.0 mg fat across 24 simulated monarch tag–recaptures) to 30% and 70% (average 79.5±13.9 mg fat across 24 simulated monarch tag–recaptures), there was an average 53.5±1.2% increase in fat consumed per tag–recapture monarch.

Over a 57 year period, we found a significant increase in SMR over time for monarchs during autumn migration linked to increasing ambient temperatures related to climate change at location-specific tagged monarchs. Overall, there was a significant increase in the OLS regression slope estimate for DEE based on tagging locations during August and September, but not October (Fig. 4; Table S2). However, analyzing the DEE in 30 year segments showed significant increases only during September and October in recent years (Table S2). Within each year, the estimated DEE was significantly higher in August than in September or October; in most years, DEE in September was higher than that in October (Table S3). Using a simple moving average, we found that the forecast showed a continuing increase in the baseline DEE across all three months (Fig. 4A).

The increasing trend for monthly average maximum temperature from 1961 to 2018 has become more significant with the recent 30 year segments in September and October, highlighting the effect of extreme temperatures on energy management for monarchs during daytime migration (Fig. 4C; Table S2). This trend was also similar for the monthly average minimum temperature at the capture location for monarchs, which also highlights the role of increasing monthly average minimum temperatures at night in increasing energy demand when monarchs should be resting (Table S2). Previous work assuming a constant energy expenditure estimated that monarchs would consume 1.56 mg fat day⁻¹. Based on the point-estimate of the forecast for energy expenditure, there was an increase from 2.44 to 2.60 mg fat day⁻¹ in August, 1.83 to 2.03 mg fat day⁻¹ in September, and 1.45 to 1.56 mg fat day⁻¹ in October. While there was a small increase from 1961 to 2040, approximately 0.11–0.20 mg fat day⁻¹, there was an increase in fat consumed depending on the month of migration, being highest in August, followed by September. However, October falls within the previous estimate. When we used the estimated average temperature increase based on the StableClim dataset, there was a similar increase in fat consumed per day (0.20 mg fat day⁻¹) to the upper end of the SMA forecast (0.11–0.20 mg fat day⁻¹). However, using month-specific ensemble estimates of average temperature from 1960 to 2040, we found that fat consumption in August increased from 2.54 to 2.86 mg fat day⁻¹, that in September increased from 1.61 to 1.99 mg fat day⁻¹, and that in October increased from 1.52 to 1.82 mg fat day^−1­­. Based on the SMA forecast, if we assume a 1 week delay using SMR for the hypothetical 500 mg butterfly with 115 mg of fat stores in the early stages of migration, this can result in an increase of 1.89 mg of fat (1.6% of energy stores) in September to 6.18 mg of fat (6.15% of energy stores) in August from the constant estimate in 1961.

Migration is an energetically taxing phenomenon and species will experience major challenges on their journey under current climate change trajectories ranging from phenological shifts to exacerbated effects of temperature on metabolic costs (Seebacher and Post, 2015; Radchuk et al., 2019). Our study found support for our hypotheses that temperature-dependent physiological processes (e.g. SMR) reduced energy reserves faster than previously estimated (Gibo and Pallett, 1979) for monarchs and that increasing temperatures as a result of climate change can exacerbate baseline DEE. First, we found that adult monarchs in reproductive diapause have around a 60% decrease in body mass from their 5th instar (Table 1). Then, for monarchs that were estimated to travel in a straight line, overall moderate pace (3.9 km h⁻¹) and the inclusion of SMR with DEE showed that temperature increases energy expenditure, as fat, by 0.02 mg h⁻¹, resulting in a potential decrease in time spent migrating through a reduction of energy reserves. Finally, over a 57 year period, we found an increasing trend in estimated DEE that aligns with increasing monthly average maximum and minimum temperatures associated with climate change. While climate change has been shown to have significant impacts on the timing and duration of animal migrations, two potential scenarios arise from our model. First, there is an increase in energy expenditure during autumn migration for monarchs. Over time, the reduction of fat reserves necessary for migratory monarchs to arrive and survive at their overwintering conditions will be in jeopardy, especially as energy reserves retained by the butterfly during overwintering are vital for the remigration north during the following spring (Brown and Chippendale, 1974). Second, there is the possibility of a phenological shift in the start of migration (Culbertson et al., 2021) as later thermal conditions are less costly or individuals that migrate later have a higher chance of survival. Although our model has limitations due to the assumptions of model movement and estimates of energy expenditure, our results reinforce the complexity of animal migration and emphasize the benefit of viewing migration in its entirety as a syndrome consisting of a suite of related physiological, behavioral and ecological traits (Dingle, 2014).

Monarchs have had a past with yearly fluctuations in temperatures and will have a similar future, as conditions begin to become more extreme under climate change, specifically in the central portion of their migratory flyway, which will likely become an extreme heat belt by 2053 (Wilson et al., 2022). Compared with previous work estimating metabolic rate, monarch metabolism as a function of temperature appears not to have changed after the 40 years since it was previously recorded over multiple temperatures under laboratory conditions (Chaplin and Wells, 1982). Thus, increases in minimum and maximum ambient temperatures due to climate change will increase the baseline energy expenditure not only during daytime migration but also during night-time rest, when energy use should be at its lowest. While we investigated estimated energy use during August, September and October of the autumn migratory season, many of the monarchs likely experience the temperatures and energy demands of September and October only, as the temperatures during August apply to migratory monarchs between latitudes 45°N and 50°N (www.monarchwatch.org). Regarding potential shifts in migratory phenology, there are likely local differences in conditions that could drive later migration (Culbertson et al., 2021; Ethier and Mitchell, 2023). On a finer temporal scale, such as migratory season, the weather conditions that promote or inhibit monarch migratory movement could possibly lengthen the time necessary to make it to their overwintering site. Based on the tag–recapture data, most monarchs flew at around 4 km h⁻¹; however, there are instances of monarchs traveling greater distances in similar or shorter time periods, as evidenced by an individual that traveled at a pace of 18 km h⁻¹ in this study and from unpublished speed estimates (Howard and Davis, 2015). While the model assumed a straight-line path, monarchs could be using assisted flight from prevailing winds to boost the distances traveled, which may also include deviations from this path. There is also the reality that monarchs are traveling at variable speeds at finer temporal scales (e.g. minute versus hour resolution), but there is a gap in our knowledge as to the extent of variable speeds under different meteorological conditions and available historical data at fine temporal resolutions.

We found using our model simulation with biologically relevant parameters, that monarchs are using more energy during autumn migration than previously estimated (Wensler, 1977; Gibo and Pallett, 1979). A reduction of fat reserves during migration linked to temperature indicates that individuals will have to potentially stop more frequently to refuel as temperatures begin to increase. Our simulations also show increases in lipid use whether using monthly ensemble estimates (i.e. StableClim) or forecasting based on average minimum and maximum temperature (i.e. WorldClim2). Brower (1985) emphasizes the dynamic process between nectar feeding and lipid conversion, as by the end of the overwintering in Mexico, nectaring becomes more plausible as a result of blooming plants (e.g. Asteraceae). However, previous work has highlighted the importance of nectaring for monarchs during migration (Brower et al., 2006; Hobson et al., 2020) and our study provides evidence for the need to ensure appropriate connectivity between stopover sites for migratory monarchs given the increased energy demands from increasing temperatures during migration. While our study focused on monarchs, other multigenerational migratory species (e.g. Vanessa cardui; Stefanescu et al., 2013) that have higher increases in metabolic rate when flying (e.g. Zebe, 1954) are likely also experiencing intensifying effects of climate change. While the thoracic temperature in lepidopterans during flight can be considerably higher than ambient temperature (Bladon et al., 2020), increases in ambient temperature increase baseline energy expenditure and thus overall energy use, even if the flight strategy taken by a species uses less energetically taxing movement, such as gliding (Rankin and Burchsted, 1992).

Changes to developmental conditions, energy management and reserves could be a selective pressure for migratory animals to drive for more efficient morphology to facilitate function. For example, wing traits that promote gliding, such as wing aspect ratio, may evolve as a result of such pressures (Satterfield and Davis, 2015; Le Roy et al., 2019a). Alternatively, natural wing deterioration may act as a selective pressure on wing morphology (Le Roy et al., 2019b). In bird species, morphological adaptations have been shown to predict accumulated fuel loads, whereby species that have more flapping flight accumulate more fat stores compared with those with gliding flight (Vincze et al., 2019). Migratory birds must also balance energetically demanding events, such as molting, through temporal separation to avoid drastically reducing energy reserves during a single event (Tonra and Reudink, 2018). In lepidopterans, different morphological adaptations of the wing can impact the flight height, speed and proportion of flight spent gliding, such as the forewings dictating flight capability and the hindwings serving as an extended airfoil during gliding flight (Le Roy et al., 2019a). It is possible that accumulated wing damage via natural deterioration could impact flight performance if the deterioration occurs on the forewing rather than the wing margins (Le Roy et al., 2019b). Given that migratory monarchs rely primarily on gliding flight (Gibo and Pallett, 1979), changes to wing morphology resulting from anthropogenic stressors could also negatively impact flight performance of migrants, thus reducing migratory success. Diet and temperature in monarchs have also been shown to affect wing morphology and flight ability (Soule et al., 2020). Diet impacts the wing shape whereby non-native tropical milkweed (Asclepias curassavica) induces shorter, wider forewings and native milkweed (Asclepias incarnata and A. syriaca) induce longer, narrower forewings which are better for gliding flight – a less energetically costly form of flight. High temperatures also reduce the flight ability of monarchs by reducing their time spent flying during a given day, which results in shorter migratory flight durations.

While fat stores have more energy per unit mass than other oxidative fuel sources (e.g. protein or carbohydrates; McWilliams et al., 2004), understanding how often monarchs refuel fat stores during migration is important not just for conservation management decisions but also for how monarch migration will be impacted in the future under climate change (Svancara et al., 2019). In monarchs, powered flight is mostly fueled by lipids, and these same energy stores are also used for overwintering (Niitepõld et al., 2022 ). During the beginning of migration, adult monarchs have lower lipid concentrations, around 23% of their body mass, but increase these in northern Texas and Mexico before arriving at their overwintering site with lipid concentrations upwards of 500% (Brower et al., 2009). It is plausible that these increases in lipid mass could require more powered flight and thus increase energy expenditure if temperatures increase. Increasing powered flight from 15% to 30% of the time showed a 53% increase in lipid use. There is evidence to suggest that the costs of migration for the eastern North American population of monarchs since the 1960s may be impacted by not only the increased temperatures exacerbating energy demands but also potentially by the shifting resources in changing landscapes. For instance, at the monarchs’ overwintering site, the less available forest cover has been shown to increase stress and reduce survival (Nicoletti et al., 2020). Overwintering monarchs need energy reserves to survive their overwintering stage (Alonso-Mejia et al., 1997), and then require initial fuel for re-migration north to locate newly emerging plants during the spring, although refueling could be possible on the spring remigration (Brower, 1985). Chaplin and Wells (1982) found that a 500 mg overwintering monarch consumed around 42.2 mg of fat over a 61 day period, although they did note observations of monarchs flying on clear sunny days and drinking at local water sources during the wintering period. However, the compound effects of natural wing deterioration during the initial migration, changes to wing morphology from diet, and potentially over-depleted energy reserves highlight the need to understand migration throughout the landscape from the perspective of the animal's physiology and morphology.

Long-distance migration in other other animals, such as birds, requires incredible phenotypic plasticity to prepare for the journey (Piersma and Van Gils, 2011). The incorporation of energy reserves and fuel utilization shows the added benefit of including an organism's physiology in terms of our understanding of migration (Sapir et al., 2011). Given the deleterious effects of climate change, it is important that the placement of stopover sites is in areas that migratory monarchs frequent (e.g. Monarch Waystation Program; www.monarchwatch.org); however, urbanized landscapes could exacerbate energy demands as a result of higher temperatures from the urban heat island effect (e.g. Merckx et al., 2018; Magura and Lövei, 2020).

We highlight several potential enhancements to our approach that would utilize different data types or experimental field data collection. One improvement would be the use of finer-resolution micro-meteorological data from the hour-interval data that we used in this study. For instance, while it is well established that tailwinds and headwinds can improve or impede flight speed, the hourly variation in the weather station data limited our application of these parameters to the model effectively because of how variable these winds may be during the hour and at different elevations. Ground-truthing the model simulation estimates will show the potential variation in flight strategies or environmental contributions to movement propensity. Another avenue of potential inquiry is the role of heterogeneous landscapes in movement, direction and speed during migration. While the benefit of using monarchs is that their migratory orientation response is relatively fixed to a southwest directionality, we do assume that the tag–recapture path is a straight line in our model, as well as a constant speed. The partial flight of butterflies, or variable flight speeds, during each hour could also contribute to distance moved as we assumed a constant hourly movement when conditions were permissible for flight. How long a migrant refuels prior to departure depends on factors such as energy loss during previous flight (Schmaljohann and Eikenaar, 2017) or increased temperatures altering the cost of flight capability, which relies on fuel mobilization, transport and metabolism (Rankin and Burchsted, 1992). However, to model refueling, there would need to be locations for stopover incorporated into a model, which was not possible for our dataset based only on when the tag–recaptures occurred. There is the potential to incorporate latitudinal effects of lipid mass increase on estimated lipid concentrations in tandem with migratory pace, as our study only focused on early stages of migration prior to entering northern Texas and Mexico. Natural observations would also be required for determining how often and how long under certain meteorological conditions monarchs will stop and stay to refuel. Other migratory lepidoptera, which can be tracked by radar (e.g. Chapman et al., 2010), could have a similar approach applied by simultaneously obtaining micro-meteorological conditions to assess large-scale migration, movement, directionality and estimated energetics. Future modelling efforts would also benefit from incorporating effects of high temperature on migration, such as the impact on orientation or search response for cooler microclimates. Nonetheless, the incorporation of the butterfly's physiology in the model improves our understanding of estimated migratory energetics in a small, long-distant migrant like the monarch butterfly.

Our study highlights the significant benefit of tag–recapture data for migratory invertebrates in tandem with the wealth of data in the literature related to energetics and micro-meteorological conditions. This technique of modelling the migratory pace and energy expenditure of monarchs can be combined with new approaches, such as the ‘catch–test–release’ method, to improve recapture probability of migrants downstream from the tagging location, to test the effect of micro-meteorological conditions at a finer temporal resolution, predict movement propensity, and calculate actual energy expenditure between the tag–recapture locations for comparison with the predicted energy expenditure across vast heterogeneous landscapes (Parlin et al., 2021). As climate change continues to create extreme thermal conditions, the importance of understanding how temperature-dependent physiological processes in small ectothermic animals are affected by extreme temperatures becomes increasingly important.

We thank O. Leek, J. Espino, W. Gray, M. McHugh and J. Malvicino for accessing weather station data for the tag–recapture monarchs. We also thank J. C. de Roode for the use of extra monarch eggs and the facilities at the University of Cincinnati and the staff at the University of Cincinnati Center for Field Studies for assistance with maintaining and capturing wild monarch butterflies. We are also appreciative of the comments and feedback from two anonymous reviewers as it greatly improved the applicability of the model. We also thank the Monarch Watch program for the use of their tagged individual.