Flight energetics of barn swallows across speeds: high aerodynamic efficiency, but low energy conversion efficiency Authors: Pablo Macías-Torres, Sonja I. Friman, L. Christoffer Johansson, Anders Hedenström Published: 2026 ABSTRACT ABSTRACTBird flight conversion efficiency (η) determines the amount of metabolic power (Pmet) transformed into mechanical power (Pmech) to remain aloft, and it is central to understanding animal flight energetics, yet it remains difficult to quantify. Using 13C-labelled sodium bicarbonate (NaBi) and particle image velocimetry (PIV), we measured Pmet and Pmech, respectively, in un-instrumented barn swallows (Hirundo rustica) flying across 7.8 to 13.6 m s−1 in a wind tunnel. Both Pmet and Pmech showed the expected U-shaped relationship with airspeed. Whole-animal η showed a shallow peak at 7% at 11 m s−1. Partial efficiency (ηp), estimated from horizontal and 4 deg climbing flight, was 27% (95% confidence interval CI: −2 to 56.1). Compared with thrush nightingales previously studied under identical conditions, barn swallows exhibited higher weight-specific flight Pmet but lower lift-specific Pmech, reflecting a higher basal metabolic rate but greater aerodynamic efficiency in barn swallows. Barn swallows showed lower peak and flatter η curve across speeds compared with that of the nightingales, suggesting a trade-off between flight energetics and flight speed: higher efficiency peak over a narrow speed range (thrush nightingales) versus lower, more consistent efficiency across a broader speed range (barn swallows). Whereas nightingales fly mainly during migration, barn swallows spend much of the day airborne and use a wider range of flight speeds, potentially favouring versatility over a narrow η peak. Although η has been studied in few species, our results suggest that η is both species and speed dependent, probably shaped by species-specific ecological demands. INTRODUCTION Flight is both an effective (high-speed movement) and efficient (low cost of transport) mode of locomotion, but quantifying the energy required to fly in animals is a challenge (Bishop and Guglielmo, 2022; Schmidt-Nielsen, 1972). Flying animals sustain flight by flapping their wings to accelerate the surrounding air, generating both lift and thrust, while adding kinetic energy to the air. This is fuelled by metabolic energy input, which includes the energy expended by the flight muscles and other body systems involved in sustaining flight. The ratio between these two quantities defines the energy conversion efficiency of flight (η), and quantifies how much chemical energy is shunted into mechanical work, needed to remain aloft, by the flying animal (Hedenström, 2025). Until recently, flight energy conversion efficiency, a critical component in understanding the energetics of flight, has not been possible to quantify directly (but see Currie et al., 2023; Hedh et al., 2020; Macías-Torres et al., 2025). This is mainly due to the technical challenges of accurately capturing both metabolic and mechanical power (Pmet and Pmech, respectively) in freely flying animals (Hedenström and Lindström, 2017). Traditional methods often relied on indirect estimates or flight models with simplified aerodynamic approximations (e.g. Pennycuick, 2008). Recent methodological advances have allowed for measuring both Pmet (Hambly et al., 2002) and Pmech (von Busse et al., 2014), in un-instrumented flying animals under near-natural flight conditions in the same individuals in wind tunnels. In this study, we quantified whole-body flight conversion efficiency in barn swallows (Hirundo rustica) using in vivo measurements of both Pmet and Pmech. Pmet was assessed via the 13C-labelled sodium bicarbonate (NaBi) method, while Pmech was calculated as aerodynamic power using particle image velocimetry (PIV) which captures the bird's wake during flight. Additionally, we estimated partial efficiency (ηp), by having the birds fly horizontally and at a climbing angle through tilting the wind tunnel. This approach was used to derive the first-ever estimate for flight efficiency, which has remained in use until now and is based on conversion efficiency being related to the relative change in Pmet and Pmech required for a known climbing flight angle (Bernstein et al., 1973; Tucker, 1972; Pennycuick, 2008). Barn swallows are highly aerial birds, spending much of their daily activity in flight while foraging, commuting and migrating (Turner, 2006). Barn swallows differ from other passerines by their wing morphology, having relatively long, pointed and comparatively high aspect ratio wings, a wing index associated with reduced induced drag and high lift production (Bruderer et al., 2001). Their wing design allows barn swallows to use gliding phases effectively between active flapping bouts, contributing to their characteristic intermittent flight style, compared with continuous or flap-bounding flapping flight in most other passerines (Bruderer et al., 2010, 2001; Tobalske, 2010). Given the barn swallow's morphological and behavioural adaptations for aerial life (Turner, 2006), we hypothesised that this species will exhibit a relatively high energy conversion efficiency. We further predicted that, consistent with prior flight energy conversion efficiency measurements in migratory songbirds, to achieve low costs of transportation and migration (Macías-Torres et al., 2025), energy conversion efficiency in barn swallows will show a peak at intermediate flight speeds, reflecting optimisation for ecologically relevant speeds. List of symbols and abbreviations BMRbasal metabolic rateCOTcost of transport F eff effective wingbeat frequency g gravitational acceleration k c isotope elimination rate m body massNaBi13C-labelled sodium bicarbonate P mech mechanical flight power P mech * lift-specific mechanical flight power P met metabolic flight power P met * weight-specific metabolic flight power P mp flight power at minimum power speed P rec post-flight recovery metabolic powerPIVparticle image velocimetryRMRresting metabolic rate U airspeed U mp minimum power speed U η,max airspeed at maximum whole-animal energy conversion efficiencyηwhole-animal energy conversion efficiencyηfmflight muscle conversion efficiencyηfm,recoveryPrec based flight muscle converison efficiencyηmaxmaximum whole-animal energy conversion efficiencyηppartial energy conversion efficiencyθwind-tunnel inclination angle MATERIALS AND METHODS We captured 12 barn swallows, Hirundo rustica Linnaeus 1758, at lake Krankesjön, South Sweden (55°42′N, 13°27′E), on 4 September 2023. Six birds, all juveniles, adapted to the conditions in captivity and successfully progressed in their flight training in the wind tunnel, while the other six birds were released. The birds were kept in a communal aviary, with a day:night light routine of 13.5 h light:10.5 h dark, simulating local daytime conditions at the time of capture. Water and food were provided ad libitum in a half-cylinder plastic feeder suspended from the ceiling, and consisted of live mealworms (Tenebrio molitor) and frozen buffalo worms (Alphitobius diaperinus). Mean body mass upon capture was 21 g (range: 18.7 to 23.6 g). During the experiments, the average body mass was 17.9 g (range: 15.7 to 20.7 g). Flight power measurements were obtained for flight speeds ranging from 7.9 to 13.6 m s−1 for Pmet and between 7.8 and 13.1 m s−1 for Pmech. All flight experiments were performed during the daytime. We used dimmed light conditions inside the wind tunnel to minimise the birds' stress, preventing them from seeing the flight operators and to facilitate their recapture after each flight experiment. Additionally, a thin plastic net (mesh approximately 15×15 mm; 0.1 mm chord thickness) was placed upstream to prevent birds from entering the settling chamber, where speed is reduced in the contraction section in front of the wind tunnel's testing section (Pennycuick et al., 1997). Barn swallows were removed from the cage in the morning and fasted for at least 4 h before a flight experiment to ensure post-absorption state and promote fat-oxidising metabolism (Jenni-Eiermann and Jenni, 2012). All birds were kept together in a dark cage during the fasting period, and each bird was returned to the aviary after its flight experiment. Prior to each experiment, the birds were weighed (±0.01 g). After the completion of all the experiments, the birds were released back into the wild and flew away seemingly unaffected. Permission for the experiments was approved by the Malmö-Lund Animal Experimentation Ethics Board (5.8.1817907/2023). Measuring flight Pmech – PIV To visualise the flow behind the bird, we used a stereo-PIV setup similar to that in Johansson et al. (2018) with four high-speed cameras placed downstream of the intended flight location of the bird (Nova R2, Photron Deutschland GmbH, Reutlingen, Germany) and a pulsed laser generating a light sheet (LDY304PIV laser, Litron Lasers Ltd, Rugby, UK) oriented perpendicularly to the wind direction, also downstream of the intended flight location. The dual pulsed laser was operated at 720 Hz and the camera array recorded the illuminated aerosol particles (di-ethyl-hexyl-sebacate, 1 μm diameter) that were previously seeded into the wind tunnel air. Processing of the recordings and vector extraction were done using DaVis (v.10.2.1, LaVision GmbH, Gottingen, Germany). Only three of the four cameras could be used for the processing because one of the cameras had been disturbed during the experiments. The stereo-PIV processing was performed, as described in the following steps, using the GPU option. The velocity vectors were calculated with a window shift in the stream-wise direction of 4 pixels for the initial pass, using an initial box size of 96×96 with two passes, followed by a final box size of 32×32 with three passes, with an overlap of 50%. Between passes, a twice-applied universal outlier detection was used: vectors that deviated more than 2 standard deviations (s.d.) from their neighbours in a filter region of 5×5 were replaced by the next-best vectors if those deviated by less than 3 s.d. The vectors were filtered by correlation value >0.25, and a vector validation was applied where vectors were accepted if the stereo reconstruction error was below 1 pixel. Finally, anisotropic denoising was applied, with the setting ‘weak’. The final vector volume size was 322×343 mm with a vector spacing of 2.4 mm (∼4.2 vectors cm−1), thus yielding 135×140 vectors. Pmech was defined as the rate at which kinetic energy was added to the wake for an integer number of wingbeats, assuming inertial power to be insignificant (Pennycuick, 2008; Pennycuick et al., 2000). Pmech was estimated from the PIV-generated velocity fields of the bird wakes, following the description in Johansson et al. (2018) and Macías-Torres et al. (2025), with the following changes. We used the ‘automask’ described in Johansson and Henningsson (2021) to isolate the wake vorticity, which was then used to reconstruct the velocity field beyond the wind tunnel limits to a square size of 2.4×2.4 m using the Helmholtz–Hedge decomposition (von Busse et al., 2014). The original measurements inside the mask were reinserted into the reconstructed flow field before calculating the kinetic energy in the flow. This method effectively removes noisy background flow while maintaining the vorticity associated with the animal wake. In addition to Pmech, vertical (lift) and horizontal (thrust) forces were estimated using the masked vorticity as applied in Johansson (2024). Calculations were performed using custom written code in MATLAB 2024b (MathWorks Inc.). In barn swallows, the flight kinematics deviate from continuous flapping and often include brief upstroke interruptions between wingbeats (Bruderer et al., 2001; Park et al., 2001), which may lead to increased forces and power output during the active beats to compensate for these pauses. Unlike previous studies from our lab (Hedh et al., 2020; Macías-Torres et al., 2025), we did not apply a filtering step to exclude PIV measurements where the vertical force deviated more than ±20% from the bird's weight. This decision was based on the fact that our analysis focused solely on active wingbeat sequences, as the pauses in between wingbeats were rarely recorded in our PIV set up. Had we applied the filter, six measurements would have been excluded, indicating that our estimates of Pmech, if anything, are probably biased upward rather than being underestimated. To contextualise the limitation of estimating Pmech from active wingbeats only, we report the wingbeat frequency across airspeeds in the sequences used for our Pmech analysis. We then compared these values with the effective wingbeat frequencies reported by Bruderer et al. (2001) for barn swallows flying horizontally in a different wind tunnel. In contrast to our measurements, Bruderer et al. (2001) estimated wingbeat frequency over 20 s intervals, thereby incorporating both active flapping phases and pauses between wingbeats. Measuring flight Pmet – NaBi Pmet was measured using the 13C-NaBi technique, following the same protocol described in Macías-Torres et al. (2025). This method uses the elimination rate of previously injected isotopic carbon as a proxy for CO2 production when direct respirometry measurement of CO2 is not available, such as when flying freely in a wind tunnel. First, an isotopic carbon solution (0.56 mol l−1 NaH2 13CO3; Sigma-Aldrich) was injected subcutaneously into the bird. The bird was then transferred to a darkened respirometry chamber of 1.5 l volume through which CO2-free air (21% O2, balance N2) from a gas cylinder (Linde Gas AB, Solna, Sweden) passed at a constant rate of 2 l min−1. The airstream was subsampled by an isotope analyser (PICARRO – G2201-I, Santa Clara, CA, USA), which measured the enrichment of 13C (δ13C; parts per thousand difference from the standard) and the CO2 concentration produced by the bird's respiration. After the injection, a peak in the isotopic concentration occurred and a stable depletion followed (Fig. S1). After 5–10 min of stable isotopic washout, the bird was extracted from the chamber and was allowed to fly in the wind tunnel at a pre-set speed. After a flight of around 2 min (mean 2.1 min, range 1.8–2.4), the bird was captured and transferred back to the respirometry chamber, where isotope levels and CO2 production were again measured. Time stamps of each step were taken. All measurements were kept under 30 min to follow the assumption of single pool kinetics in small animals (Hambly and Voigt, 2011). The pre- and post-flight δ13C data were natural log transformed and fitted with a linear regression. The regression slopes represent the log converted isotope elimination rate (kc; δ13C min−1) in the bird. We used the R-package LolinR (Olito et al., 2017) to select the best linear fit regression and avoid biases due to arbitrary data selection in the δ13C values. After fitting the linear regressions, pre- and post-flight kc were extrapolated until the time of flight start and flight termination, respectively, so a value of δ13C was inferred for the flight start and flight stop timestamps. The isotopic elimination rate during flight (flight kc) was calculated as the slope between the calculated pre- and post-flight δ13C values at the flight start and flight stop time stamps. As the isotope elimination rate is proportional to the amount of CO2 produced (Hambly and Voigt, 2011), flight CO2 was calculated by multiplying the flight isotopic elimination rate (flight kc) by the ratio between the post-flight CO2 production and the post-flight isotope elimination rate. This estimated flight CO2 production was then converted to Pmet (W) following the protocol described in Macías-Torres et al. (2025). All preliminary NaBi experiments (n=61) were visually inspected and any data showing unexpected patterns in the pre- and post-flight enrichment curves (i.e. a secondary enrichment in the isotope curve leading to a double peak), as well as irregularities in CO2 measurements were excluded from further analyses. In total, six measurements were discarded by this filtering. Furthermore, we assumed that a bird in flight consumes energy at a higher rate than in pre- and post-flight, during which birds remained mostly motionless inside the respirometry chamber. We therefore applied an additional criterion: estimated CO2 production during the flight had to be higher than the values directly measured before and after flight. Based on this criterion, nine samples were discarded. This resulted in a dataset of 46 NaBi experiments used to analyse horizontal flight Pmet. A second dataset consisted of nine NaBi samples collected from birds flying at a positive incline of 4 deg, which were used to calculate partial efficiency (see below). Outlier filtering We screened for extreme values in flight Pmet using the modified Z-score of Iglewicz and Hoaglin (1993), applied for binned airspeed centred at half-integers (U0.5). For each binned airspeed, we computed the bin median and the unscaled median absolute deviation (MAD): (1)MAD=mediani(|xi−x~|). For each observation we then calculated the modified Z-score: (2)Mi=(0.6745×(xi−x~))/MAD and flagged observations as outliers when Mi>3. Flagged observations were excluded from subsequent analyses. This screening was applied once and resulted in the removal of one outlier (2.92 W at 9 m s−1; Mi=3.4); no further observations exceeded the threshold thereafter. Modelling power After outlier removal (final Pmet n=45 from 6 birds; Pmech n=27 from 6 birds), we fitted a model for flight Pmet and Pmech as well as for the weight-specific flight power as a function of airspeed (U). We fitted flight Pmet and Pmech and weight/lift-specific metabolic/mechanical power (Pmet* and Pmech*) with a simple linear and a quadratic polynomial in U. Models were fitted both with and without random intercepts for individual birds to control for the effect of between-bird variation. For Pmech and Pmech* models, the number of wingbeats was included as weighted in the model contributing to that estimate, assuming measurement precision increases with sample size. Model decisions were guided by fixed-effect significance, explained variance and the likelihood ratio test (see Supplementary Materials and Methods for details). Whole-animal energy conversion efficiency Whole-animal energy conversion efficiency was calculated by dividing Pmech estimates by Pmet estimates from the best fitted model at a given airspeed (U): (3)η(U)=Pmech(U)Pmet(U). We then determined the maximum η(U) over the overlapping observed speed range (U) of the two datasets, ηmax, and its corresponding speed Uη,max. Partial efficiency We calculated the partial energy conversion efficiency (ηp) during flight following the approach of Tucker (1972) and Bernstein et al. (1973). ηp was defined as the ratio between the mechanical power change (ΔPmech) attributable to a tunnel inclination (4 deg) and the corresponding change in metabolic power (ΔPmet): (4)ηp=ΔPmechΔPmet, where the mechanical power change was calculated as: (5)ΔPmech=WbV=Wbsin(θ)U, where Wb=mg is the body weight, m is the mean body mass of all birds measured in kg (mean 0.0179 kg), g is the gravitational acceleration (9.81 m s−2), θ is the tunnel tilt angle (4 deg) measured with a clinometer, and U is the airspeed (11.4 m s−1). We measured Pmet during horizontal flight and climbing flight using the NaBi method (see above). Because the dataset consisted of limited samples per bird and some birds showed small or negative within-individual differences in Pmet between horizontal and inclined flight, partial efficiency (ηp) was estimated at the population level using a mixed-effects model. We fitted a linear mixed-effects model with flight power as the response variable, inclination as a fixed effect and a random intercept for bird individuals. The change in flight metabolic power (ΔPmet) was estimated as the difference between the estimated marginal means in Pmet flying at 0 deg and 4 deg (climbing flight) from the linear mixed model: (6)ΔPmet=P¯metθ=4−P¯metθ=0. Uncertainty for ηp, as the standard error (s.e.), was obtained using the delta method: s.e.(ηp)=ηp·[s.e.(ΔPmet)/ΔPmet]. The 95% CI was calculated as ηp±1.96s.e.(ηp). Muscle efficiency We estimated flight muscle conversion efficiency as described in Morris et al. (2010). Flight muscle conversion efficiency was the ratio between Pmech output to Pmet input, minus the basal metabolic rate (BMR). To account for overhead respiratory and circulatory costs, a factor of 1.1 was applied (Tucker, 1973). Hence, muscle efficiency (ηfm) was calculated as: (7)ηfm(U)=1.1(Pmet(U)−BMR)Pmech(U), where Pmet and Pmech are the metabolic and mechanical power estimates from the best-fitting models at a given airspeed in watts. BMR (W) represents the mean basal metabolic rate across individuals, predicted from body mass (m, in grams), using the allometric equation of McKechnie and Wolf (2004): (8)log(BMR)=−1.461+0.669log(m). In addition to the allometrically predicted BMR, we calculated flight muscle conversion efficiency by subtracting the mean post-flight recovery Pmet (Prec, see below) from the flight Pmet: (9)ηfm,recovery(U)=Pmech(U)(Pmet(U)−Prec), where Prec is the mean post-flight recovery metabolic power across individuals (see below). Here, no overhead respiratory and circulatory factor was used as Prec was measured while birds were performing these vital functions. Pmet while not flying In order to compare flight Pmet with non-flight Pmet, we estimated the Pmet of barn swallows while not in flight, during the recovery period right after a flight experiment. We refer to this measurement as post-flight recovery metabolic power (Prec). Following each trial, the bird was placed in a darkened respirometry chamber, and its metabolic rate was quantified from direct CO2 measurements using the isotope analyser system. The mean value of CO2 production rate was estimated from the best-fitting linear model using the R-package LolinR (Olito et al., 2017) and converted to Pmet using the same calculation as applied to the flight power measurements. We refer to this measurement as Prec, rather than resting metabolic rate (RMR), because the birds were not measured under standard resting conditions (Ellis and Gabrielsen, 2019). Specifically, the birds had been flying just a few seconds beforehand and had also been handled before being placed in the chamber. Thus, this value probably includes additional costs associated with recovery and handling stress. We used the post-flight recovery phase rather than the pre-flight phase because pre-flight metabolic rates were consistently higher (as measured by CO2 production; Fig. S1), probably due to a higher stress level caused by longer handling, injection and the preparation period immediately before a flight experiment. In contrast, just before the post-flight phase, the birds were captured while flying and rapidly transferred to the respirometry chamber, which resulted in a lower metabolic rate during the post-flight than during the pre-flight phase. This is also the reason why the post-flight, rather than the pre-flight phase is used when estimating flight Pmet (Elowe et al., 2023; Friman et al., 2024; Macías-Torres et al., 2025). Therefore, although post-flight Prec should not be interpreted as true RMR because there may be some intrinsic stress involved, it provides the most conservative available individual estimate of Pmet while not flying. Flight energetics in barn swallow versus thrush nightingale We compared Pmet, Pmech, Pmet*, Pmech* and energy conversion efficiency (η) between two species for which flight power was measured using the same methods in the same wind tunnel: barn swallows (this study) and thrush nightingales (Macías-Torres et al., 2025). We also compared the post-flight Prec between the two species to evaluate differences in flight versus non-flight metabolism under the same circumstances in both species. Additionally, cost of transport (COT) was calculated for each species considering both Pmet and Pmech. COT is the energy expenditure per unit weight over unit distance. It was calculated as: (10)COT=PowerW×U, where Power is Pmet or Pmech in watts (J s−1), W is the weight in newtons and U is the flight speed in m s−1. Thus, COT is a dimensionless number. Statistical analysis Uncertainty in ηmax, Uη,max and ηfm​ was quantified using a non-parametric case bootstrap (bootstrap iterations=50,000). For each replicate, we resampled rows independently within the mechanical and metabolic datasets (sampling with replacement), refitted both quadratic models and recomputed the efficiency calculation. We report bootstrap means and 95th percentile confidence intervals (CI). Additionally, uncertainty in ηp was obtained via a stratified non-parametric bootstrap: we resampled metabolic flight power within each inclination separately with replacement preserving the original sample sizes (bootstrap iterations=50,000). This resampling draws samples from the empirical distribution of the observed data, so that an observation can appear multiple times. For each replicate, we computed the bootstrapped mean difference ΔPmet(b) and then calculated ηp(b)=ΔPmech/ΔPmet(b) while keeping ΔPmech fixed from the previously calculated value. We report the 95% CI from the bootstrap analysis. Uncertainty in minimum power speed (Ump) and its corresponding flight power (Pmp) was quantified via a non-parametric case bootstrap. We drew B=5000 bootstrap samples by resampling rows of the Pmet and Pmech dataset with replacement (preserving sample size), refitted the same quadratic model to each resample, and recomputed Ump and Pmp. Because a meaningful minimum requires a convex parabola, bootstrap fits with b2≤0 were discarded and excluded from summaries; we report the proportion of such replicates. For valid replicates, we summarised uncertainty using the bootstrap 95% CI for both Ump and Pmp. Analyses were performed in R studio (v4.5.0). Mixed-effects models were fitted with lme4 (lmer) and P-values/d.f. obtained via lmerTest (Bates et al., 2015). Estimated marginal means and contrasts (ΔPmet) were computed with emmeans (https://cran.r-project.org/package=emmeans). From all the fitted models, the best-performing one was a simple polynomial model without random effects, as including individual as a random intercept resulted in model singularity (Supplementary Materials and Methods). RESULTS P met Pmet was 1.52 W (bootstrap mean: 1.49 W; 95% CI: 1.23–1.77 W) at Ump=9.9 m s−1 (bootstrap mean: 9.6 m s−1; 95% CI: 7.4–10.7 m s−1), while Pmet* was 8.72 W N−1 (bootstrap mean: 8.49 W N−1; 95% CI: 6.83–10.19 W N−1) at Ump=9.97 m s−1 (bootstrap mean: 9.63 m s−1; 95% CI: 7.5–10.69 m s−1) (Fig. 1). Power versus air speed in barn swallows (Hirundo rustica) flying in a wind tunnel. (A) Metabolic power Pmet=11.21–1.95U+0.098U2 and (B) weight-specific metabolic power Pmet*=66.44–11.58U+0.58U2 (where U is airspeed). (C) Mechanical power Pmech=0.54–0.092U+0.0048U2 and (D) lift-specific mechanical power Pmech*=3.35–0.566U+0.0287U2. Grey circles show individual measurements, while black lines indicate quadratic polynomial fits with 95% confidence intervals (CI; shaded areas). P mech Pmech was 0.10 W (bootstrap mean: 0.10 W; 95% CI: 0.08–0.12 W) at Ump=9.5 m s−1 (bootstrap mean: 9.4 m s−1; 95% CI: 7.5–10.3 m s−1), while Pmech* was 0.57 W N−1 (bootstrap mean: 0.57 W N−1; 95% CI: 0.51–0.61 W) at Ump=10 m s−1 (bootstrap mean: 10 m s−1; 95% CI: 9.4–10.5 m s−1) (Fig. 1). Whole-animal energy conversion efficiency (η) Whole-animal energy conversion efficiency η for raw values of Pmet and Pmech resulted in a peak of 7% (bootstrap mean: 7.7%; 95% CI: 6.4–9.6%) at 11 m s−1 (bootstrap mean: 10.8 m s−1; 95% CI: 7.9–13.1 m s−1). Whole-animal energy conversion efficiency η based on Pmet* and Pmech* showed a peak of 6.5% (bootstrap mean: 7.3%; 95% CI: 6–9.6%) at 10.4 m s−1 (bootstrap mean: 10.4 m s−1; 95% CI: 7.9–13.6 m s−1) (Fig. 2, Table 1). Flight power and whole-animal conversion efficiency versus airspeed in barn swallows (H. rustica) flying in a wind tunnel. (A) Pmet and Pmech, and whole-animal energy conversion efficiency (η). (B) Pmet* and Pmech*, and η. Black circles represent metabolic power; blue triangles represent mechanical power; red lines represent the whole-animal conversion efficiency calculated by dividing the lines fitted to the mechanical and metabolic power in each panel. The red circle denotes the peak efficiency and the speed it occurs at. Table 1. Energy conversion efficiency estimates calculated in this study Maximum value (95% CI)Airspeed (U) at maximum value, ηmax (95% CI)Whole-animal energy conversion efficiency (η)7% (6–9.6)11 m s−1 (7.9–13.1)Muscle conversion efficiency (ηfm)9% (7.1–11.6)*10 m s−112.7% (9.1–19.5)‡10 m s−1Partial conversion efficiency§ (ηp)27% (delta method: −2–56.1; bootstrapping: 12–153.8)11.4 m s−1 (only one airspeed measured)Whole-body energy conversion efficiency is derived from the best-fitting models for mechanical and metabolic power. *Muscle conversion efficiency was calculated based on Morris and Askew (2010). ‡Muscle conversion efficiency was calculated relative to the metabolic rate when the barn swallows were not flying (post-flight recovery metabolic power, Prec; see Materials and Methods). §Partial conversion efficiency was estimated based on the tilted wind tunnel method (Bernstein et al., 1973; Tucker, 1972). CI, confidence interval. Energy conversion efficiency estimates calculated in this study Whole-body energy conversion efficiency is derived from the best-fitting models for mechanical and metabolic power. *Muscle conversion efficiency was calculated based on Morris and Askew (2010). ‡Muscle conversion efficiency was calculated relative to the metabolic rate when the barn swallows were not flying (post-flight recovery metabolic power, Prec; see Materials and Methods). §Partial conversion efficiency was estimated based on the tilted wind tunnel method (Bernstein et al., 1973; Tucker, 1972). CI, confidence interval. Partial efficiency (ηp) Flight Pmet was measured during horizontal flight (0 deg) and a positive climb angle (4 deg) at 11.4 m s−1. We analysed 20 flights from 6 birds, 11 horizontal and 9 climbing flights (Table S1). The mixed model indicated a positive, marginal effect of inclination on flight Pmet: estimate=0.13 W, s.e.=0.07 W (t16.49=1.95t, P=0.068). Expressed as 4 deg−0 deg values, the estimated marginal mean difference in metabolic power (ΔPmet) was 0.52 W (s.e.=0.28 W; Fig. S2). The calculated change in Pmech due to a climbing flight of 4 deg (ΔPmech) was 0.14 W. This resulted in a ηp of 27% (delta-method 95% CI: −2–56%). The non-parametric bootstrapping yielded a median of ΔPmet(b)=0.53 W with a 95% CI: 0.02–1 W; bootstrap median ηp=26% and 95% CI: 12–154%. Flight muscle efficiency (ηfm) Muscle efficiency calculations resulted in a maximum of 9% (bootstrapping 95% CI: 7.1–11.6%) at 11 m s−1, showing a shallow concave function in the mean values (Fig. 3, Table 1). Flight muscle efficiency estimates in barn swallows (H. rustica). Flight muscle conversion efficiency ηfm=1.1×Pmech/(Pmet−BMR) (following Morris et al., 2010) (blue); and recovery ηfm,recovery=Pmech/(Pmet−Prec) (black dashed), across flight speeds from 7 to 14 m s−1, where Prec is the metabolic power after a flight experiment, during recovery (see Materials and Methods), and BMR is basal metabolic rate. The lines represent the estimated mean values from bootstrapping methods; shaded areas show bootstrapping 95% CI for each corresponding approach. Flight versus non-flight Pmet Mean Prec was on average 0.7 W (s.e.=0.01 W) for barn swallows. Metabolic flight cost at Ump was 2.19 times more costly than Prec. Estimated BMR from allometric equations was 0.24 W (s.e.=0.002 W). Thus, metabolic flight cost at Ump was 6.4 times more costly than estimated BMR. Barn swallow versus thrush nightingale Mean Pmet did not differ significantly between barn swallows (mean 1.89 W) and thrush nightingales (mean 2.01 W; Welch two-sample t-test: t=−0.82, d.f.=76.7, P=0.41; 95% CI: −0.41 to 0.17; Fig. 4A). In contrast, mean Pmet* was significantly higher in barn swallows (mean 10.9 W N−1) than in thrush nightingales (mean 6.3 W N−1; Welch two-sample t-test: t=6.04, d.f.=59.5, P<0.0001; 95% CI: 3.07 to 6.12; Fig. 4B). Comparison of flight metabolic and mechanical power between barn swallows and thrush nightingales across airspeeds. (A) Pmet, (B) Pmet*, (C) cost of transport (COT) based on Pmet, (D) Pmech, (E) Pmech* and (F) COT based on Pmech plotted against airspeed (U). COT (dimensionless) is the ratio between the flight power (metabolic or mechanical) and the bird's weight (N) multiplied by the flight speed. Coloured lines are the best-fitting polynomial models, and the grey shading is the standard error of the model. Pmech differed significantly between the two species. Barn swallows showed lower mean Pmech (0.12 W) compared with thrush nightingales (0.27 W; Welch two-sample t-test: t=−15.53, d.f.=65.9, P<0.0001; 95% CI: −0.16 to −0.12; Fig. 4D). Additionally, Pmech* was significantly lower in barn swallows (mean 0.68 W N−1) than in thrush nightingales (mean 0.86 W N−1; Welch two-sample t-test: t=−5.82, d.f.=53.8, P<0.0001; 95% CI: −0.25 to −0.12; Fig. 4E). Average Pmet-derived COT across all flight speeds was higher in barn swallows than in thrush nightingales, with a mean Pmet-derived COT of 1.1 (interquartile range IQR: 0.71–1.25) compared with a Pmet-derived COT of 0.78 (IQR: 0.61–0.86) in thrush nightingales (Fig. 4C). The opposite pattern was observed for the Pmech-derived COT across all flight speeds, where thrush nightingales had a mean Pmech-derived COT of 0.11 (IQR: 0.09–0.13); barn swallows mean Pmech-derived COT was 0.07 (IQR: 0.06–0.9; Fig. 4F). Post-flight recovery metabolic power (Prec; calculated from the measured amount of CO2 produced after performing an experimental flight) was significantly lower in barn swallows (mean 0.7 W) than in thrush nightingales (mean 0.82 W; Welch two-sample t-test: t=−8.05, d.f.=77.0, P<0.0001; 95% CI: −0.15 to −0.09). However, when accounting for body mass, weight-specific Prec was significantly higher in barn swallows (mean 4 W N−1) compared with thrush nightingales (mean 2.5 W N−1; Welch two-sample t-test: t=16.61, d.f.=61.6, P<0.0001; 95% CI: 1.3 to 1.6; Fig. S3). DISCUSSION Our wind-tunnel measurements provide direct estimates of Pmet and Pmech in un-instrumented flying barn swallows. As expected, both Pmet and Pmech followed classic U-shaped speed–power curves (Fig. 1). We observed that whole-animal conversion efficiency η had a shallow peak at ∼7% at a speed of 11 m s−1, which is lower than direct estimates in other passerines species (Hedh et al., 2020; Macías-Torres et al., 2025), while consistent with the predicted body-mass scaling of conversion efficiency (Currie et al., 2023; Groom et al., 2018). Our results indicate that barn swallows, although less efficient fliers than hypothesised, have a wider range of speeds at which they can fly while performing relatively well. This pattern suggests that flight energy conversion efficiency is both species and speed dependent, and it is probably tuned to match the range of ecologically relevant speeds used by each species. Power measurement validation Barn swallow Pmet has been estimated previously using different methods, resulting in a range of values: 1.34 W (mass loss; Lyuleeva, 1970), 1.3 W (doubly labelled water; Hails, 1979) and 1.62 W (doubly labelled water; Turner, 1982). These measurements were performed on freely flying birds in the wild, and thus the flight speed is unknown, but it is expected that the birds flew at speeds between minimum power and maximum range speeds (Hedenström and Alerstam, 1995). Thus, the measurements of free flying barn swallows agree well with the measured values in the wind tunnel in this study (Pmet at Ump 1.52 W). Another wind tunnel study, using doubly labelled water to measure Pmet in barn swallows, yielded a value of 2.13 W (range 1.74–2.55 W, s.d. 0.21 W at 10.3 m s−1; Schmidt-Wellenburg et al., 2007). Additionally, Pmet can be estimated from allometric equations, such as in Bishop and Butler (2015): Pmet=52.6M0.74, where M is mass (kg). This model predicts a Pmet of 2.69 W for a bird of mass 18 g. Our measured Pmet at Ump (1.52 W) lies within the range of previous measurements and estimates, suggesting that our measurements are realistic. Additionally, it is worth noting that our study birds were all juveniles, and we cannot exclude that flight metabolic costs may be slightly higher than would be expected in fully mature adults. This is because juvenile birds may still be refining flight behaviour, neuromuscular control and flight-related morphology (see Ruaux et al., 2023). Pmech measurements were estimated in barn swallows in a previous study performed in the same wind tunnel with values ranging from 0.17 to 0.27 W across a range of speeds between 6 and 11 m s−1 for a 19 g individual (Pennycuick et al., 2000). Pmech values previously reported were obtained using a different method from that in this study: kinematics of the body and humerus were recorded with high-speed video, and these data were then combined with aerodynamic and inertial principles to calculate the forces, moments and angular displacements, from which Pmech output was derived (Pennycuick et al., 2000). They report that their Pmech measurements ‘turned out to be more than that predicted on the basis of current estimates of body drag coefficient and profile power ratio, possibly because the bird was not flying steadily in a minimum-drag configuration’. Thus, our Pmech measurements (0.07–0.19 W), although lower than previously measured, seem to be realistic. We acknowledge that wake dissipation may have reduced the measured wake energy relative to the energy present immediately behind the wings. However, because measurements were made relatively close to the birds, this effect is likely to be small. Moreover, any dissipation-related underestimation should be broadly consistent across speeds and comparable to that in previous Pmech values reported under the same experimental approach (Currie et al., 2023; Hedh et al., 2020; Johansson et al., 2018; Macías-Torres et al., 2025). In this study, Pmech was calculated as the wake energy per wingbeat multiplied by the wingbeat frequency of the analysed sequence, i.e. the rate at which energy is added to the wake during the analysed sequence. Because our Pmech measurements were based on active flapping wingbeats only, they may overestimate the real wingbeat frequency of the species. To compare our values with realistic wingbeat frequencies (effective wingbeat frequencies that include intermittent pauses between wingbeats, Feff), we compared our values with Feff from Bruderer et al. (2001) for barn swallows in horizontal flight. Across the same airspeeds as our PIV measurements, the estimated Feff from Bruderer et al. (2001) averaged 6.76±0.30 Hz, whereas wingbeat frequency measured during our PIV sequences averaged 8.86±1.09 Hz (Fig. S4). Thus, observed wingbeat frequencies during active sequences were on average 2.10 Hz higher, corresponding to approximately 31% greater values than the effective wingbeat frequencies estimated from Bruderer et al. (2001). Consequently, our Pmech estimates represent active-flapping Pmech and may overestimate the time-averaged Pmech over complete flight cycles that include pauses. Additionally, wingbeat frequency and wake energy are not independent, especially if the active wingbeats compensate for the mid-upstroke pauses observed in this species. If that is the case, the energy added to the wake per active wingbeat may be higher than it would be during continuous flapping without pauses. So, if active wingbeats have both higher frequency and higher energy per beat, estimated Pmech should be potentially higher than for continuous flapping flight. Furthermore, simply replacing the measured wingbeat frequency with Feff, while keeping the measured wake energy per wingbeat unchanged, would not provide a realistic estimate of time-averaged Pmech. Instead, it would assume that each active wingbeat generates the same wake energy regardless of whether pauses are included, which is unlikely, in addition to the unknown power generated during the pauses only. Consequently, our Pmech measurements are probably an overestimation of true flight Pmech of the flap-gliding flight style of barn swallows flying in a wind tunnel. Whole-animal energy conversion efficiency Barn swallows showed a shallow peak in conversion efficiency at intermediate speeds, consistent with observations from another species (thrush nightingales) studied using the same methodology, although this peak was not sharply defined (cf. Macías-Torres et al., 2025). Instead, the bootstrapping method indicates that barn swallows have a broad range of speeds at which they can fly while performing close to their peak efficiency, spanning much of the tested speed range. Being similarly efficient at a wide range of flight speeds probably benefits aerial insectivores such as barn swallows, which exploit a wide speed range during prey capture (Turner, 1982). The flight conversion efficiency value in barn swallows in this study was roughly half of previous estimates for thrush nightingales (15%; average mass 33 g; Macías-Torres et al., 2025) and blackcaps (Sylvia atricapilla; 14–22%; average mass 19 g; Hedh et al., 2020), suggesting that barn swallows are less efficient overall. This pattern is consistent with the body-mass scaling of conversion efficiency, which predicts that smaller birds are less efficient than larger ones (Bishop, 2005; Currie et al., 2023; Groom et al., 2018). Applying the mass scaling equation by Currie et al. (2023) to barn swallows (body mass 17.9 g) yields a predicted whole-animal conversion efficiency of 6.2% (s.d. 0.1), which aligns well with our measurements, providing support for our findings. By contrast, using the same mass scaling equation for thrush nightingales predicts an efficiency of ∼7.2%, well below the 15% in Macías-Torres et al. (2025). This discrepancy between mass-scaling predictions and measured conversion efficiency emphasises the need for additional cross-species measurements of energy conversion efficiency. Additionally, our support for a body-mass scaling of efficiency (Currie et al., 2023) offers a mechanistic explanation for why intermediate size birds, and not the smallest birds, perform the longest non-stop migrations, as seen in several wader species (Gill et al., 2009; Hedenström, 2010; Klaassen et al., 2011). Barn swallows in this study performed intermittent flap-gliding flight, alternating active flapping phases with fixed-wing bounds in the middle of the upstroke (Bruderer et al., 2001). The gliding phase can provide support for up to 20% of the body weight as lift (Tobalske, 2010, 2022). These ‘fixed-wing’, partly outstretched wings, gliding phases were excluded from our Pmech calculations (consisting of just 2–6 wingbeats), but were included in the Pmet estimates, as these lasted about 2 min. Consequently, our Pmech probably reflects an upper bound relative to the time-averaged costs of free flight. If our measurements overestimate the Pmech of barn swallows, then flight energy conversion efficiency would be even lower than the values calculated here. Similarly, an overestimation of Pmet due to handling stress is also a possibility; however, published values for freely flying barn swallows (Hails, 1979; Lyuleeva, 1970; Turner, 1982) align well with our Pmet measurements. Finally, it is possible that barn swallows incorporate some energy-saving flight behaviours during migration, such as using fixed-wing bounds (Tobalske, 2022) or by exploiting wind-speed gradients while gliding with fully extended wings (Warrick et al., 2016). Such behaviours could reduce the overall power required for sustained flight compared with continuous flapping. Yet another possibility is that the natural flight style of barn swallows, with frequent and rapid manoeuvres, with pauses in the middle of the upstroke sequence (Bruderer et al., 2001), prevents them from using energy-saving adaptations suggested for other birds. For example, it has been shown that birds can store energy in the tendon of the supracoracoideus muscle when continuously flapping, allowing them to operate at higher efficiency (e.g. Deetjen et al., 2024). However, the flap-gliding flight seen in our swallows may require active breaking of the wingbeat sequence and overcoming higher inertial power than in other birds (or during steady migratory flight). Similarly, muscle incurs energetic costs not only when producing mechanical work but also during activation, force development and relaxation, with brief downstroke pauses and reactivation possibly increasing the relative cost of muscle contraction compared with the time available for positive mechanical work (Usherwood, 2016). Thus, a muscle activation cost from resuming multiple wingbeats after pauses in barn swallows may result in a relatively high Pmet, partly explaining the apparently low energy conversion efficiency. At the same time, this may not be reflected in the aerodynamic power to the same degree, as Pmech is largely determined by the aerodynamic properties of the wings and body. Lastly, barn swallows often showed lateral movement in the wind tunnel, moving back and forth between the sides of the test section, ‘surfing’, rather than maintaining a completely steady position. Such behaviour may increase Pmet because it involves repeated adjustment in body orientation, force production and flight control, all involving muscle activation. However, these costs may not be fully reflected in Pmech, which was calculated from the wake during selected relatively stable flight sequences and therefore mainly represents the aerodynamic power of the analysed wingbeats. This mismatch between whole-animal metabolic costs and locally measured aerodynamic power could contribute to the relatively low apparent energy conversion efficiency. We stress, however, that these lateral movements were not equivalent to the rapid manoeuvres barn swallows perform in the field, but may represent an additional cost relative to the steadier flight of species previously measured under similar conditions. Partial and muscle efficiency The first ever estimate of partial efficiency (ηp) resulted in a mean value of 23% (Bernstein et al., 1973; Tucker, 1972) and has been used as the reference value in flight power models ever since (Pennycuick, 1975, 2008). ηp is the ratio of the change in power output to power input at different flight efforts: horizontal flight versus climbing flight. The partial efficiency is thus independent of whatever basal metabolism the bird has, assuming that the basal metabolism remains the same at the two climbing rates. Thus, it relates to the flight engine efficiency solely and should reflect similar values to those of flight muscle efficiency. Mean ηp measured here (27%) resembles the values previously obtained, although the uncertainty in our estimate limits its reliability. Flight muscle efficiency estimated in this study only increased the efficiency by a small degree compared with whole-animal conversion efficiency, reaching a maximum of 12.7% at intermediate speeds, aligned with the maximum whole-animal efficiency speed. The discrepancy between partial and muscle/whole-animal efficiency should be interpreted with caution, as partial efficiency estimates were less reliable. Flying in a wind tunnel In this study, barn swallows were restricted to fly in the confined space of a wind tunnel, which inevitably constrains their natural flight behaviour, as they cannot fully exploit the surrounding air to gain kinetic or potential energy as they would do under natural conditions (Masman and Klaassen, 1987). Barn swallows are capable of a diverse flight repertoire in the wild, including gliding, partial bounding, exploitation of near-ground wind gradients during manoeuvres, and reducing wingbeat frequency to capitalise on wind speed gradients close to the surface (Warrick et al., 2016). More broadly, many diurnal migrants shift between continuous flapping, intermittent flight and soaring to reduce energetic costs (Hedenström, 1993; Sapir et al., 2010; Shamoun-Baranes et al., 2016; Tobalske, 2010), behaviours that are inextricably linked to the limitations of flying in a wind tunnel. The range of airspeeds recorded for wild-flying barn swallows spans from 3.7 to 19.4 m s−1, with a mean of 9.5±2.7 m s−1 (Warrick et al., 2016), which overlaps with the flight speeds measured in this study. Notably, the mean wild flight speed aligns with the minimum power speed identified in this study, suggesting comparable baseline flight performance despite experimental constraints. Thus, we believe that flight power estimates from wind tunnel flights in this study provide valuable insights into the baseline of barn swallow flight energetics, although they may not fully capture the full range of flight flexibility that barn swallows make use of in the wild. Flight power in barn swallows versus thrush nightingales Flight Pmet and Pmech is hard to measure in free flying birds and usually comparisons are based on studies using various methodologies (Bishop and Guglielmo, 2022; Tobalske et al., 2003). Here, we discuss the flight power results obtained for two passerine species measured under the same circumstances and using the same methodology. We showed that the two species differed in the range of flight speeds that they could fly at, with species-specific U-shaped power curves (Fig. 4). Pmet differed between the two species, with weight-specific flight Pmet being higher in barn swallows than in thrush nightingales (Fig. 4B), supporting the hypothesis that flight power scales with body mass with an exponent less than 1 (Bishop and Butler, 2015; Hedenström, 2010). Additionally, smaller birds have higher BMR in proportion to their mass (Lasiewski and Dawson, 1967), resulting in their cost of flight being a smaller multiple of their BMR than otherwise expected. This was not fully supported by our measurements as the flight Pmet was a similar multiple of their estimated BMR from allometric equations (in barn swallows, flight Pmet at Ump was 6.4×BMR, while in thrush nightingales it was 6.1×BMR). The post-flight recovery metabolic rate (inferred from measured CO2) is sampled a few minutes after the bird's flight experiment and probably reflects values related to the field metabolic rate, when not performing any flight. This study's post-flight recovery metabolic rates (mean±s.e.m. 4±0.08 W N−1) were 33% lower than field metabolic rates measured in the same species during an entire day using doubly labelled water (104 kJ day−1 for a 20.4 g bird, equivalent to 6 W N−1; Nagy, 1987). Absolute post-flight recovery metabolic rate was lower in barn swallows than in thrush nightingales, as expected for a bird with lower mass. However, weight-specific post-flight recovery metabolic rate was higher in barn swallows than in thrush nightingales, as expected in smaller birds showing higher relative metabolism (Lasiewski and Dawson, 1967). This resulted in barn swallows having almost the same flight power if compared with their post-flight metabolic rate as in thrush nightingales (in barn swallows flight Pmet at Ump is 2.19 times post-flight Prec, while in thrush nightingales it is 2.07 times post-flight Prec). All this indicates that barn swallows have a higher background metabolic rate while flying compared with thrush nightingales, which supports our finding of barn swallows showing a lower whole-animal energy conversion efficiency than thrush nightingales. Pmech* illustrates that barn swallows fly at a lower cost per unit of lift than thrush nightingales, suggesting a higher aerodynamic efficiency in barn swallows compared with the thrush nightingales (Fig. 4D). One cause of this better aerodynamic performance could be a lower body drag, due to a more streamlined body shape, or a more efficient lift production due to the wings having a higher aspect ratio. This was noted in the first wind tunnel studies in this species (Pennycuick et al., 2000). Conclusion Overall, our results emphasise the importance of accounting for scaling when considering conversion efficiency in flying animals. This becomes particularly important when using flight models, such as ‘afpt’ (Klein Heerenbrink et al., 2015; Tools for Modelling of Animal Flight Performance: https://cran.r-project.org/package=afpt), to make predictions about flight energetics. In addition, the difference we found between our barn swallows and previously studied species (i.e. thrush nightingale; Macías-Torres et al., 2025) points towards not only species-specific optimal (i.e. peak efficiency) flight speeds but also ecologically derived variation in the shape of the efficiency curve across flight speeds. While barn swallows may benefit from a wide range of flight speeds that they can efficiently fly at, thrush nightingales perform intensive continuous flights only during migration and may benefit from having a narrower range of speeds with higher efficiency. This introduces the idea that there might be a trade-off between specialisation and flexibility, with higher efficiency at a narrow speed range versus lower efficiency spread across a broader range. However, few species have been studied thus far, and more studies are needed to draw firm conclusions on how ecology, physiology or other factors (e.g. phylogeny) shape the conversion efficiency of a species. Supplementary Material