How might turbulence affect animal flight in a changing world?

The aerial environment profoundly impacts the energy required to fly and the speed at which animals cover ground (Tucker, 1966; Furness and Bryant, 1996; Chapman et al., 2011). Animals adjust their behaviour accordingly by, for example, modifying their flight routes, timings, airspeed and/or flight mode (e.g. flapping or gliding) in relation to airflows (Sapir et al., 2014; Sapir et al., 2011; Bohrer et al., 2011). Yet, while responses to wind and updrafts are relatively well understood (e.g. Thorne et al., 2023; Pennycuick, 2008), relatively little is known about responses to turbulence (Ortega-Jimenez et al., 2016; Combes and Dudley, 2009). Turbulence can be beneficial for flying animals; for instance, the introduction of micro-scale turbulence in the boundary layer over wings can improve the flow attachment and reduce drag (Schlichting et al., 2000). Birds can also extract energy from turbulence where it includes an upward component; for instance, by exploiting the upwash in turbulent wakes of animals flying ahead of them (Friman et al., 2024; Portugal et al., 2014), or the air rising in thermal updrafts (Sapir et al., 2011). Nonetheless, the few metabolic measurements to date show that turbulence can be costly for flapping fliers, because kinematic adjustments are required to maintain stability and control in response to rapid changes in wind velocity at small scales (see ‘The effects of turbulence on the costs of flapping flight’, below).

Anthropogenic activities are providing a new impetus for understanding how turbulence impacts animal flight. Habitat modification and infrastructure projects are affecting the distribution, characteristics and predictability of turbulence near the ground (see ‘Turbulence in the field: how does it vary in space and time?’, below). Climate change is impacting turbulence near ground level, as a result of changes in wind speed and convection, and increasing the frequency and intensity of turbulence at higher altitudes, associated with, for example, thunderstorms and greater atmospheric instability (Trapp et al., 2007; Masson-Delmotte et al., 2021). In this Commentary, I will (1) consider what animals encounter, by summarising naturally occurring and anthropogenic drivers of turbulence and the implications for its spatial and temporal distribution at flight heights, (2) assess what is known about how turbulence affects the metabolic costs of flight and (3) evaluate animals' behavioural responses to turbulence, concentrating on sustained turbulence, rather than discrete gusts (Reynolds et al., 2014). I also make some preliminary predictions about how animals respond to turbulence, assuming that responses to turbulence are related to flight costs and capacities. Overall, the aim is to highlight the importance of turbulence and the challenges and prospects for tackling key knowledge gaps. The focus is on birds, which have been studied in laboratory and field settings, but many of the points will apply to other flying animals.

Most animal flight occurs within the atmospheric boundary layer (ABL), but there is increasing evidence that birds including waders, passerines and nightjars migrate above the ABL at heights of 4–7 km above ground (Sjöberg et al., 2023, 2021; Norevik et al., 2023). Here, wind speeds tend to be stronger and more uniform as a result of the reduction in surface effects. A range of studies have hypothesised that birds operating at these heights could benefit from lower turbulence (Kerlinger and Moore, 1989; Schmaljohann et al., 2007; Bowlin and Wikelski, 2008). Nonetheless, high-flying birds can still encounter turbulence associated with storm clouds, lee waves and frontal systems. Such turbulence can show marked regional variation (Fig. 1) that could also affect flight altitudes in long-distance migrants.

Within the ABL, turbulence dominates because of the effects of surface friction and heat flux on the mean flow. This influence typically extends to 1 km above ground, and to ∼3 km over deserts, dry fields and boreal forests. Characteristic turbulence intensities may be ∼5–20% of the mean wind speed, with most of the turbulent kinetic energy being associated with eddies from metres to ∼1 km in size, with the length scale of eddies increasing with height (LeMone et al., 2019). All animals will therefore experience eddies that vary in length scale and velocity, whether they operate over flat, open landscapes, or near topography or obstacles that perturb the flow. Nonetheless, there will be spatial and temporal patterns in the likelihood, magnitude and nature of turbulence depending on the surface characteristics and regional atmospheric conditions, i.e. wind speed and solar heat flux.

Areas with strong solar heating will experience convective turbulence, or thermal turbulence, driven by the uneven heating of the substrate. This type of turbulence dominates in low wind conditions and can cause perturbations over scales from millimetres to kilometres (Ortega-Jiménez and Combes, 2018). In unstable atmospheric conditions, air heated by warmer areas of the substrate rises, drawing in cooler air at the base. This creates convective ‘cells’ that expand with altitude and can span the depth of the boundary layer. These cells can cover a large proportion of the ground surface area; for instance, thermals with a minimum diameter of 40 m covered 40–50% of a survey area in Colorado, with other studies reporting similar fractional coverages (Young, 1988). The flow within the convective cells can become turbulent in cases with extreme surface heating (Chillà and Schumacher, 2012).

Once wind speeds increase above a few metres per second, mechanical mixing becomes the predominant type of turbulence generation. Surface friction slows the flow close to the substrate, producing wind shear and turbulent eddies, with the turbulence intensity increasing with the wind speed (all other factors being equal). Turbulence is also generated by topography, forming in the lee of hills and other obstacles in the flow including buildings and trees. There are similarities between the flow over cities and that over plant canopies (Roth, 2000), in that a shear layer develops near the top of the canopy, and the canopy converts the mean flow to turbulent flow and reduces the wind speed. Nonetheless, the size of the eddies is larger in urban environments as this varies with the size of the obstacle, with the length scale in forests being linked to the leaves, twigs and branches that produced them (Poggi et al., 2004). In cities, wind-driven turbulence interacts with the urban heat island effect, which increases atmospheric instability near the surface and generates further mixing, amplifying the turbulent eddies (Roth, 2000). The net effect of the many sources of turbulence in the urban environment is that the rate of turbulence production and the turbulence intensity are up to twice as great as in flat, rural environments (Roth, 2000).

Wind farms deserve particular consideration as the turbulence profile of the wakes, particularly that close to the turbines, is distinct from anything that animals will have encountered in the natural world (Fig. 2). Furthermore, the need to move towards low carbon economies is driving large-scale developments (Drewitt and Langston, 2006), which can introduce additional turbulence in areas of hundreds to thousands of square kilometres. Wind farms have important impacts on the ABL. The main upstream effect is a small reduction in wind speed due the blockage effect of turbines, but this may only amount to a 3% reduction in wind speed (Porté-Agel et al., 2020). The wind speed also decreases downstream, being most marked in the near-wake region, which is 2–4 rotor diameters downwind of the turbine. This region is characterised by periodic helicoidal vortex structures, which are continuously shed from the tip and the root of the rotor blades (Fig. 2). Turbulence intensity peaks in the transition region between near- and far-wakes, with the greatest turbulence near the upper edge of the wake. Nonetheless, turbulence remains high in the far-wake: turbines introduce turbulence with scales of 1–10 m (Stevens and Meneveau, 2017) and turbulence intensities are some 10–30% above that in the incoming flow for offshore wind farms. Far wakes can extend up to 70 km downstream for offshore wind farms (Platis et al., 2018), with the length and the dominant turbulent scales affected by the wind farm layout and atmospheric stability. Wakes persist for longer when the atmosphere is stable because large eddies in unstable atmospheric conditions (100 m to kilometres in scale) break down the small-scale structures introduced by turbines (Stevens and Meneveau, 2017). This explains why wakes are longer in offshore wind farms. Wind farms also introduce turbulence into the air above them, as a shear layer forms between the slower, wake-affected air and the faster-moving air above it, generating an internal boundary layer that can be a few metres thick or as deep as the ABL.

Turbulence can increase the costs of flapping flight, with the impact varying with the eddy length scale (see below). To give some context about the magnitude of these costs, I will first estimate how wind affects flight costs, considered per unit time (i.e. flight power). Wind affects flight power because headwinds and tailwinds change the speed that provides the minimum costs per unit distance travelled, i.e. the maximum range speed, U_mr (Hedenstrom and Alerstam, 1995). The change in costs can be estimated using the afpt package (https://CRAN.R-project.org/package=afpt), which is an implementation of the model from Klein Heerenbrink et al. (2015). This uses aeronautical theory to predict the power curve from morphometric data, and then applies an optimality approach to derive the maximum range speed, and associated power, for a given headwind or tailwind component. Not all species will be able to increase their airspeed in the manner predicted, as the ability to increase speed, and the costs of doing so, vary with mass (Fig. 3). Indeed, Pennycuick et al. (2013) showed that only species with a mass <1.1 kg migrated at speeds around U_mr, whereas birds over 1.3 kg flew at speeds lower than this. Species less than ∼1 kg should therefore have the available power to increase their maximum range speeds when flying in headwinds.

Using afpt, I estimated the costs of birds increasing their airspeed to fly at their maximum range speed in a 5 m s⁻¹ headwind. This increased flight costs by 11–38% compared with the costs of flying at the maximum range speed in still air (taking the 21 species with a mass <1 kg in Pennycuick et al., 2013), increasing to 74% for the pied wagtail (Fig. 3). A headwind of 5 m s⁻¹ represents a condition that many animals will experience regularly, being classified as a gentle breeze on the Beaufort scale, although it represents ∼75% of the predicted maximum flight speeds for species in Fig. 3 (Table S1).

Aeronautical frameworks have proved to be very powerful, allowing us to model how wind affects the costs of flapping flight per unit time and per unit distance, according to wind support and bird morphology. However, these fundamental frameworks, which have underpinned studies of provisioning and migration (e.g. Hein et al., 2012), all assume that the air is laminar. Clearly, this assumption is violated for nearly all animal flight, given that turbulence is almost ubiquitous. This leads to the question of how much turbulence can affect flight costs. The limited measurements that are available demonstrate that turbulence can increase flight costs considerably. In the wild, the heart rates of migrating Swainson's thrushes (Catharus ustulatus) are higher when flying in turbulent conditions (Bowlin and Wikelski, 2008). Budgerigars (Melopsittacus undulatus) flying in highly turbulent air (43% rms) increase their oxygen consumption by 100% compared with birds flying in smooth air (Tucker, 1966, 1968). Finally, Ortega-Jimenez et al. (2014) quantified the costs of flight in Anna's hummingbirds (Calypte anna) flying behind three wake-generating cylinders with diameters equal to 38%, 77% and 173% of the wing length and found that flight costs increased by 25% behind the largest cylinder compared with control conditions.

Turbulence incurs a cost because animals must respond to the aerial variability. Very small eddies do not seem to make much difference to animals in flight, but when eddies reach a similar magnitude to the wing chord or span, the two wings may experience different airflows, and animals must make kinematic and postural adjustments to maintain flight stability and control (Fig. 4). This explains why Tucker (1972) found no effects of small airspeed fluctuations (i.e. 0.88–1.44% rms/mean values) on flight metabolism in the laughing gull (Larus atricilla), while eddies the scale of the wing span increased flight costs in both insects and hummingbirds (e.g. Ortega-Jimenez et al., 2013, 2014). For instance, orchid bees (Euglossa imperialis) become increasingly unstable about their roll axis as turbulence increases, eventually crashing or being shot out of the jet (Combes and Dudley, 2009). Orchid bees improve their roll stability by extending their hindlegs, a response also documented in hawkmoths, Manduca sexta (Ortega-Jimenez et al., 2016), but this increases drag and associated power requirements by 30% (Combes and Dudley, 2009).

Birds show a number of kinematic responses to increase their stability in turbulent flow. Hummingbirds vary the orientation and fan angle of their tails to increase passive stability, which increases drag (Ravi et al., 2015). Birds also increase their mean wingbeat frequency and/or amplitude, both of which can enhance manoeuvrability and stability (Ravi et al., 2015; Ortega-Jimenez et al., 2014; Lempidakis et al., 2024). However, these changes are typically small compared with the variability between strokes, i.e. the variability in wingbeat amplitude, wingbeat frequency and stroke plane (Ravi et al., 2015; Ortega-Jimenez et al., 2013, 2014). For instance, hummingbirds flying behind cylinders increase the standard deviation in their wingbeat amplitude and frequency by as much as 200% and 50%, respectively (Ortega-Jimenez et al., 2014). Such adjustments between wingbeat cycles is thought to be more costly than a continuous flight style; indeed, it may account for much of the 25% increase in flight costs for hummingbirds flying behind a large cylinder (Ortega-Jimenez et al., 2014). Nonetheless, there is no simple way of estimating the costs associated with variable kinematics, not least because part of the kinematic response could be passive. Separating the effects of active, behavioural control from aerodynamic effects of the flow on the wings is complex, as it requires direct quantification of neuromuscular activity (Liao et al., 2003). This highlights the need for more metabolic measurements, which can at least provide an estimate of the net cost of operating in turbulent flows and could therefore provide insight into the effects of eddies according to their length scale, bird morphology and flight style.

Overall, the few metabolic measurements to date demonstrate that turbulence can have a substantial impact on flight power, increasing it by up to 25% in hummingbirds and potentially much more in budgerigars, when eddies are of a similar size to the birds themselves. As eddy size increases beyond this, birds will take longer to traverse them and will experience them as changes in air velocity. This should require fewer adjustments to maintain flight stability and control, resulting in a lower energetic penalty (Fig. 4). Metabolic costs may also vary with flight morphology; for instance, moths showed large variations in yaw but only moderate oscillations in roll as flow variability increased, whereas bees showed increasing instability about their roll axis (Ortega-Jimenez et al., 2013; Combes and Dudley, 2009). Birds also diverge in their manoeuvring style (Dakin et al., 2018) and it may be that this or traits such as the range of motion (Baliga et al., 2019) affect their responses to turbulence, with implications for the size and magnitude of eddies that birds can respond to and the associated energetic costs.

Finally, turbulence should have important consequences for the costs of transport (the costs per unit distance), as well as flight power, as there is a general tendency for animals to reduce their maximal speeds in stronger turbulence (but see Lempidakis et al., 2024). Indeed, there are parallels between turbulence and climbing flight, as animals reduce their airspeed to accommodate the additional power required in both scenarios (Berg and Biewener, 2008). In the case of turbulence, hawkmoths fly at 3 m s⁻¹ in laminar flow but can only achieve 2 m s⁻¹ when flying in large vortices (Ortega-Jimenez et al., 2013). Similarly, Anna's hummingbirds can fly and feed at 12 m s⁻¹ in unperturbed flow (Clark and Dudley, 2009) but could only feed continuously at 9 m s⁻¹ in the wake of a large cylinder (Ortega-Jimenez et al., 2014). The reduction in speed is substantial in both cases (33% and 25%, respectively). This, combined with the increase in power required to fly in turbulence, should result in a marked increase in the costs of flying a given unit distance.

Is there evidence that animals avoid turbulent eddies, and the associated costs? Flying animals must be able to respond to the flight conditions they encounter routinely. Strong turbulence occurs in the ABL and at high altitudes (Figs 1, 2 and 4), and animals may adjust their flight timing or route to reduce the probability of encountering it. It has long been theorised that this affects migration, providing advantages to migrating at night and/or at high altitudes where the flow is more laminar (Kerlinger and Moore, 1989; Raynor, 1956). However, it has been hard to distinguish the role of turbulence from other environmental factors such as wind and temperature (Van Belle et al., 2007; Shamoun-Baranes et al., 2017), and recent evidence suggests that heat balance plays a major role in determining the flight height of daytime migrants (Sjöberg et al., 2023). Nonetheless, few studies have explicitly tested whether animals adapt their flight trajectories in relation to turbulence (Larkin, 1982) beyond the responses of soaring birds to rising air (e.g. Nourani et al., 2021; cf. Crall et al., 2017). This may be due in part to the complexities of estimating time-varying turbulence. Advances in computational fluid dynamics models should provide new prospects here (Duraisamy et al., 2019). Furthermore, measurements made onboard birds themselves could also open up new avenues for investigation, as the body accelerations and fine-scale fluctuations in height, akin to the bumpiness we encounter in aircraft, both vary with turbulence (Laurent et al., 2021; Lempidakis et al., 2022).

Offshore wind farms provide model scenarios to test turbulence avoidance, as spatially predictable patches of turbulence occur within wind farms and their wakes. These patches are particularly marked in the marine environment where low levels of thermal convection mean that the surrounding flow tends to be relatively laminar. The near-wake area behind turbines should be particularly costly to fly through due to the combination of relatively low integral length scales and moderate to strong wind speeds (Fig. 2). We know from onboard loggers and radar tracking that certain taxa make striking detours around wind farms (Desholm and Kahlert, 2005; Plonczkier and Simms, 2012). This has mainly been attributed to birds responding to visual stimuli (May, 2015; Martin, 2011), yet some of the responses are also consistent with an avoidance of strong turbulence. For instance, there is an increase in avoidance when turbines are active (Dierschke et al., 2016) and hence generating turbulence. Furthermore, the strongest, most predictable avoidance occurs in heavy species with high wing loading, including waterfowl (mainly geese and common eiders, Somateria mollissima) (Desholm and Kahlert, 2005) and northern gannets (Morus bassanus) (Peschko et al., 2021), which have lower available power, and hence less capacity to increase their flight power.

Tracking data could be used to test whether turbulence predicts the distances that birds divert around wind farms. If turbulence matters then the diversion distance should vary with wind speed and bird approach angle (Drewitt and Langston, 2006). Birds approaching down-wind will encounter the far-wake before they reach the wind farm itself and they should therefore divert their path around wind farms at greater distances than those approaching up-wind or from the side (noting that wake length is determined by wind speed, atmospheric conditions and wind farm characteristics).

Wake characteristics could also affect how birds move within wind farms. The wakes behind individual turbines can converge, producing ‘full wake’ conditions, with turbulent flow throughout the wind farm, or ‘partial wake’ conditions, with corridors of low turbulence between them. It would therefore be interesting to assess whether birds are more likely to fly within wind farms when these corridors appear and whether the flight trajectories align with their orientation (determined by the wind direction, Fig. 2). The routes birds take would also need to be considered in relation to whether birds are flying with a headwind component. This will have its own costs, as birds increase their airspeed with headwinds (Hedenstrom and Alerstam, 1995), which may make them more sensitive to factors that increase their costs further. Headwinds will also reduce the effective length scale of the eddies, potentially producing eddies that are closer to the dimensions of birds. Overall, turbulence could therefore influence avoidance at multiple scales, from the ability to manoeuvre away from the moving parts (Furness et al., 2013), sometimes termed micro-avoidance, to the avoidance of individual turbines and/or whole wind farms (i.e. meso and macro scales), all of which could influence collision risk.

In a general sense, turbulence could be used as a cue even if it is not the ultimate driver for avoidance. Birds are likely to be tuned in to changing turbulence levels as turbulence is associated with risks and costs in natural systems. For example, the arrival of frontal systems is marked by gustiness, and changes in pressure and temperature, which often precede changes in wind, and in the case of cold fronts, precipitation. Most of these factors can affect flight costs and capacities, and can affect the intensity of passerine migration (Van Belle et al., 2007; Shamoun-Baranes et al., 2017). Convective turbulence could even be linked to predation risk in some systems as flight height and activity of raptors is also linked to the development of thermal updrafts (e.g. Shamoun-Baranes et al., 2003).

Turbulence should levy a cost for all flapping fliers, because it pervades all aerial habitats, including those that might appear sheltered, e.g. underneath forest canopies. Animals are therefore paying these ‘overheads’ on a daily basis, but their magnitude is unknown. One important consequence of this is that estimates of flight costs made from birds flying in the wild include the costs of flying in turbulence. The baseline costs of flight in still air will therefore be lower than these estimates suggest. If we are to predict how turbulence affects animal behaviour, we need more data on the consequences for the metabolic costs of flight, the costs of transport and particularly how these vary with eddy scale and velocity in relation to flight morphology.

Testing whether animals detour or land to avoid strong turbulence could provide valuable information on whether and when turbulence becomes costly, including for large fliers that cannot easily be studied in the laboratory. It will be important to couple tracking data with estimates of the flow characteristics in areas where birds do not fly, e.g. using analytical or computational fluid dynamics models, to see whether route selection, and in particular route adjustments, can be predicted by specific turbulence characteristics. Where animals do fly through turbulent areas, their flight characteristics could provide insight into associated costs; for example, if birds reduce their airspeeds in relation to the turbulence profile. If flying animals do modulate their flight decisions in relation to the metabolic costs of turbulence, then responses should also vary with body mass and available power.

Turbulence affects flight stability and control, as well as costs. Indeed, the two are inherently related (see ‘The effects of turbulence on the costs of flapping flight’, above). While this Commentary has focused on energetics, birds could also adjust their behaviour in relation to a reduction in stability, which becomes particularly pertinent when animals operate close to the substrate or obstacles in the flow (Shepard et al., 2016). Animals may therefore prioritise currencies of costs or control in different scenarios and/or flight heights.

Field measurements and flow models will also provide invaluable information on the turbulence that animals encounter, including the envelope of conditions they tolerate and those they avoid. This will provide important context for the design of future laboratory tests. One of the challenges facing experimental biologists is how to simulate conditions that animals experience in the wild. The turbulence generated by grids is more akin to freestream turbulence than the regular and predictable eddies in von Kármán vortex streets (cf. Ravi et al., 2015). Nonetheless, grids may not be good options for producing eddies with maximum length scales larger than wingspans within the confines of wind tunnel test sections. Gust generators could be used to produce gusts that vary in length scale and magnitude (Balatti et al., 2022), enabling the effects of velocity and length scale to be investigated independently from flight speed (Ortega-Jimenez et al., 2013, 2014) to establish which elements are the most challenging for flight control and the most metabolically costly. The combination of laboratory and field approaches should therefore provide insight into the costs and risks associated with turbulence, which is much needed against the backdrop of increases in turbulence frequency and intensity due to land-use change and climate change.

I am very grateful to The Company of Biologists for the stimulating presentations and discussions at the Journal of Experimental Biology 2024 Symposium entitled: Integrating Biomechanics, Energetics and Ecology in Locomotion. I would also like to thank Anders Hedenström, Simon Watkins and two anonymous reviewers for their constructive feedback.