Freezing heat : Thermally imposed constraints on the daily activity patterns of a free-ranging grassland bird

Heat stress is a risk for birds exposed to high ambient temperatures, especially for those that live in open environments with limited protection from direct sun radiation. This makes them particularly vulnerable to climate warming. We studied how ambient temperature affects the daily activity of a threatened grassland bird, the little bustard Tetrax tetrax. The activity of 20 birds tracked by GPS satellite telemetry between 2009 and 2012 was monitored throughout the yearly cycle in the Iberian Peninsula. We found that temperatures over ;258C strongly inhibited the activity of birds during the breeding and postbreeding seasons. High temperatures were mostly frequent during mid-day, which often forced birds to reduce activity during this period, especially during summer. We show that the expected future rise in temperatures may result in a substantial reduction in the duration of the period of the day thermally adequate for maintaining activity. With climate change inactivity levels during breeding and post-breeding are expected to rise 37% and 59%, respectivly, compared to present standards. This may pose serious time constraints, particularly on functions like breeding and foraging, with potential consequences for individual fitness and population dynamics. Global warming may thus affect the range of birds not just through habitat change but also by limiting their activity.


INTRODUCTION
Climate warming has been related to biogeographic range shifts and phenological changes of many terrestrial species (e.g., Parmesan 2006, Burrows et al. 2011).However the mechanistic causes of these temperature related phenomena are still greatly unknown, with few examples directly linking species' physiology with vulnerability to climate change (Calosi et al. 2008, Cahill et al. 2013), particularly with endotherms (Huey et al. 2012).A better knowledge on the species thermal tolerance limits and acclimatory capacity will improve the capacity to predict species distribution shifts and assess how well these species will perform within warmer ecosystems (Somero 2010).
In arid zones species have high metabolic levels, requiring efficient mechanisms of temperature regulation (e.g., Rastogui 2007).Besides physiological challenges to preventing overheating, behavioral thermoregulation plays an important role in regulating body temperature within thermally heterogeneous environments (Dawson and Whittow 1999).Temperature refugia can play an important role in avoiding overheating, but, on the other hand, it may also deprive animals of foraging or to perform other costly metabolic functions such as reproduction, which may undermine population growth rate and raise the risk of extinction (Sinervo et al. 2010).However, in open habitats with part of the ground surface exposed to solar radiation, thermal heterogeneity can be limited during the warmest periods of the day (Grant et al. 1988).
A few studies carried out in open grasslands with birds suggest that they significantly reduce their activity during hot periods or perform seasonal movements to escape the summer heat (Combreau andLaunay 1996, Alonso et al. 2009).The generality of these findings still needs to be assessed, but such responses to high temperatures may have profound implications on the life history traits of grassland birds.Knowledge of how birds cope with high temperatures is valuable to understanding their potential responses to the predicted rises in mean ambient temperature and frequency of extreme heat events due to climate change (Christensen et al. 2013).
Animal activity patterns are often adaptive responses to their environment, and are thought to affect fitness (Daan and Aschoff 1982, Kronfeld-Schor and Dayan 2003, Lehmann et al. 2012).Although many circadian rhythms are endogenous, and may persist in the absence of environmental cues (Aschoff 1966, Santiago-Quesada et al. 2012), they are often adjusted to local environmental conditions.Factors known to influence these rhythms include, among others, temperature, food availability, predation risk, competition and level of disturbance (Pienkowski 1983, Burger and Gochfeld 1991, Olsson et al. 2000, Van Der Veen 2000, Brandt and Cresswell 2009).Temporal patterns in animal activity during the circadian cycle are quite variable; many species have a single peak of activity in a 24-hour period, although two peaks or even more are also common (Aschoff 1966).The bimodal activity pattern has classically been related, in arid zone species, to variations in ambient temperature or solar radiation (Combreau andLaunay 1996, McMaster andDowns 2013).The capacity to behaviorally adapt to different climatic conditions through the regula-tion of biological rhythms may have ecological and evolutionary implications for many species (Kronfeld-Schor and Dayan 2003).As the climate is changing quickly in many regions, species may have to adapt these rhythms in the near future.
Most studies of bird activity patterns have been done with animals in captivity (e.g., Donkoh 1989).Those carried out in the wild are often based on focal observations restricted to the daytime and are limited to part of the yearly cycle (e.g., Hidalgo de Trucios and Carranza 1991, Speakman et al. 2000).
The main goal of this study was to analyze the daily activity patterns of a free-ranging grassland Mediterranean bird, the little bustard Tetrax tetrax, across seasons and investigate how ambient temperature may affect these patterns.If high summer temperatures pose a thermoregulatory challenge to little bustards, we predicted that activity will decrease at peak temperatures.Additionally, we analyzed how the projected scenarios of temperature rise due to climate change may affect this species' activity patterns.The analysis of combined data on activity and temperature can bring important insights into the species' phenotypic plasticity and its capacity to cope with future global warming.

Study area and species
Little bustards were captured in Castro Verde (Alentejo region), the most important breeding site for the little bustard population in Portugal (Silva et al. 2006).The area is dominated by extensive cereal farming and pastures.This region has one of the highest breeding densities known for the species (Moreira and Leitão 1996).It has a Mediterranean climate, with fairly cold winters and hot summers (AEMET and IM 2011).
The little bustard is a medium-sized grassland bird of the Bustard family (Otididae) which lives in the open dry grasslands and extensive arable lands of Asia and southern Europe (Iñigo and Barov 2010).It is listed as Near Threatened because it is considered to be experiencing a moderately rapid overall population decline (BirdLife International 2014).Most of its Western European population is now concentrated in the Iberian Peninsula, which is an important global stronghold for the species (Iñigo and Barov 2010).
In Iberia it is widely distributed, and reaches its highest known densities in the grasslands of Alentejo, in southern Portugal (Moreira et al. 2012).
This species breeds in an exploded lekking system in which males display in loosely clustered sites that are attended by females for the single purpose of mating (Schulz 1985, Jiguet et al. 2000).Parental care is carried out exclusively by the females.The diet of chicks is dominated by insects during the first weeks of life, while adults feed mostly on green plants throughout the year (Jiguet 2002).In Castro Verde, during the dry summer (corresponding to the post-breeding period) most of the vegetation dries off, leading to regional migratory movements towards areas with greater food availability (Silva et al. 2007).In autumn or early winter they return to their breeding grounds (Silva et al. 2007) where they congregate in flocks until the onset of the following breeding season (Silva et al. 2004).

Capture, tracking and activity patterns
Between 2009 and 2011, 20 adult male little bustards were captured in Castro Verde and fitted with 30 gr solar GPS PTT (Platform Transmitter Terminal, Microwave Telemetry).Capturing females is more difficult due to their secretive nature.Most captures took place at the beginning of the breeding season in early May.Birds were captured with snares, using a stuffed female as a decoy.Handling time was less than 20 min., to reduce the risk of capture myopathy (Ponjoan et al. 2008).The total weight of the tag and harness was less than 4.7% of the mass of the bird, following the recommendations of Kenward (2001) and in accordance with the license issued by the national nature conservation agency (ICNF).PTT were set to obtain a GPS location every 2 h.Individuals were tracked for a variable period, between April 2009 and August 2012 (Appendix A).
For the purpose of the present study we selected the core periods of three distinct biological phases of the yearly cycle (Schulz 1985): breeding (April-May), post-breeding (July-August) and winter (December-January).Spatial movement is a key element of animal activity (Wilson et al. 2006), and we used it as a proxy of high and low activity.We considered that a bird was active if two consecutive locations were separated by at least 24 m.We used this distance threshold because, using static PTTs located in the grasslands used by the species, we found that shorter separations could be due to GPS location errors.In fact, only 83.3% of 360 locations of these static PTTs were within the 18m of potential error indicated by the manufacturer (Microwave Telemetry 2014).Considering an error of 24 m included 91.4% of the locations.

Data analysis
We started by estimating the relationships between activity level (presence or absence of activity), as the response variable and two explanatory variables: time of day (the second hour of each 2-h period) and temperature (in degrees Celsius, registered at the end of each 2-h period at the nearest weather station).Generalized Additive Mixed Models (GAMM) (Wood 2006, Zuur et al. 2009) were used because an exploratory data analysis suggested that some relationships between response and explanatory variables were non-linear.This analysis was carried out separately for the three phases of the life cycle referred to above (breeding, postbreeding and winter).The mgcv package and 'gamm' function (Wood 2006) were used to fit GAMM in R 2.15.1 (R Development Core Team 2012), using thin plate regression splines and with the optimal amount of smoothing estimated by generalized cross validation (Wood 2006).A basis dimension of k ¼ 4 was set for the smoother for temperature, to allow for some complexity in the function while providing a realistic prediction of temperature effects and avoiding overfitting of the data.To further minimize the risks of over-fitting, a gamma value of 1.4 was set, as recommended by Wood (2006).Activity was modelled using a binomial distribution and a logit link function.Both bird individual code and day (nested on bird) were included as random effects.Temporal correlation in model residuals was allowed by using an autoregressive model of order 1 (AR-1), which improved the AIC of the model (Zuur et al. 2009).Confidence regions for each function were estimated based on 95% Bayesian credible intervals (Wood 2006).Model adequacy was evaluated by plotting residuals versus fitted values and explanatory variables.Model fit was evaluated by the adjusted r 2 and by using the Area Under the Curve (AUC) generated by the Receiver Operating Characteristic (ROC; Pearce and Ferrier 2000), estimated using the ROCR package (Sing et al. 2005).
A GAM between time of day and temperature showed an obvious association between both variables (adjusted R 2 of 0.40 for breeding, 0.33 for post breeding, and 0.68 for winter), at least for specific daylight time periods, which hindered the estimation of the net temperature effect.To overcome this potential problem, the effect of temperature on activity level was estimated separately for each time of day during daylight hours, again using GAMM with bird individual code as a random effect.To account for possible type I error effects due to multiple comparisons, a Bonferroni correction (Miller 1991) for the number of models being calculated in each season (11 or 12) was applied for estimating the significance levels of the temperature smoother.
The models relating temperature to activity identified the temperature values above which the activity of the little bustard was substantially inhibited.To determine the potential impact of global warming on activity, we graphed the proportion of observations spent by the tracked birds above that temperature threshold representing the observed level of activity, and the equivalent proportion of activity if temperatures rose as predicted by the recent IPCC assessment (Christensen et al. 2013).This comparison was done only for daylight hours, from dawn to dusk, which is when the birds carry out most of their activity (present work; Schulz 1985, Jiguet andBretagnolle 2001).We therefore identified for both breeding and post-breeding seasons which time slots registered by the PTT was predomi-nant with civil twilight.Temperature projections of the mean surface air temperature were based on the CMIP5 global model for the RCP4.5 scenario and considered the maximum response estimates of the models that were used to forecast the period of 2081-2100 of summer and annual temperatures (Christensen et al. 2013).We then simulated the proportion of the day that the bird was inactive considering a temperature raise of 58C for summer and 48C for spring (as no predictions are available for spring, we used the mean annual temperature increase).These increase estimates were added to all temperature records and the proportion of observations above the threshold was recalculated.

Influence of time of day and temperature on activity
Minimum, mean and maximum temperatures during the breeding, post-breeding and winter periods show a large variability, roughly 08 to 408C (Table 1) and reflect the Mediterranean character of the study area.
The results of the GAMM show significant effects of time of day (hour) and daily mean temperature on activity patterns of little bustards in all seasons except winter, when the effect of temperature was negligible (Table 2).Overall, the two predictors included in the model explained 12-35% of the variance in activity patterns, depending on the season (Table 2).During both the breeding and post-breeding periods, daily activity showed a bimodal pattern with two peaks of activity, one soon after sunrise and one in the evening.However, while in the breeding season the mid-day reduction in activity was slight, during the post-breeding period it was highly accentuated (Fig. 1).This resting period coincided with the hottest part of the day (Appendix B).In winter, short days resulted in a contraction of the total period of activity, with no evident activity peaks (Fig. 1).Temperature had contrasting impacts across seasons: an increase in temperature was associated with lower activity levels during breeding and postbreeding, particularly over ;258C, but during winter was associated with a slight increase in activity (Fig. 1).
The effect of temperature for each 2-h period assessed for winter, breeding and post-breeding seasons in shown in Fig. 2.During winter there was evidence of a positive effect of temperature on activity levels, although it was only significant for the early morning period (Fig. 2).During breeding, significant temperature effects were registered for the four time periods between 7:00 and 17:00, although the shape of the fitted curve varied from a monotonic positive effect in early morning (7:00-9:00) to quadratic effects at the other times of day, with activity increasing until ;20-258C and decreasing thereafter (Fig. 2).
During the post-breeding season, little bustard activity levels were drastically reduced with increasing temperatures throughout most of the day (from 9:00 to 17:00).

Potential consequences of global warming on activity
Regardless of the season or time of day, our models consistently indicated that temperatures above ;258C strongly inhibited little bustard activity.During the breeding season the proportion of observations above this threshold reached 22% (Fig. 3).However, if temperatures were to rise by about 48C, as predicted by climate change models, the same animals, in the same locations would reach levels of inactivity of 45%.The climate warming scenario represents an overall raise of their inactivity levels of 37% during the day period (including civil crepuscular) (Fig. 3) compared with the present situation.
During the post-breeding season the proportion of observations of our tracked birds above 258C reached 93% during the hottest hours of the day (Fig. 3).This is in agreement with the dramatic reduction of activity observed during mid-day (Fig. 1).With climate warming and temperatures increasing by 58C, at the same locations of our study birds, the expected proportion of observations would reach a level of inactivity of over 90% for a 10-h period.Overall, inactivity levels would raise 59% between dawn and dusk (Fig. 3) compared to today's scenario.

Influence of time of day and temperature on activity
Our work is the first to quantify how temperature can affect daily and seasonal activity patterns based on a dataset collected from freeranging birds and throughout the yearly cycle.It has been suggested that the dry and hot summer weather of Southern Europe has an indirect impact on the little bustard (through a detrimental effect on habitat quality; Delgado andMoreira 2000, Delgado et al. 2009), and we now provide evidence that high temperatures directly constrain its periods of activity.
Changes in seasonal activity patterns of little bustards, from bimodal in breeding and postbreeding periods to a roughly unimodal pattern in winter, may partly be explained by behavioral adaptations to gradual transitions in the annual cycle of photoperiod and temperature.Little bustard daily activity is mostly restricted to the hours of light, and day length may influence seasonal activity.Short days in winter may force birds to remain active throughout the day to fulfil their energetic/foraging needs, or they may remain active simply because air temperatures never reach limiting levels.
In the longer spring and summer days, temperature seems to directly influence daily and seasonal routines.During the breeding season, courtship is the main activity of male little bustards (Jiguet and Bretagnolle 2001) which presumably demands high energy expenditure (Vehrencamp et al. 1989).Our results suggest that the breeding activity of males sharply declines when air temperature 258C, probably due to an increase in energetic costs when displaying under high temperatures or to minimize the risk of overheating.Hidalgo de Trucios and Carranza (1991) showed that the courtship behavior of great bustards Otis tarda also follows a bimodal daily pattern and that v www.esajournals.orgmales select temperatures between 158 and 218C to display.Therefore, temperature may be an important element of the conditions for display (Hidalgo de Trucios and Carranza 1991), with even moderately high mid-day temperatures inhibiting courtship behavior.Schulz (1985) observed an increased conspicuity of little bustards in Alentejo (southern Portugal) during the first three hours after dawn and the last three hours before dusk, a pattern that is also compatible with a reduction of mid-day activity to avoid high temperatures.
During the post-breeding period, in the hot Mediterranean summer, high temperatures seem to constrain little bustard activity even more than during the breeding season.This is evidenced by the much more accentuated and long period of inactivity during mid-day (Fig. 1), and by the Fig. 2. Effect of temperature (8C) on little bustard activity levels during the breeding (April-May), postbreeding (July-August) and winter (December-January) periods, for the times of day where a significant smoother (after Bonferroni correction) was obtained from generalized additive mixed model analysis.Ticks in the x-axis represent the location of observations along the temperature predictor.The shaded areas represent 6 95% confidence envelopes.
v www.esajournals.orggreater reduction in activity in response to increasing air temperatures 2).In fact, while during the breeding season air temperatures seldom reached the apparently critical 258C threshold, they did so on virtually all summer days (Fig. 3).Activity during summer may partly be driven by the needs of thermoregulation, as high air temperatures pose a physiological conflict between evaporating water to maintain body temperature and the need to conserve water and avoid dehydration (McKechnie and Wolf 2010).In hot environments birds often minimize their heat loads and requirements for evaporative cooling by remaining inactive and in the shade of vegetation during the hottest periods of the day (Dawson and Whittow 1999).However this might prove difficult in open habitats since they may be forced to remain exposed to solar radiation at midday (Hueyet al. 2012).So thermal tolerance seems to be important for grassland birds, which varies depending on their morphology, physiology and behavior (Rastogui 2007).
Bimodal routines in birds have been explained as a response to environmental factors, such as starvation-predation trade-offs (Houston et al. 1993, McNamara et al. 1994, Van Der Veen 2000), food availability (Hutto 1981), digestive constraints (Ward 1978) and temperature (Brandt and Cresswell 2009).However, some authors consider them to be a persistent property of the circadian oscillating system, independent of daily changes in environmental conditions (Aschoff 1966, Santiago-Quesada et al. 2012).While our study cannot rule out the presence of such a persistent rhythm in the little bustard, it does demonstrate that environmental conditions, particularly temperature, play an important role in shaping the patterns of daily activity.This is evident in the strong negative relationship between temperature and activity.Moreover, the disappearance of the clear bimodal pattern in winter indicates a high behavioral flexibility that was also reported for other species of bustards (Mian 1988, Martı ´nez 2000).
Besides day length and temperature, seasonal changes in food availability, habitat composition, timing of life cycle events, and disturbance events might influence daily activity patterns of little bustards, and thus contribute to the unexplained variability in the models which we obtained.In addition, the fact that temperatures were not measured in the exact bird locations (in some cases the nearest weather station was a few Fig. 3. Comparison of the level of inactivity during breeding (April-May) and post-breeding (July-August) periods, given by the proportion of observations spent by the tracked birds above the temperature threshold of 258C (solid line), and the equivalent proportion with the temperature rise predicted by the IPCC assessment (dashed line) representing the future expected level of inactivity.The scenarios of global warming for 2081-2100 predict a maximum temperature increase of 48C during breeding and 58C for the post-breeding season.Light grey bar represents the hours of the day with civil twilight while the dark grey bar represents the dark night hours.The remaining hours correspond to the daylight period.The expected raise of inactivity during the daylight period due to climate warming during breeding will be of 37% while for post-breeding the inactivity raises to 59%.v www.esajournals.orgtens of kilometers away) may have also contributed this unexplained variance.

Potential consequences of global warming
Predicted climate warming is likely to aggravate the daily thermal stress of many grassland birds, possibly constraining energy budgets and negatively impacting their fitness (Huey et al. 2012).According to our models, the expected warming of the western range of the little bustard is likely to result in a substantial reduction of the periods of the day with a temperature adequate for maintaining activity, both during the breeding and post-breeding periods.In spring, increasing temperatures could potentially inhibit courtship behavior during part of the day and thus potentially decrease male ability to attract females and copulation rates.Although some authors have observed courtship display in the Otididae family during the night (Schulz 1985, Jiguet andBretagnolle 2001), activity in this period is highly dependent on moonlight (Combreau and Launay 1996), thus reducing the total time available for displaying.However, the most significant impact of increasing temperatures on little bustard behavior is likely to happen during the post-breeding period.On the vast majority of summer days, little bustards will only have temperatures suitable for maintaining activity in the early morning.A tendency towards decreasing activity levels, through forced resting periods lasting most of the day, is expected as a response to summer heat.
With limiting foraging time during the day, birds could select riskier habitats or locations and reduce vigilance as a trade-off between predation risk and energy intake (Lima and Dill 1990).Behavioral adaptations might also involve feeding at night (Davies 1982), which has been recorded for the houbara bustard (Combreau and Launay 1996).However, nocturnal foraging in birds is thought to be limited by low visual acuity even in bright moonlight and might not compensate the reduction in foraging during the daytime (Pienkowski 1983, Martin 1990).Moreover, foraging at night could increase the risk of predation (Combreau and Launay 1996).Other behavioral adaptations could involve making long flights during the cool of the day (Davies 1982).Previous studies demonstrated that in the post-breeding period little bustard males make regional migrations to benefit from higher food availability in more northern and coastal areas (Silva et al. 2007).However, these summer migrations might also be a form of behavioral thermoregulation, as suggested for the Iberian great bustard (Alonso et al. 2009), particularly when moving towards cooler coastal areas.Nonetheless, whilst migratory movements to escape the summer heat could reduce the likelihood of heat stress, the fitness consequences of moving, for example, to lower quality habitats are unknown.In any case, birds that are limited to habitats exposed to elevated temperatures during the dry season of the year are likely to suffer a negative cumulative effect due to reduced trophic availability and limited activity rate.
Warmer winters, on the other hand, are expected to slightly increase the daily activity of little bustards, particularly in the early mornings, which could potentially increase foraging time during this season.However, this potential effect is quite small when compared with those observed during the breeding and post-breeding seasons.
The rise of temperature has been identified as an indirect cause of the decline and ultimate extinctions of lizard populations, restricting the daily time window of activity and consequently constraining costly metabolic functions, such as growth and breeding (Sinervo et al. 2010).Endotherms have a more indirect response to ambient temperature variation because of their high physiological capacities, buffering environmental heterogeneity (Angilletta 2009).Birds in particular respond to elevated temperatures with high Evaporative Water Loss which is, however, a short-term cooling mechanism (Dawson 1982).Our work shows that elevated temperatures can also significantly restrict the daily activity time window for endotherms, which consequently is also likely to make them vulnerable to climate change.

Fig. 1 .
Fig. 1.Effects of (A) time of day (hour) and (B) ambient temperature (8C) on daily activity patterns of little bustards (solid line) during the breeding (April-May), post-breeding (July-August) and winter (December-January) obtained from generalized additive mixed models (GAMM).The y-axis shows the contribution of the fitted centered smooth terms s (names of the predictor, estimated degrees of freedom) to the response variable (occurrence of activity in each 2-hour period; see Methods).Ticks in the x-axis represent the location of observations along the predictors.The shaded area represents 6 95% confidence envelopes.

Table 1 .
Mean (6 SD), minimum and maximum temperatures (8C) in the breeding (April-May), post-breeding (July-August) and winter (December-January) periods, according to the readings of the meteorological stations next to the locations were the little bustards were tracked (between April 2009 and August 2012).

Table 2 .
Results of the GAMM analysis used to predict the daily activity patterns during the breeding, winter and post-breeding seasons, indicating the estimated degrees of freedom (with F-values and approximate significance) of the smoothers for hour of day and temperature (Appendix B).The intercept (6 SE), adjusted r 2 , and area under the ROC (AUC) are also shown.Sample size refers to the total number of movement records per season.