Introduction

The current pace of climate change is more rapid than has been experienced for the past 65 million years (Diffenbaugh and Field 2013; Kemp et al. 2015). Since climate is a major determinant of the natural distribution of species, recent poleward shifts in distribution have already been detected across a diverse range of taxa in response to climate change (e.g. Walther et al. 2002; Parmesan and Yohe 2003; Hickling et al. 2006; Pearce-Higgins et al. 2014; Platts et al. 2019). Rapid changes in temperatures can affect the survival or reproduction of organisms and pose a serious threat to the persistence of species, particularly for species that can neither disperse nor adapt with sufficient speed (Bale et al. 2002; Parmesan 2006; Berg et al. 2010; Bellard et al. 2012). In particular, high temperatures can have both lethal and sub-lethal effects on wildlife, resulting in mass mortality in extreme cases (McKechnie et al. 2012; Conradie et al. 2019).

Songbirds, due to their small size, high metabolic rate and primarily diurnal habits, can be especially sensitive to changes in temperature (McKechnie and Wolf 2010; du Plessis et al. 2012; Bourne et al. 2020a, b). To date, most studies on this topic have examined the effects of temperature on physiological traits (Dubois et al. 2016; Xie et al. 2017), but it can also induce changes in behavioral such as in foraging (du Plessis et al. 2012; Edwards et al. 2015), cognitive performance (Soravia et al. 2021; Blackburn et al. 2022), and acoustic communication (Garson and Hunter 1979; Luther and Danner 2016; Coomes et al. 2019a, b; Coomes and Derryberry 2021). Recent research found that high temperatures reduce the ability of birds to discriminate conspecific and heterospecific songs (Coomes et al. 2019a, b), but little is known about the effect of heat on responsiveness to other types of vocalizations. Understanding the effect of temperature on responsiveness to vocal information is important, since vocal signals are a primary form of communication for many bird species (Bradbury and Vehrencamp 1998). Among different contexts of communication, responsiveness to alarm calls is an especially important behavior to consider because alarm calls are anti-predator signals that contain critical information for avoiding predation (Templeton et al. 2005; Magrath et al. 2015). Any failure to discriminate the information content of these calls at high temperatures could thus lead to reduced survival.

Alarm calls are often classified in two ways: (i) flee alarm calls, which are associated with the caller escaping while signaling other individuals to freeze or flee, or (ii) mobbing calls, which are associated with the caller approaching and vocalizing or displaying toward the predator while recruiting others to join it (Curio 1978; Magrath et al. 2015). The principal benefit of mobbing is to induce the predator to leave the area (Curio et al. 1978; Pettifor 1990). Thus, mobbing may reduce the immediate predation risk for the mobbers (Pavey and Smyth 1998) and may limit the risk of future attacks (Flasskamp 1994). Although no information is currently available on the energetic costs of mobbing (reviewed in Carlson and Griesser 2022), this behavior may be energetically expensive because it comprises repeated locomotion, movements of multiple body parts simultaneously (e.g. head bobbing, tail flagging) and vocalizations (Smith and Graves 1978; Crofoot 2012). During hot days, birds may be expected to mob less because recent evidence suggests that, similar to humans, animals can experience cognitive decline during heat stress (Coomes and Derryberry 2021; Soravia et al. 2021; Blackburn et al. 2022). This could reduce the ability to perceive and respond to stimuli in the surrounding environment, such as cues of predator presence. During heat stress, individuals also have to trade-off the physiological need to offset heat against their investment in other behaviors (Cunningham et al. 2021), and hence may be less likely to invest in particularly active behaviors such as mobbing.

In this study, we investigated whether ambient temperatures affect the ability of great tits (Parus major) to respond to playback of conspecific mobbing calls (Fig. 1a). Great tits are known to mob a range of predators (Lind et al. 2005; Dutour et al. 2016; Kalb and Randler 2019), and to respond to mobbing calls from both conspecifics and heterospecifics (Randler and Vollmer 2013; Dutour et al. 2019). We predicted that great tit responses would be lower during hot days, and in particular that when hearing mobbing calls on hotter days would approach the speaker less closely in terms of distance, and also emit fewer mobbing calls. Previous studies have suggested that temperature affects the frequency of avian song both via the immediate effects of current temperature (e.g. Gottlander 1987) or via carry-over effects of prolonged high temperature events (time lag effect, e.g. Garson and Hunter 1979). Thus, both current and past temperature states are important considerations when studying the potential effects of high temperatures on avian communication. Finally, Strauss et al. (2020) suggested that temperature might have a non-linear effect on singing, with a reduced number of songs under both cold and warm conditions, mainly because of the higher energetic demands leading to a trade-off between foraging and vocalizing. However, Strauss et al. (2020) did not find non-linear patterns in their data, perhaps because the temperature during their behavioral experiments never exceeded 20 °C. We therefore expanded on Strauss et al.’s (2020) experiments by investigating the non-linear relationship between temperature and responses to mobbing calls across a larger range of temperatures.

Fig. 1
figure 1

Spectrogram of typical (a) mobbing call uttered by great tits, and (b) song uttered by common chaffinches. The mobbing call of great tit is made of the combination of introductory note followed by broadband frequency notes (D notes). The song of common chaffinch consists of two or three trill-like phrases, followed by an end phrase of more complex and dissimilar notes

Full size image

Material and methods

General methods

This study measured the response to mobbing calls of great tits at air temperatures ranging from 18 °C to 36 °C. The great tit's thermoneutral zone range is 15–30 °C (Broggi et al. 2005; Mathot et al. 2016). The maximum temperature of 36 °C is considered a high temperature in the region where this study was conducted (see https://www.meteo-sain-bel.fr/). Data were collected during playback tests conducted on wild great tits inhabiting mixed deciduous-coniferous forests located near Lyon (45.818992°N, 4.517753°E; France). All tests were conducted between 24 June and 4 September 2020, during a typical dry summer period (except two additional control tests conducted in early October; see hereafter). Since the focal birds were unbanded, we kept a minimum distance of at least 200 m between experimental sites to minimize the chance of testing the same individuals more than once (as in Dutour et al. 2017; Kalb and Randler 2019). As a control, songs of common chaffinch (Fringilla coelebs) were played back to great tits (Fig. 1b, Kalb and Randler 2019). Chaffinches are widespread throughout the study area and great tits were assumed to be familiar with chaffinch song. A total of 92 great tit individuals were exposed to audio playback tests (mobbing calls: 60 individuals, control: 32 individuals). The playbacks (mobbing calls or control) were evenly distributed across the study period to avoid a temporal effect on the tests (see Online Resource 1).

Call collection and stimuli preparation

Mobbing calls used in the study were recorded from great tits responding to playback of mobbing calls of conspecifics (Dutour et al. 2017), given by individuals in response to predator mounts (Kalb et al. 2019), or were obtained from the Xeno-Canto online database (http://www.xeno-canto.org/) with search criteria specifying the type of vocalization as “alarm call”. In Xeno-Canto, we reduced our selection to sound files that we could identify as associated with a mobbing event (most of these mobbing calls have already been used in previous mobbing studies; e.g. Dutour et al. 2021; Salis et al. 2021), that were quality “A” (i.e. highest recording quality), and that had no other bird species audible in the recording. We used mobbing calls of great tits that were composed of combinations of frequency-modulated elements followed by a string of a repeated loud broadband elements (FME-D call; Dutour et al. 2019). From these recording files, 40 unique soundtracks of mobbing calls were built using Avisoft-SASLab software (i.e. with one calling individual per soundtrack). Chaffinch songs were obtained from the Xeno-Canto online database with search criteria specifying the type of vocalization as “song”, quality “A”, and without other bird species in the recording. From these recording files, 32 unique soundtracks of chaffinch songs were compiled (one calling individual per soundtrack). The song of chaffinch consists of two or three trill-like phrases, followed by an end phrase of more complex and dissimilar notes (Nottebohm 1968). On each soundtrack, low frequency noise had been systematically removed (below 1 kHz, Kalb and Randler 2019) and natural calling rates were used for each species: great tit mobbing calls were repeated at a rate of 26 calls/min (Dutour et al. 2019), and chaffinch songs at a rate of 8 songs/min (rates observed in previous records; unpublished data). On average, the duration of a great tit mobbing call was 0.88 s and the duration of a chaffinch song was 2.5 s (Fig. 1). Consequently, focal great tits heard 23 s of great tit mobbing call per minute compared to 20 s of chaffinch song. Each sound file consisted of one minute of playback and was saved in Waveform audio file format.

Playback experiment

All playback tests were conducted between 11 am and 2 pm during days with cloudless sky and without strong wind or rainfall. After finding a great tit, a remotely controlled speaker (Shopinnov 20W, frequency response 100 Hz - 15 kHz) was hung from a tree 1.5 m from the ground (Suzuki et al. 2016) and ~ 20 m away from the bird (Dutour et al. 2021). Before the beginning of the test, the baseline behavior of the focal individual was observed during a pre-trial period lasting one minute to ensure that the bird was not showing vigilance or antipredator behaviors. To determine the responses of the focal tit to the different playback treatments (mobbing calls or control), two behaviors were measured: the minimum distance to the speaker using a tape measure after the playback (Kalb et al. 2019), and the number of mobbing vocalizations during 1 min of playback (Carlson et al. 2017). We presumed that a closer approach to the loudspeaker and a greater number of mobbing calls reflect a higher mobbing response from the focal bird. It was not possible to record data blind because our study involved focal animals in the field. All signals were broadcast with the same intensity (~ 80 dB sound pressure level, measured at 1 m from the loudspeaker, which is within the range of natural mobbing calls produced by great tits; Templeton et al. 2016, and within the range of songs produced by chaffinches; Brumm and Ritschard 2011). In order to limit pseudoreplication (Hurlbert 1984; Kroodsma et al. 2001), each control playback was used only once. For the mobbing call playbacks, 20 exemplars were used once, while the other 20 were used twice. Immediately after each playback, the time of the playback was recorded, as well as the air temperature (T0). To investigate time lag effects of ambient temperature, the mean temperatures observed for the 5, 12, 24 and 48 h before the test were calculated for each playback test (T5 to T48) from hourly records collected from the nearest weather station (< 6 km from the study area; https://www.meteo-sain-bel.fr/; 45.818000°N, 4.588517°E). We calculated T5 over the five hours before the test to ensure a mean temperature calculated on diurnal temperatures only for all the tests (T6 would have included nocturnal temperatures in some tests). Humidity levels during the tests were also recorded (these levels were obtained from the local weather station in the same way that temperatures were obtained).

Statistical analyses

To test whether the mobbing calls generated the expected mobbing response on the behavioral variables tested (B1: the minimum distance to the speaker and B2: the observed number of mobbing vocalizations) we checked first whether responses to the mobbing calls were different to responses to the control. We ran (generalized) mixed models with treatment (mobbing calls or control) as a predictor term for (i) the minimum distance to the speaker (fitted with a Normal distribution) and (ii) the number of mobbing vocalizations (fitted with a Poisson distribution). In both models, the identity of the playbacks (exemplar ID) was systematically added as random effect.

Once we had confirmed that the mobbing calls generated the expected mobbing response and the control sounds triggered lower responses than conspecific mobbing calls, we analyzed the impact of temperature on the behavioral responses to mobbing call playbacks using two linear mixed models (M1 associated with B1 the closest distance to the speaker; M2 associated with B2 the number of mobbing vocalizations). In the two models, the identities of the playbacks (exemplar ID) were included as random effects. The time of day the playback test was conducted, and the temperature were included as dependent variables. As temperature scales (T0: Temperature during the test, T5-48: mean temperatures of the 5, 12, 24 and 48 h before the test) were not independent and were highly correlated, a set of 5 mixed models were ran for each behavior (B1 and B2), each of them based on a different scale of temperatures introduced in the model as dependent variables. As temperature response curves are typically quadratic, with an intermediate optimum (Uiterwaal and DeLong 2020), a quadratic effect of temperature was systematically included in the models. For each behavior (B1 and B2), a model selection process (AICc ranking; package bbmle, Bolker et al. 2017) was conducted on these five models, and an additional null model and a model without temperature effect, to compare model fit and identify the model that explained the most variation in data patterns (i.e. the model incorporating the more relevant temporal scale of temperature). The significance of each explanatory term (temperature, quadratic effect of temperature and time of day) was tested using z tests. Since high humidity can make it more difficult for birds to dissipate heat due to evaporative cooling and may have affected behavior, and date can have an effect on mobbing behavior (Coomes JR et al. 2019), the impacts of humidity and date have been controlled through additional models derivate from the final models selected, to avoid over-parametrization of the models (all details available in Online Resources 2 and 3). All statistics were carried out using R v.3.5 software (R Development Core Team 2017). All models were analyzed using the package glmmTMB (Brook et al. 2017), and the quality of the model estimates was monitored using Pearson residuals. For all statistical tests, the level of significance was set at P < 0.05.

Results

Responses to mobbing calls versus songs (control)

Birds approached the speaker more closely during playback of great tit mobbing calls than during playback of control (chaffinch) songs (mobbing calls – mean distance = 12.58 m vs. control – mean distance = 19.24 m, generalized linear mixed model: χ2 = 23.394, P < 0.001; Fig. 2a). Similarly, focal individuals produced significantly more mobbing vocalizations in response to the mobbing call playback (mean number of vocalizations = 5.30) than to the control playback (mean number of vocalizations = 0.21) (generalized linear mixed model using Poisson family: χ2 = 57.271, P < 0.001; Fig. 2b).

Fig. 2
figure 2

Great tit response to mobbing calls and control treatments (N = 92 individuals) in terms of (a) minimum distance (in meters) to the loudspeaker and (b) Number of mobbing vocalizations produced during 1 min of playback. The boxes range from Q1 (the first quartile) to Q3 (the third quartile) of the distribution and the range represents the IQR (interquartile range). The median is indicated by a line across the box. The “whiskers” on box plots extend from Q1 and Q3 to 1.5 * IQR. Outliers are represented as white dots. The mean and standard error for each treatment are shown in red. Different letters (a and b) indicate significant differences

Full size image

Temperature effect on responses to mobbing calls

The different temporal scales of temperature tested (from the temperature during the tests to the mean temperature in the 48 h before the test) showed distinct patterns with the behavioral variables. The minimum distance to the speaker reached by focus individuals was best explained by the temperature experienced in the five hours prior to the test (T5, Table 1), whereas the number of mobbing vocalizations given in response to playback was best explained by temperature during the test (T0, Table 1). Accordingly, the final models implemented, as a temperature effect, (i) the mean temperatures observed for the 5 h before the test when considering minimum distance to the speaker reached by focal individuals (B1) as a dependent variable, and (ii) the immediate temperature during the test when considering the number of vocalizations (B2) as a dependent variable.

Table 1 ΔAICc ranking based on the two set of models investigating behavioral response to playback of mobbing calls: M1 = minimum distance to the speaker by the focal bird during playback and M2 = number of observed vocalizations by the focal bird during playback, and comparing the null model and a model without temperature effect with models incorporating the effect of temperatures at several temporal scales (T0: temperature observed during the playback to T48: mean of the 48 h before the playback, N = 60). A null value indicates the model ranked as the best model; models in which the difference in AICc relative to AICcmin is < 2 can be considered to have substantial support (Burnham et al. 2011)
Full size table

The temperature during the five hours prior the playback was related to the minimum distance of approach to the speaker during mobbing call playback (Table 2; Fig. 3). This effect of the temperature was quadratic (Y = 41.033—3.233X + 0.083X2), with individuals approaching closer to the speaker during mobbing call playback when the mean temperature in the five hours before the playback averaged 19.5 °C (Max{x} ∈ R(41.033—3.233X + 0.083 X2)). At high temperatures, the mean temperature in the five hours before the playback thus negatively correlated with the approach of individuals. The time of day at which the playbacks were conducted did not affect approach behavior (Table 2).

Table 2 Test results and parameter estimates of the two models M1 and M2. M1 = terms influencing the minimum distance to the speaker (R-squared = 0.250), M2 = terms influencing the number of mobbing vocalizations (R-squared = 0.105). Temperature (T0 = temperature during the test, T5 = mean temperature during the 5 hours before the test), and hour when the test was performed have been used as independent variables. Significance code: * = < 0.05 
Full size table
Fig. 3
figure 3

Relationship between the mean temperature during the five hours before the playback test and the minimum distance between the focal great tit and the speaker reached during the playback test (N = 60). The confidence band (shaded – 95% CI) and the regression line (bold) have been calculated based on the values predicted by the model

Full size image

The temperature during playback affected the number of mobbing calls emitted by focal individuals (Table 2; Fig. 4). This effect of the temperature was quadratic (Y =—38.376 + 3.472X—0.067X2), with individuals calling more often during mobbing call playback when the temperature averaged 25.9 °C (Max{x} ∈ R(− 38.376 + 3.472X − 0.067X2)). At both high and low temperatures, immediate current temperature (T0) negatively correlated with the number of mobbing calls given by the focal individual. The time of day at which the playbacks were conducted did not affect calling behavior (Table 2).

Fig. 4
figure 4

Relationship between the immediate temperature during the playback test and the number of mobbing vocalizations produced by the focal great tit during the playback (N = 60). The confidence band (shaded – 95% CI) and the regression line (bold) have been calculated based on the values predicted by the model

Full size image

Discussion

Our main finding is that great tits, when hearing playback of conspecific mobbing calls, varied two response measures (approach to the playback speaker, and emission of their own mobbing calls) in ways that covary with ambient temperature. More specifically, temperatures from 5 h prior to tests predicted distances of approach, and temperatures during tests predicted the emission of mobbing calls. In both cases, birds’ responses during the highest temperatures were seen to diminish in intensity. This result indicates that longer term heat is a greater influence on movement response than immediate heat, perhaps because it may cause a cumulative effect of heat stress (Conradie et al. 2019; Fernandez et al. 2021).

The quadratic effect of temperature suggests that the behavioral response of great tits to mobbing calls is the strongest at temperatures of approximately 19.5 °C for the approach behavior (typical average temperature in June for the region) and 26 °C for calling behavior, suggesting an impact of temperature on the mobbing behavior of great tits. This result contrasts with the study of Strauss et al. (2020), who reported that the minimum approach distance of great tits for territory defense behaviors was not affected by temperature. This could be due to the fact that temperature in their study did not exceed 27 °C. Indeed, in our research, mobbing behavior was particularly inhibited during the playback of mobbing calls at high temperatures (Fig. 3, 4). However, we cannot determine whether the lower number of calls and approach distance during high temperatures were due to behavioral flexibility in mobbing behavior (for example, a behavioral trade-off between off-setting heat and other behaviors, sensu du Plessis et al. 2012), or due to a temperature-related physiological mechanism that limited birds’ movements. It is possible that excessive heat leads to a generalized drop in motivation to respond to stimuli. There could also be an increase of sound attenuation during signal propagation at high temperature. Indeed, a previous study showed that air temperature affects the active space of acoustic signal (Larom et al. 1997). Nevertheless, while modification of active space on long-range communication is mainly due to temperature inversions and low surface wind, Harris (1966) showed that signal attenuation for frequency range between 2 and 5 kHz (corresponding to great tit mobbing calls) is impacted only minimally by variation in temperature between 15 and 30 degrees (0.2 to 1 dB of excess attenuation per 100 m with a 50% humidity level). It thus seems unlikely that our results could be explained by temperature-dependent variation in playback signal propagation.

High temperatures can have strong and varied physiological impacts on wildlife (Stillman 2019), leading to mortality in extreme cases (Welbergen et al. 2008; McKechnie et al. 2012). In this study we investigated the impact of high temperatures in the short-term, with a very short latency time. Average temperatures experienced immediately during the test or during the previous few hours did affect behavior, suggesting short-term impacts of temperature on behavior. Short term impacts could lead to longer term effects on survival and development, as have been observed in other species (Bourne et al. 2020a, b).

High temperature events will become increasingly frequent under future climate scenarios. Predicting how these high temperatures and heatwaves will impact animal behavior is therefore crucial (Cunningham et al 2021; Soravia et al 2021). Our study suggests high temperatures are likely to induce, in great tits, behavioral shifts in how birds respond to antipredator signals. The predicted increase in summer heatwave number and intensity may alter mobbing behavior perpetuation in this species in the future, and hence may impact survival rates in birds because this behavior is known to repel predators in great tits (personal obs. MD) and other species (Flasskamp 1994; Pavey and Smyth 1998). It remains unknown what impact temperature will have on the behavior of great tit predators.

Mobbing calls are anti-predator signals that contain critical information for survival (Templeton et al. 2005; Magrath et al. 2015). Failure to detect and interpret the risk-based information encoded in mobbing calls at high temperatures could ultimately lead to reduced survival rates, inducing for instance a decrease in the flight initiation distances with increasing temperatures (Díaz et al. 2021). Fortunately, we did not observe a complete failure at higher temperatures to detect and interpret information encoded in mobbing calls. Indeed, such failure would have induced a complete lack of behavioral response from the targeted great tits, with consequences that could be lethal. Our results suggest instead that tested birds changed their investment in approaching behavior, perhaps so as to be able to direct energetic resources towards staying cool. This result supports an increasing body of research indicating that animals invest less in active behaviors at higher temperature. For example, temperate passerine birds sing earlier with rising temperatures in spring (Bruni et al. 2014), and in pied babblers (Turdoides bicolor), parents provision young less at high temperatures (Wiley and Ridley 2016). Our results are, to our knowledge, the first to provide experimental support in a wild bird population for the hypothesis that high temperatures affect the behavioral responses to alarm calls. In the current context of rapid climate change, we argue that there is an urgent need to quantify the costs of heat exposure in natural populations, in order to improve our understanding of the contexts in which animals change their behavior in response to climate conditions.