In endothermic animals, low ambient temperatures increase energy requirements to maintain high core body temperature (Tb)1. This is particularly problematic for herbivores which often simultaneously face severe food shortages. The result is a direct survival risk that may affect fitness in following seasons2. Daily torpor or hibernation are strategies for down-regulating metabolic rate (MR) and Tb, thereby adjusting energy management when faced with food shortage and/or low ambient temperatures (for review see3,4). An evident cost to both strategies is constrained manoeuvrability and behavioural reactivity. Therefore, daily torpor and hibernation are typically associated with the availability of shelters. However, large non-hibernating mammals, such as red deer (Cervus elaphus), Przewalski’s horse (Equus ferus przewalskii), Alpine ibex (Capra ibex), or Shetland pony (Equus caballus) also show substantial reduction of winter energy expenditure. These mammals do so by employing similar physiological mechanisms compared to hibernators or daily heterotherms, i.e. via a reduction of endogenous heat production and tolerance of lower Tb, particularly in peripheral parts of the body5,6,7,8,9,10,11. Nevertheless, the reduction of core Tb is only minor in these species and manoeuverability is certainly not impaired to the extent present in hibernators and daily heterotherms.

In birds, reduced basal metabolic rate (BMR)12, and associated facultative hypothermia occurs in many species as a reaction to food shortage or low ambient temperatures; one species, the common poorwill (Phalaenoptilus nuttallii) is even known to hibernate (reviewed in4,13). However, most studies of seasonal metabolic adjustments in birds found that BMR is up-regulated during winter (reviewed in14). It has been suggested that seasonal changes in standard metabolic rate (SMR)12 depend on body size, i.e. increasing during winter in small birds, but decreasing in large birds (>200 g15). Hitherto, the empirical evidence for this remains equivocal14 and the picture is even more obscured by seasonal migration, a strategy to escape harsh winter conditions.

Besides extreme temperatures and food shortage, reproduction and moult pose major energetic challenges16. Birds are known to increase energy expenditure during different phases of reproduction, e.g. egg production, incubation, nestling provisioning17,18,19. Two competing hypotheses about energy allocation in seasonally breeding birds have been suggested: The ‘increased demand hypothesis’ predicts that the total energy demand is greater during reproduction than during winter and hence, a seasonal peak in MR is expected during the reproductive phase. Crucially, changes in energy expenditure over the annual cycle strongly depend, according to this hypothesis, on seasonal changes in activity patterns and the energy content of food20,21. However, the geese in our study were ad libitum fed also during winter. Therefore, we can rule out the availability and energy content of the food to be the driver for changes in MR. In line with the ‘increase demand hypothesis’ we would expect little Tb variation across the year, but an increase in MR during the reproductive period.

In contrast, the ‘reallocation hypothesis’ predicts relative little variation in Tb and MR during the annual cycle. During winter, birds are expected to have higher energy expenditure due to increased thermoregulatory needs. This additional energy expenditure is, according to the ‘reallocation hypothesis’, reallocated to reproductive activity during the breeding season, because birds breed when ambient temperatures are moderate22. Hitherto, little empirical support has been produced for either hypothesis21,23. Based on what is known about the winter physiology of non-hibernating mammals, we here suggest ‘winter hypometabolism’ as a third hypothesis for explaining seasonal changes of MR. For large birds, such as geese, this hypothesis predicts a decrease in MR during winter mainly due to reduced endogenous heat production and a tolerance of lower Tb8,24, particularly during the nocturnal rest phase5,7. Further the winter hypometabolism hypothesis predicts higher MR and Tb during summer independent of reproduction, i.e. also occurring in not reproducing birds. Therefore, in contrast to previous studies which mostly considered changes in activity patterns and energy content of food as drivers for changes in MR, we here predict that a change of endogenous heat production, accompanied by a down-regulation of Tb, is the most important modulator of MR.

In the present study, we test these hypotheses via data loggers implanted into 25 individuals of a flock of free-living, non-migrating, individually marked greylag geese (Anser anser). The loggers stored 2-minute means of fH, an acknowledged proxy of field metabolic rate (FMR)25,26,27,28, and abdominal Tb for a maximum of 18 months. FMR is measured in free-living animals under natural conditions, in contrast to BMR and SMR, which measure MR under standardized conditions, e.g. during rest, in a specific environment regarding temperature.


Daily mean heart rate (fH) and body core temperature (Tb) both varied profoundly over the year, with peaks during summer and troughs during winter. The magnitude of these peaks and troughs was similar in reproductive and non-reproductive females and males (Fig. 1). Compared to the summer peaks, mean fH was about 22% lower at the winter troughs. Conversely, despite virtually identical seasonal patterns, average Tb at the winter troughs was only about 1 °C below the summer peaks (Fig. 1). Troughs of fH and Tb occurred during the period of December to March. Peak fH and Tb occurred between July and September for most birds, with the exception of reproductive females, which had highest fH and Tb during the incubation period in April/May (Fig. 1). The period of moult coincided largely with the period of highest fH for males and non-reproductive females (Fig. 1), although this was not a statistically significant effect (p = 0.636, Table 1).

Figure 1
figure 1

Annual course of heart rates and core body temperatures of female (red) and male geese (blue). Plotted values are daily individual means. Periods of incubation and moult are shown as box-plots at the bottom of the upper graphs. Shaded areas indicate 95% confidence belts of the overall mean courses determined by spline fitting (see Methods for details). Horizontal bars within belts indicate 95% confidence limits of the location of local peaks and troughs, vertical bars 95% confidence limits of the height of these peaks and troughs, respectively. If horizontal bars extend from December to January only the part in either of these months is plotted. Dark shaded and white bars: reproductive animals; light shaded and black bars: non-reproductive animals.

Table 1 Results of mixed linear modelling of power transformed daily means of heart rate.

Among all variables and factors tested for independent associations with daily mean fH, daily mean Tb was by far the strongest predictor of fH. Tb correlated positively with fH in both sexes with a slightly stronger effect in females (Table 1). In females the second strongest predictor of fH was photoperiod and the magnitude of this effect was much greater than in males (Table 1). Mean air temperature and the amount of precipitation per day were negatively related to daily mean fH in both sexes, but again with stronger effects of air temperature observed for females (Table 1). During incubation females showed comparatively lower fH after correcting for their higher Tb during this period the same effect was found in the attending male partner (Table 1). Leading goslings after hatching in both sexes was associated with higher fH (Table 1). Age had no discernible effect on fH in either sex. The percentage of won interactions during aggressive interactions had only weak effects. Interestingly, this indicator of rank in the flock was positively associated with fH in males but negatively in females (Table 1).

The correspondence between fH and Tb on an annual scale (Fig. 1) was also evident on a daily scale (Fig. 2). The pattern found in hourly means of the fH and Tb during summer (July/August) and winter days (December/January) resembled the typical courses of fH and Tb of diurnal animals, i.e. lower fH and Tb during the night. However, Tb was significantly lower over the day during December/January as compared to July/August. The same difference was present in fH except during early morning and late afternoon hours in December/January when pronounced peaks of fH occurred (Fig. 2). Both, fH and Tb were highest in incubating females with apparently minor differences between day and night, in contrast to the patterns found in their mates over the same time period (Fig. 2).

Figure 2
figure 2

Daily course of heart rates and core body temperatures of female and male geese during December and January (blue), July and August (red), and of incubating females and their mates (orange). Plotted are hourly means of heart rate and body temperature. Error bars represent 95% confidence intervals of these means and reflect variation between individuals. Vertical bars indicate the range of onset and end of civil twilight in the morning and evening, respectively, at the roosting site Almsee. Open bars: December/January); hatched bars: July/August.


The changes in heart rate (fH) recorded over the annual cycle indicate a reduction of field metabolic rate (FMR) during winter in the observed population of greylag geese. Among all variables and factors tested for independent associations with daily mean fH, daily mean Tb was by far the strongest predictor of fH. Such a correlated winter drop in Tb and fH contradicts both, the ‘increased demand hypothesis’ and the ‘reallocation hypothesis’; but supports our novel hypothesis of ‘winter hypometabolism’. Among the potential predictors of fH tested, Tb had by far the strongest effect on fH in both sexes (Table 1). This result suggests that the reduction of fH is mainly due to a decreased endogenous heat production and simultaneous down-regulation of Tb. The closely associated patterns of fH and Tb evident over the day and over the year support this view. Changes in locomotor activity may also have contributed to changes of fH and Tb10,11,29. As we have no quantitative simultaneous measurements of activity along with the measurement of fH and Tb, we could not include this factor into our analysis. Still, activity may significantly contribute to seasonal changes of fH, although the contribution was presumably minor as compared to that of Tb, as has been found in other species5,7,8,9,24. Furthermore, the higher morning and afternoon peaks of fH during January and December, as compared to July and August, most likely resulted from increased activity triggered by our feeding schedule. The food provided during the winter months was particularly attractive to the geese and likely caused increased locomotor activity during the two daily feeding periods29. Together with heat increment of feeding this may have constrained the otherwise more continuous metabolic depression during winter.

A profound decrease of FMR during winter, however, was found despite our ad libitum feeding, indicating that photoperiod rather than food availability is the major cue governing the annual cycle of fH and Tb30,31,32. After correcting for other factors, photoperiod indeed showed the second-strongest influence on fH in females and the third-strongest influence in males. Independent of these effects, unfavourable weather conditions, i.e. low air temperatures and considerable precipitation also increased fH33,34. However, high precipitation increased mean fH similarly in both sexes, whereas low air temperatures had a greater effect on female than male fH. This sex difference may be due to lower body mass of females35,36, rendering them more prone to heat loss.

We observed a modulation of endogenous heat production not only as a result of photoperiod-induced seasonal acclimatization and unfavourable weather conditions, but also in response to other energetically costly situations, such as reproduction. For example, we found a significantly earlier annual peak of fH in incubating females as compared to non-incubating females. The higher energy expenditure of incubating females in contrast to non-incubating birds apparently results from the need to maintain high endogenous heat production and Tb even through the night in order to keep eggs warm. Furthermore, the energetic cost of, again photoperiod-induced, follicle development may have also contributed to the spring peak of fH in reproducing females37. On a first glance, these results seem to support the ‘increased demand hypothesis’ and parallel previous findings of an increased energy expenditure during reproduction in female birds, for example in great tits Parus major38, Arabian babblers Turdoides squamiceps39, or macaroni penguins Eudyptes chrysolophus40. However, it has to be pointed out that in contrast to females, the incubation period in males was characterized by low fH. This presumably reflects the low energy expenditure due to decreased locomotor activity, which is not only true for the incubating female, but also for the attending partner, as in greylag geese, the male pair-partner tends to stay close to the incubating female during the entire incubation period. Also, contrary to what would be predicted by the ‘increased demand hypothesis’, Tb during incubation actually increased in females. This again supports the ‘hypometabolism hypothesis’ predicting that MR is more strongly associated with Tb, rather than locomotor activity and caloric intake.

In geese, only females incubate the eggs and brood the offspring. It has been argued that male investment into reproduction mostly consists of vigilance and agonistic behaviour against conspecific competitors, which ensures paternity on side of the male41, and access to resources such as food and nesting sites on the female side42. However, the energetic cost of such behaviour was apparently low, because we did not find a spring peak of fH in males but a seasonal course similar to that of non-reproductive females (Fig. 1). Moreover, leading goslings was associated with low fH in males. Therefore, there is obviously no ‘increased demand’ for males during the period of reproduction, in contrast to females.

Further, in all our geese, we found a late summer high of fH (Fig. 1). This peak was not associated with moult as in Barnacle geese (Branta leucopsis)43. The period of moult roughly coincided with the parental period in the birds studied, and had no discernible effect on fH independent of other factors tested. Barnacle geese, however, moult later than greylag geese and the coincidence of the annual peak of MR with moult in Barnacle geese may therefore be spurious. In greylag geese, other factors than reproduction or moult must be responsible for the late summer peak of fH. We suggest pre-migration fattening44 as a possible explanation for this, i.e. an ‘increased demand’, but later in the year as originally proposed by this hypothesis. Increased frequency of social interactions in the newly formed summer flock45 may further contribute to the late summer peak of fH. Finally, we found a sex-dependent effect of dominance rank, i.e. percentage of aggressive interactions won, onto fH: The higher-ranking a male the higher fH. Although this effect was minor, it supports previous results indicating sex-specific social investment in geese46.

In conclusion, our results indicate that the pronounced changes of fH during the year are mainly caused by photoperiod-induced changes of endogenous heat production. Tolerance of lower Tb during winter is a major factor in this. Hence, our results support our ‘hypometabolism hypothesis’. Comparison with males and non-reproducing females demonstrates that in reproducing females Tb and FMR is further elevated during incubation, superimposed on baseline seasonal changes.


Study population

A non-migratory flock of greylag geese was introduced in the Almtal (Upper Austria) by late Konrad Lorenz in 1973. The geese are unrestrained and roam the valley between the Konrad Lorenz Forschungsstelle (47°48′N, 13°56′E) and a lake approximately 10 km to the South (Almsee; 47°44′N 13°57′E), where they roost at night. At the time of our study the flock consisted of approximately 170 individuals, marked with coloured leg bands for identification. The flock was supplemented with pellets and grain twice daily (0800–0900; 1500–1700), creating a condition of nearly ad libitum food provisioning year round. Both hand-raised and goose-raised flock members were habituated to the close presence of humans and neither show avoidance if approached up to 1 m distance, nor did they significantly change fH when familiar humans approached47.

Heart rate and body temperature measurements

Twenty-five individuals (8 females/17 males) were fitted with a battery-powered electronic package (60 × 30 × 11 mm, ~60 g; for further details see36), fully implanted into the abdominal cavity. Packages contained a VHF-transmitter, battery, antenna, a temperature sensor, memory, and were encapsulated in a physiologically inert medical-grade silicone rubber. Two electrode plates (surgical steel, approximately 8 mm in diameter and 1 mm thick), fixed towards the lower rib cage in the vicinity of the heart, detected the QRS complex of the electrocardiogram and were connected to the electronic package with approximately 10 cm long coiled silicone rubber-insulated wire of multistranded stainless steel, fitted into silicone rubber tubing to form an elastic and flexible lead. Heartbeats were counted by a microcontroller and average values for consecutive 2-minute intervals were stored on board. Accuracy of fH measurement was ± 2 beats per minute (bpm). Output of the temperature sensor was also stored on board at the end of each consecutive 2- minute interval. Each sensor was calibrated before implantation in the range of 35–41 °C. We were able to re-calibrate 6 devices after surgical removal upon termination of the experiment. We could not detect a significant drift in any of these devices.


All procedures performed in this study involving animals were in accordance with Austrian law and the institutional guidelines of the University of Veterinary Medicine, Vienna. Implantations were performed by an experienced team of veterinarians48 under the animal experiment license issued by the Austrian Ministry of Science GZ68.210/41-BrGT/2003).

Weather data

Daily mean air temperature and per day-accumulated precipitation was provided by a weather station at the Almsee, run by the ‘Hyrdographischer Dienst’ of Upper Austria.


For each day of the year we calculated onset and end of civil twilight at the roosting site Almsee according to to determine minutes of daylight and night and to calculate the length of photoperiod.


Presence of each individual goose and pair-bond status was monitored every other day in a standard way by experienced observers49. Additionally, during the reproductive season, nests were monitored and information on numbers of eggs, onset of incubation, hatching date and numbers of hatchlings was collected. For each family, the number of young (goslings) was recorded every day until fledging of the young, at approximately three months of age. When focal individuals had dependent offspring, this parental period of the year was defined as ‘leading goslings’49. Lengths of the incubation periods were determined from these observations for each breeding female and its partner. We considered a female and its pair-partner as reproductive during a year when the female was incubating, regardless of whether goslings hatched and survived. Data from reproductive individuals were not available for the months September to December (Fig. 1). A couple of weeks after hatching of the young, geese moult their wing feathers and are unable to fly for a few weeks50. Duration of the moulting period was determined for each individual of the flock by observations every other day.

The percentage of won aggressive encounters of each individual was evaluated during summer, fall, and mating season by observing social interactions before and during the daily feedings on five days per period (for details see51). The overall percentage was used as an indicator of an individual’s dominance rank in the flock in a given period.

Data analyses

For the analyses presented here, we used fH and Tb data downloaded from the surgically removed electronic packages. We discarded the first days of measurements after implantation of a device because of potential aftermaths of surgery on fH and Tb. Periods before surgical removal of the transmitters, containing obvious outlier values, which are likely due to dwindling battery power, were also discarded. Raw fH data were purged with a moving average filter to remove biologically implausible outlier values. All statistical analyses were preformed using R 3.3.052.

We tested continuous and categorical variables potentially associated with seasonal changes of fH in linear mixed models (R-package ‘nlme’53), with daily mean fH calculated for each individual as response variable. We explicitly included photoperiod as a predictor of fH to correct for the influence of seasonal differences in day length and hence the period of daily activity and associated fH and Tb in these diurnal birds. By including 2-way interactions of all predictors with sex we further tested whether effects differed between females and males. For these analyses, we power-transformed daily mean fH in order to achieve a normal distribution of residuals.

For visualizing seasonal changes of fH and Tb, we calculated general additive mixed models (R-package ‘mcgv’54 with spline fits to day of the year as predictor variable of fH and Tb for reproductive and non-reproductive females and males, and 95% confidence intervals (CI) of these fits. CI of the timing and height of local trough and peak fH and Tb, respectively, were determined by bootstrapping from respective distributions of troughs and peaks produced by simulating 10000 replicates of model coefficient vectors from the posterior using ‘mvrnorm’ from R-package MASS55.

In all models, we accounted for repeated measures by including individual identity as a random intercept factor and corrected for temporal correlation detected in the residual error term of models by including an auto-regressive correlation structure (‘corARMA’ from R-package ‘nlme’).