Ecology, 82(6), 2001, pp. 1642±1654 ᭧ 2001 by the Ecological Society of America

BEHAVIORAL THERMOREGULATION BY : TRADE-OFF BETWEEN SAVING ENERGY AND REDUCING CRYPSIS

LUIS M. CARRASCAL,1,3 JOSE A. DõÂAZ,2 DANIEL L. HUERTAS,2 AND INGRID MOZETICH1 1Museo Nacional de Ciencias Naturales (CSIC), Jose GutieÂrrez Abascal 2, E-28006 Madrid, Spain 2Departamento de BiologõÂa I (Vertebrados), Facultad de BiologõÂa, Universidad Complutense, E-28040 Madrid, Spain

Abstract. We studied the effect of solar radiation on the winter biology of Short-toed Treecreepers brachydactyla inhabiting a montane forest in Spain. We hypothesized that, in temperate latitudes of cold winter climate, with low cloudiness and under windless conditions, should select sunlit sites (i.e., forest sectors or trunk patches with high levels of exposure to sunlight) to reduce the metabolic cost of thermoregulation. At a within- habitat scale, a hypothesis of ``only metabolic bene®ts'' predicts that birds should select sunlit

patches at shade temperatures (Tshade) below the birds' lower critical temperature (Tlc). They should shift to a random use of sunlit and shaded patches at temperatures above Tlc. Alter- natively, there could be added costs (e.g., travel costs, predation risk) in using sunlit patches. If higher visibility leads to diminished crypsis at sunlit patches (``trade-off with predation

risk'' hypothesis), birds should select only shaded patches at Tshade values above Tlc (to enhance crypsis). They should increasingly select sunlit patches as Tshade decreases below Tlc. Treecreepers were selective in their use of sun±shade patches, across different spatial scales. At the among-plots scale, abundance was positively related to the avail- ability of sunlit trunks, after we controlled for the effects of tree density, prey availability, and altitude. At the within-plot scale (i.e., selection of foraging patches on trunks and thick

branches), Ivlev's electivity for sunlit patches decreased linearly as Tshade increased. Birds preferred to forage on sun-exposed surfaces (electivity Ͼ0) when Tshade was lower than ϳ4ЊC, but they tended to forage on shaded surfaces (electivity Ͻ 0) when Tshade was higher than ϳ9ЊC. The selection of sunlit trunk patches at low temperatures was not a by-product of higher food availability, because numbers of prey were much less predictable than temperature as a function of trunk exposure (sun vs. shade), and the negative relationship between use of sunlit patches and temperature remained signi®cant when we controlled for the effects of prey availability. Thus, the selection of sunlit patches at low temperatures can be interpreted as a behavioral thermoregulation strategy allowing birds to save energy. However, the Ivlev's electivity for sunlit patches became negative at temperatures well

below Tlc, and no sunlit patches were used when Tshade ഠ Tlc, which led us to reject the ``only metabolic bene®ts'' hypothesis. Photometric measurements of treecreeper taxidermic mounts realistically positioned on trunk surfaces, and detection times by simulated (human) predators, suggest that treecree- pers were more detectable under direct solar radiation than in deep shade. Crypsis dimin- ished in sunlit patches because of a higher image contrast and an increased difference in perceivable coloration between and background, relative to shaded patches. Average detection times were signi®cantly lower for sunlit mounts. Moreover, focal birds scanned more frequently in sunlit than in shaded patches. Thus, the observed temperature-dependent variation in the selection of sunlit substrata is consistent with the ``trade-off with predation risk'' hypothesis predicting that prey should avoid patches where they are more detectable to potential predators. We interpret the distribution and behavior of treecreepers as indicative of a trade-off between the energy savings (through higher operative temperature and reduced metabolic costs) and the increased risk of predation (through higher visibility and diminished crypsis) afforded by sunlit foraging patches. Key words: avian energetics; behavioral thermoregulation; bird abundance; Certhia brachydac- tyla; crypsis; endotherms; food availability; predation risk; temperature; trade-off; treecreepers.

INTRODUCTION and distribution. Small have high relative The need to maintain a favorable energy balance is energy requirements (Walsberg 1983, Blem 1990, Hed- one of the potential factors in¯uencing bird abundance enstroÈm 1992) because of the small thermal inertia caused by their high surface/volume ratio and the high Manuscript received 21 June 1999; revised 31 March 2000; metabolic cost of ¯ight (9±23 times the basal metabolic accepted 4 May 2000; ®nal version received 3 July 2000. rate; Tatner and Bryant 1986, Carlsson and Moreno 3 E-mail: [email protected] 1992). In winter, these requirements rise as a conse- 1642 June 2001 THERMOREGULATION AND CRYPSIS IN BIRDS 1643 quence of the inverse relationship between metabolic due to overheating or evaporative water loss; Austin rate and ambient temperature (Calder and King 1974, 1976, Weathers and Greene 1998). Moreno et al. 1988). The long, cold winter nights im- Given the shape of the metabolism±temperature re- pose additional energy costs, causing some small pas- lationship (Fig. 1A) and the positive value of ⌬T, the serines to lose Ͼ10% of their body mass overnight (see thermoregulation bene®t of using sunlit patches should Bednekoff et al. 1994 and references therein), and vary with temperature, as shown in Fig. 1B: constantly eventually threaten their winter survival. maximum when Tshade Ͻ Tlc Ϫ⌬T, monotonically de-

Thermoregulation made up ϳ42% of the total daily creasing between Tlc Ϫ⌬T and Tlc, and no bene®t when energy expenditure in aviary-housed Loggerhead Tshade Ն Tlc. Under windless conditions, a theoretically Shrikes (Lanius ludovicianus) with probable low lo- optimal pattern of behavior (using metabolic savings comotory costs and in Chukars (Alectoris chukar) dur- as the only relevant currency for short-term decisions) ing winter time, with daily operative temperatures ϳ5± would imply selecting only sunlit patches when Tshade 25ЊC (Weathers et al. 1984 and Carmi-Winkler et al. is lower than Tlc, because thermoregulation bene®ts on 1987, respectively). Thus, any physiological or behav- sunlit patches are always positive. When Tshade reaches ioral strategy that could reduce the metabolic cost of the thermoneutral range, birds should shift to a random thermoregulation might help to enhance winter surviv- use of sun±shade patches. This shift could follow a al. One such strategy is to gain environmental heat from step function (continuous line in Fig. 1C: ``only met- external sources, particularly radiation, which has a abolic bene®ts'' hypothesis) if there were no added pronounced effect in reducing thermoregulatory costs costs to the use of sunlit patches, an assumption that at low temperatures (Hamilton and Heppner 1967, Lus- is unlikely on two major grounds (for zero±one rules, tick et al. 1970, Ohmart and Lasiewski 1971, De Jong see Stephens and Krebs 1986). First, the availability 1976, Lustick et al. 1978, 1980). Thus, birds inhabiting of sun-exposed surfaces is expected to change with cold environments should select the patches or local- time of day. Early in the morning, for instance, there ities with a higher incidence of solar radiation. This is a low availability of sunlit patches and environmental form of behavioral thermoregulation could be espe- temperatures are low. Therefore, a continuous exploi- cially relevant in mid-latitude montane regions, where tation of sunlit patches is constrained by their scarcity winter temperatures usually remain below 0ЊC, but and associated high travel costs between them. there is a greater availability of calm, longer sunny Second, and perhaps more important, detectability days than in higher latitudes. to potential predators might increase at sunlit patches At a regional, coarse-grained scale, we therefore pre- relative to shaded ones, and this could lead to an in- dict that the winter abundance of small spe- creased risk of predation. Maximum predation risk nat- cies should be positively related to the amount of solar urally occurs at the highest light intensities, and light radiation received by different localities or habitat sec- intensity can be used as a cue for predator avoidance tors. However, such a relationship could be obscured (e.g., Clarke 1983, Watanuki 1986, Endler 1987, Clark by between-plot differences in food availability or hab- and Levy 1988, Kotler et al. 1988, 1994, Wolfe and itat structure, which means that any study designed to Summerlin 1989, Lima and Dill 1990, M. DõÂaz 1992; examine its importance should control statistically for but see Bouskila 1995). However, surprisingly few the effects of these confounding variables. studies document the existence of a trade-off between thermal bene®ts and increased visibility to predators A model for the behavioral selection of sunlit in sunlit locations. Here, we hypothesize that such a vs. shaded patches trade-off might substantially modify the relationship between temperature and selection of sunlit patches by At the within-habitat scale, other factors may in¯u- small birds. To proceed with this argument, we need ence the exploitation of sun±shade patches. To work to assume that detectability increases, and crypsis de- out our predictions, we make the following assump- creases, with light intensity (i.e., exposure to sunlight), tions. (1) It is well known (Calder and King 1974) that, so that predation risk should be higher at sunlit patches. in the absence of convective heat losses (e.g., wind- A second assumption is that predation risk is indepen- speed ഠ0 m/s), metabolic expenditure decreases with dent of temperature, because it is illumination, not tem- increasing temperature at temperatures below the lower perature, that causes birds to be more vulnerable to critical temperature (Tlc), and it remains constant within predators. the thermoneutral range (Fig. 1A). (2) The environ- Figure 1D depicts the variation in the relative bene®t mental temperature experienced by birds is always of exposure to sun at different temperatures. When Tshade higher in sunlit than in shaded patches (⌬T ϭ Tsun Ϫ is lower than Tlc Ϫ⌬T, thermal bene®ts increase mono- Tshade Ͼ 0; ⌬T ഠ 10±15ЊC in our study area; Mozetich tonically with time spent in full sun. At temperatures

1995). (3) The temperatures available in temperate cold in the shade between Tlc Ϫ⌬T and Tlc (e.g., T1 and T2, areas are always below the upper critical temperature with Tlc Ϫ⌬T Ͻ T1 Ͻ T2 Ͻ Tlc), the increase in thermal

(Tuc in winter), which implies that there are no ther- bene®t reaches a plateau at different values of sun ex- moregulatory costs to the use of sunlit patches (e.g., posure, depending on the distance between Tshade ϩ⌬T 1644 LUIS M. CARRASCAL ET AL. Ecology, Vol. 82, No. 6

FIG. 1. A graphic model relating ambient temperature in the shade (Tshade), energy metabolism, and diminished predation risk due to enhanced crypsis at shaded patches with treecreeper exploitation of the environmental mosaic of sun±shade patches. This model applies only to ambient temperatures (both in sunlit patches [Tsun] and in the shade) lower than the upper critical temperature Tuc), such as those prevailing in temperate regions during the winter. Here, EX*1 is the maximum value for the ``metabolic bene®t minus predation cost'' term attained for a given temperature T1; ⌬T is the temperature difference between sunlit and shaded patches; Tlc is the lower critical temperature; and TR is the temperature at which the use of sunlit patches is unselective (i.e., proportional to its availability). Bene®ts refer to any common scale (e.g., energy or survivorship) on which energy saving and predation risk can be measured.

and Tlc. When we combine predation costs due to in- is lower than Tlc Ϫ⌬T, the balance increases to reach creased visibility at sunlit patches with metabolic ben- a maximum at 100% exposure to sun, so that birds e®ts at different temperatures (Fig. 1E), we note that should only use sunlit patches. At temperature T1, the the optimal amount of exposure to sunlight becomes balance increases up to a value EX*1 of exposure to smaller as Tshade approaches Tlc. For example, when Tshade sunlight (the maximum, or optimal sun±shade exposure June 2001 THERMOREGULATION AND CRYPSIS IN BIRDS 1645 that maximizes the trade-off between metabolic bene®t tion structure (e.g., shelter) that might also condition and predation cost), and then decreases as exposure to the behavior of birds. This allowed us to identify a sun keeps rising above EX*1. Finally, when Tshade is in ``pure'' effect of illumination on crypsis by eliminating the range between Tlc and Tuc Ϫ⌬T, birds face only the confusion between visibility and structural safety. costs at sunlit patches (diminished crypsis with no ther- In other situations, the shaded site differed from the mal bene®t), and they should remain in the shade. Rep- sunlit one not only in visibility, but also in the security resenting the values of exposure to sunlight that max- afforded by the structures that provide shade and refuge imize the balance between metabolic bene®ts and pre- (e.g., rock crevices, dense cover below shrubs, etc.; see dation costs (EX*i values for any i temperature) against Pianka 1986, Carrascal et al. 1992, J. A. DõÂaz 1992,

Tshade, we obtain Fig. 1F. MartõÂn and LoÂpez 1999). Thus, if all these assumptions are met, we predict Our speci®c goals were: three important changes in the shape of the relationship 1) To analyze whether the spatial variation in tree- between selectivity for sunlit patches and temperature. creeper abundance is related to the amount of solar First, selectivity for sunlit patches should decrease radiation received by different woodland plots, con- within the range Tlc Ϫ⌬T Յ Tshade Յ Tlc, because the trolling for the effects of food availability and habitat thermal bene®ts of sun exposure decrease as temper- structure. ature approaches Tlc, whereas the risk of predation in- 2) To analyze whether the selection of sunlit trunk creases with exposure to sun. Second, birds within the patches covaries with environmental temperature ac- thermoneutral zone should stay in the shade in order cording to the predictions of a model considering only to enhance crypsis (because Tshade Ն Tlc and sunlit patch- the metabolic bene®ts derived from exposure to sun- es provide no thermal advantage). Finally, the tem- light. perature at which birds randomly use sunlit and shaded 3) To test whether treecreepers are more detectable patches (TR in Fig. 1F) will be lower than Tlc. De- or perceive higher risk in sunlit than in shaded patches, pending on how rapidly the risk of predation increases which would imply a trade-off between the thermal with sun exposure, this displacement toward lower tem- bene®ts of sun exposure and the predation costs derived peratures will be more or less pronounced. Compare from diminished crypsis at sunlit patches. An addi- the predictions derived from the ``trade-off with pre- tional objective was to analyze whether a model con- dation risk'' hypothesis (Fig. 1F) with those derived sidering such a trade-off can successfully predict the from the step function model accounting only for met- observed pattern of covariation between temperature abolic bene®ts (i.e., zero±one rule, Fig. 1C). and selection of sunlit patches. It should be noted that our predictions will neces- sarily be qualitative because of a lack of information METHODS about thermal bene®ts and predation costs on a com- Study area mon scale. Moreover, an animal's Tlc may vary sea- sonally, with time of day, and depending on whether Data were collected in the Guadarrama range (central the animal is in darkness or light, and fasted or replete. Spain, approximately 40Њ47Ј N, 04Њ00Ј W). This is a However, our graphic model (Fig. 1) is able to evaluate montane area mainly covered with coniferous forests the ``only metabolic bene®t'' hypothesis, indepen- of scots pine Pinus sylvestris from 1000 to 2000 m dently of the exact values of parameters Tlc, Tuc and the a.s.l. The axis of the range is west±east oriented, but slope of the relationship of metabolism vs. temperature. there are many valleys with a north±south orientation. Imprecision in the assessment of those parameters af- This land pattern de®nes a considerable amount of var- fects the slope and set point values of the curve in Fig. iation in the cardinal orientations of slopes and the 1F, but not its markedly different graphic pattern as amount of solar radiation received by different forest compared with Fig. 1C. tracts. Average duration of day : night time in this area dur- Objectives ing our study period (December±February) is 9.9: 14.1 To test these predictions, we analyzed a natural, eco- h (daytime range: 9.25±10.5 h, nighttime range: 14.75± logically relevant scenario, considering a well-estab- 13.5 h). During the winter season, the area has a cold- lished physiological basis. Our model organism (a continental mediterranean climate. Mean average tem- small insectivorous passerine) and our ecological sys- perature in December±February of 1997±1998 was temÐ(a cold montane forest in south western Europe) 1.1ЊC at 1850 m a.s.l. (Navacerrada Pass), and mean are appropriate for the questions being addressed. The minimum temperature during this period was Ϫ1.4ЊC. bird of choice is the Short-toed Treecreeper (Certhia Although harsh winter snowstorms in¯uence the cli- brachydactyla), an active insectivorous searcher that mate of the area, anticyclonic conditions predominate forages almost exclusively on tree trunks. On these during the winter season. Thus, 42% of the days in feeding substrates, the availability of sunlit and shaded December±February of 1997±1998 were cloudless, patches varies naturally within and between trunks, with snow precipitation affecting 39% of the days. Feb- without being confounded with differences in vegeta- ruary is usually the sunniest winter month, with up to 1646 LUIS M. CARRASCAL ET AL. Ecology, Vol. 82, No. 6

74% cloudless days in 1997±1998. These data are rep- that these devices are not proper operative temperature resentative of the long-term meteorological conditions thermometers (due to differences in coloration, insu- prevailing at the study area, as indicated by the mean lation, and boundary-layer effects; Bakken et al. 1985, of 46% cloudless daytime hours computed for the Bakken 1989). However, previous measurements whole Guadarrama range (1994±1995, CLIMATE da- showed that, on calm days, copper cylinders and taxi- tabase version 2.1, W. Cramer, Potsdam, Germany, per- dermic mounts (copper models covered by treecreeper sonal communication). skins) eventually reached the same equilibrium tem- Short-toed Treecreepers were censused in pine for- peratures (see also Vitt and Sartorius 1999). Thus, cop- ests throughout the area within the 1500±2000 m a.s.l. per models provided reasonably close approximations altitudinal range. Below 1500 m, the Guadarrama range to the taxidermic mounts under the same conditions of reaches the valleys of the Northern Spanish plateau, radiant heat loads. At equilibrium, the temperatures of providing little variation in the orientation of slopes, copper cylinders and taxidermic mounts were almost and hence no differences in insolation. The behavior perfectly correlated (R2 ϭ 0.9995, n ϭ 15, range 3± of treecreepers exploiting sunlit and shaded trunk sur- 40ЊC; Tmount ϭ 0.49 ϩ 0.99 Tcylinder). Therefore, our mod- faces was sampled in a large area of Valsain-La Granja el temperatures re¯ect and quantify the thermal dif- mixed (pine±oak) forest at 1150±1500 m a.s.l. This ference between sunlit and shaded locations under area is covered by an ancient forest with very large windless, cold, winter conditions. Nevertheless, we ac- trees, many higher than 20 m. Tree density is relatively knowledge that our measurements could be re®ned for low (100±200 trees/ha), thus facilitating the monitoring wind plus temperature effects by incorporating a stan- of long sequences of treecreeper's foraging behavior. dard temperature scale that would predict selection of Finally, photographs of stuffed treecreepers on sunlit sites for increasing solar exposure and wind avoidance. and shaded trunks were taken in Los Asientos (Valsain- Prey availability for treecreepers was estimated by La Granja area). counting all arthropods Ͼ1 mm long found during 2 min at a height of ϳ1.5 m on the selected trunk (see Study species Cooper and Whitmore 1990 and references therein). The Short-toed Treecreeper Certhia brachydactyla is Arachnids and Diptera accounted for Ͼ80% of arthro- a small (ϳ8.2 g), insectivorous bird common in tem- pod numbers both in sunlit and shaded locations. The perate forests of the southwestern Palaearctic. It occurs scarcity of prey in shaded locations precluded any re- at middle and lower middle altitudes of temperate liable statistical comparison of prey size or type be- zones, becoming more montane at the southern edge tween sunlit and shaded patches. However, it is unlikely of its range (Hagemeijer and Blair 1997). It is a strictly that these variables differ markedly between sun and arboreal passerine that forages for arthropods on the shade, because exposure to sun at a given trunk surface trunks of practically any species of tree (Cramp and changes several times within a single day as a conse- Perrins 1993). Winter populations show little evidence quence of the shadows cast by other trees. of movement other than postbreeding dispersal. In the To assess the predictability of model temperature and Guadarrama montane woodlands, winter foraging is re- food availability as a function of exposure to sun (sunlit stricted to the trunks and thick branches of trees (main- or shaded trunk surfaces) and variation among indi- ly Scots pine, Pinus sylvestris, and Pyrenean oak, Quer- vidual trees, we recorded the temperature of the copper cus pyrenaica; Carrascal et al. 1987). The strictly in- tubes and the availability of potential prey on 25 ran- sectivorous diet and the small size of this bird species domly selected trees in the Valsain area (1350 m a.s.l.) make it an adequate model for the objectives of this in February 1998. Estimates were repeated twice on study. the sunlit half of the trunk and twice on its shaded half, The lower critical temperature for the Short-toed thus providing an overall sample size of 100 values for Treecreeper is 21.5ЊC (data from winter-acclimated, temperature and arthropods. postfeeding birds during nighttime; Mozetich 1995), a Signi®cance of the effects was tested using a mixed repeated-measures ANCOVA model, with sun±shade value that falls within the range 20±25ЊC obtained from as the ®xed within-subjects factor, individual tree as allometric equations and from empirical data of phy- the random factor, and time of day as the covariate. logenetically related, similarly sized species (Calder Variation between the two samples from each tree and and King 1974, Kendeigh et al. 1977). sun±shade level de®ned the residual variance of the Arthropod and ambient temperature variation ANCOVA model. The partition of variance was esti- with exposure to sun mated by means of a multiple regression approach to ANCOVA, with time of day, exposure to sun±shade, As an approximation to the environmental temper- and individual tree (these later two factors coded as ature experienced by foraging birds, we used hollow dummy variables) as independent variables. copper cylinders (length 5 cm, diameter 1 cm) closed at both ends except for a small ®ssure that allowed us Censuses to insert the sensing tip of an electronic digital ther- The abundance of treecreepers was estimated by one mometer (digi-thermo; Ϯ0.1ЊC precision). We assume of us (D. L. Huertas), at 20 different 5-ha transects, by June 2001 THERMOREGULATION AND CRYPSIS IN BIRDS 1647 counting all birds detected along two 25-m survey min, in all pine trees used by treecreepers, both in full belts, one on each side of the progression line. Census sun and in the shade. transects were 1 km long (5 ha). Nearly all treecreepers Availability of exploitable trunk surfaces exposed to were ®rst detected by their calls. Treecreepers are not the sun was visually estimated (at 20% intervals) in very noisy birds, but they continuously emit a clear the 10 nearest trunks around the pines exploited by call and are easy to locate (about one call every 2±5 treecreepers. For those focal birds that foraged in more min). Censuses were carried out between 0800 and than one pine, we averaged estimates of sun availability 1200 (GMT, Greenwich mean time), always on calm, over all of the exploited trunks. clear days. Each plot was censused twice (®rst in Jan- In summary, for each focal bird, we estimated the uary and then in February 1998) and we used the mean percentage of time foraging in full sun, the average of the two values as an estimate of treecreeper abun- availability of arthropod prey on sunlit and shaded bark dance. The incidence of solar radiation on each census surfaces, the model temperature in full sun and in the plot was quanti®ed by counting the number of trunks shade, and an approximation of the average availability with at least half of their ``exploitable'' surface exposed of trunk surfaces exposed to the sun. to the sun along a 10 m wide belt at each side of the Selection for sun patches was calculated by means progression line. The exploitable surface of Scots pines of Ivlev's electivity index (Ivlev, 1961): extends over the entire circumference surrounding the IE ϭ (P Ϫ P )/(P ϩ P ) lower part of trunks (ϳ4 m), which has bark that does fafa not ¯ake away. The upper, ¯aked parts of trunks are where Pf is the proportion of time spent foraging in avoided by foraging treecreepers (Carrascal 1988). The the sun and Pa is the proportion of the exploitable trunk number of sunlit trunks was highly correlated with the surface that is exposed to full sun. duration of the daily insolation period (in hours per day) in each census plot (r ϭ 0.915, n ϭ 20, P Ͻ 0.001). Treecreeper detectability on sunlit Thus, plots receiving direct sunlight for a longer daily and shaded trunk surfaces period, also had more sunlit trunks at midday. In ad- We used 20 slides of a stuffed treecreeper simulating dition, we also noted the altitude of each census plot a foraging posture on a tree trunk to examine whether and we controlled two other variables that might also the detectability of treecreepers differed between sunlit condition treecreeper abundance (Carrascal 1987): the and shaded patches (Fig. 2; sample size was 10 in both number of trees Ͼ30 cm in diameter at breast height, illumination conditions). Stuffed birds were obtained counted within a 25 m radius circle every 100 m along from the bird collection at the Museo Nacional de Cien- the transects, and the availability of potential prey, es- cias Naturales (Madrid, Spain). We asked a group of timated by counting arthropods on the nearest trunk at skilled birdwatchers (staff of the Vertebrate section at every 100 m along the transects. the Universidad Complutense of Madrid) to ®nd the bird in the projected slides, and measured their detec- Selection of sunlit and shaded trunk surfaces by tion times. Our aim was to detect differences between foraging treecreepers sun and shade in detectability for humans and perhaps Focal samples of individual treecreepers were ob- other diurnal, visually guided predators, e.g., Sparrow- tained by two of us (L. M. Carrascal and I. Mozetich) hawks Accipiter nisus, weasels Mustela nivalis, and on clear, calm days in three different years (February feral and wild cats Felis spp. We also applied a pho- of 1996, 1997, and 1998). Although focal birds were togrammetric approach to test the hypothesis that such not marked for individual identi®cation, the fact that differences could be based on differences between sun samples were obtained in three different years, and and shade in (1) image contrast and (2) color disparity throughout a large extension of pinewood (ϳ100 ha), between bird and background. reduced the probability of repeatedly sampling the Pictures were taken with a 35-mm tripod-mounted same treecreepers and thus precluded pseudoreplica- camera (Nikon F90N) with a 70±210 mm telezoom. tion. We tried to follow all individual treecreepers The ®lm used was Fujichrome Sensia II 100 ASA. All while we walked throughout the forest. Birds initially pictures were taken with a 81A ®lter. The diaphragm were detected mainly by ear. We followed focal birds was f11 in all pictures to maximize depth of ®eld and from a distance of Ն15 m, without disturbing them, sharpness (especially in slide corners). We used the while they were foraging on pine trunks. Sampling circular markings of the spot-metering area on the times ranged from 1 to 10 min (145 Ϯ 120 s, XÅ Ϯ 1 view®nder as visual guides to establish the relative size SD; n ϭ 36 birds). of treecreepers and to achieve a similar size in all pic- For each focal bird, we used a stopwatch to record tures. The tripod was located at a distance of 4.5±7 m the time spent in full sun and in shaded portions of from the trunk where the stuffed treecreeper was placed trunks. At the end of each sample (truncated at 10 min, (at 1.0±1.7 m above the ground), and the focal length or when it became impossible to keep following the of the lens was zoomed between 135 and 180 mm. The focal bird), we measured model temperatures and picture area was divided into 12 sectors (three vertical counted all arthropods found at breast height during 2 by four horizontal sectors), and we randomly asigned 1648 LUIS M. CARRASCAL ET AL. Ecology, Vol. 82, No. 6

FIG. 2. An example of the stuffed bird slides used to obtain photogrammetric measurements and detection times of simulated (human) predators, both at sunlit (upper) and shaded (lower) locations. treecreeper position to one of these sectors. Therefore, the surface of the taxidermic mount (number of pixels there were no spatial cues to locate treecreepers on ϭ 20,541 Ϯ 455, mean Ϯ 1 SE; ϳ75 ϫ 275 pixels) to slides when pictures were presented to observers. carry out photogrammetric measurements of image Slides were digitized using a Polaroid Sprint Scan contrast. We used Adobe Photoshop v4.0 (Adobe, San 35 plus ®lm scanner (1774 ϫ 1124 pixels; ®le size 5.7 Jose, California, USA) to calculate, for each slide, the Mb; Polaroid Corporation, Cambridge, Massachusetts, interquartile range of the distribution of pixels on a USA). We de®ned an area approximately three times gray-scale single-channel image consisting of up to 256 June 2001 THERMOREGULATION AND CRYPSIS IN BIRDS 1649 levels of gray. This range provides a direct index of image contrast. The dark shadow produced by a bird in full sun, for example, originates a bimodal distri- bution of pixels, with a ®rst peak at the lowest part of the scale (the shaded portion of the image) and a second peak at its brighter half (the sunlit portion). We also measured color differences between bird (mount) and background (trunk surface excluding the mount). Color disparity was estimated as the absolute value of the difference between the bird and trunk mean levels of the distribution of pixels in the red, green, and blue channels (RGB scale) provided by Adobe Pho- toshop v4.0. We tested for the effect of exposure to sun FIG. 3. Regression of treecreeper abundance against the or shade on color difference using a MANOVA model number of sunlit trunks per transect. with red, green, and blue as the dependent variables. Human detection times were tested by passing the than others. In addition, the magnitude of the temper- slides to a group of 14 birdwatchers who were asked ature difference between sun and shade varied among to ®nd (and to con®rm with a laser pointer) the location individual trees. These latter effects are probably re- of the bird from a distance of 4 m. Slides were classi®ed lated to the previous thermal history of each trunk (e.g., according to the situation of the taxidermic mount (in trunks that were in the shade when temperature was the sun or in the shade). Birdwatchers were questioned registered, but were exposed to sunlight a short time one by one. The size of the projected image on the before). The combined effects of exposure to sun, in- screen was 1.5 ϫ 1 m, the actual length of the stuffed dividual tree, and time of day explained 92.3% of the treecreeper being ϳ17 cm. When birds were detected, variance in operative temperatures, of which as much detection times were measured with a stopwatch to the as 82.2% corresponded to differences between sun and nearest 0.5 s, up to a maximum searching time of 1 shade. min. Two factors, exposure to sun and individual tree, had Vigilance and foraging on sunlit and a signi®cant effect on arthropod counts (exposure to shaded trunk surfaces sun: F1,24 ϭ 33.8, P Ͻ 0.001; tree: F24,24 ϭ 2.24, P ϭ 0.027), with larger numbers of prey on sunlit trunk In January±February 1998, we (L. M. Carrascal and surfaces (xÅ ϭ 2.9 arthropods/2 min of search vs. xÅ ϭ J. A. Diaz) observed the behavior of focal birds as they 0.5 in the shade). Neither the interaction between both foraged in the sun (n ϭ 24 sequences) or in full shade factors (F24,25 ϭ 1.7, P ϭ 0.107) nor the effect of time (n ϭ 28 sequences) in a large tract of a mature forest of day (multivariate effect of covariate: Rao's R ϭ 2.7, where treecreepers were very abundant. We tried to df ϭ 2, 23, P ϭ 0.126) was signi®cant. The combined avoid repeated sampling of the same individuals by effects of exposure to sun, individual tree, and time of searching in different forest zones. The duration of be- day explained 51.4% of the variance in arthropod havioral sequences was 53.2 Ϯ 29.4 s (mean Ϯ 1 SD; counts, of which only 18.4% corresponded to differ- range 15±126 s). For each focal bird, we noted sun ences between sun and shade, whereas 32.9% was ac- exposure and time of day, counted the number of pecks, counted for by differences among trees. Therefore, and measured with a stopwatch the total amount of time food availability was less predictable than temperature, spent vigilant and the overall duration of the sequence. on the basis of the simple cue provided by the sun± Scanning behavior was de®ned as the time that a bird shade dichotomy. spent motionless with its beak raised to eye level or On a regional scale (i.e., differences among the 20 higher (for a similar approach, see Hogstad 1988). census plots), radiation and elevation together ex- plained 33% of the among-plots variance in mean num- RESULTS bers of arthropods (F2,17 ϭ 4.2, P ϭ 0.033), of which Predictability of temperature and arthropod 13.6% corresponded to the effect of radiation (partial availability as a function of exposure to sunlight standardized regression coef®cient, ␤ϭ0.380, P ϭ 0.072) and 19.4% to the effect of elevation (␤ϭ Exposure to sun (F ϭ 2289.4), individual tree 1,24 Ϫ0.452, P ϭ 0.036). Thus, the altitudinal location of (F ϭ 9.4), the interaction between both factors 24,24 census plots was a better predictor of food availability (F ϭ 7.2), and time of day (multivariate effect of 24,25 than the incidence of solar radiation. covariate: Rao's R ϭ 7.7, df ϭ 2, 23) had a signi®cant effect on the temperatures available (all P's Ͻ 0.005). Covariation of treecreeper abundance Thus, temperature was consistently higher in the sun and radiation levels (xÅ ϭ 26.7ЊC in the sun vs. xÅ ϭ 12.1ЊC in the shade), The abundance of treecreepers was positively related and some particular trunks had higher temperatures to the availability of sunlit trunks (Fig. 3) when we 1650 LUIS M. CARRASCAL ET AL. Ecology, Vol. 82, No. 6

TABLE 1. Effect of independent variables on treecreeper abundance.

Independent variable ␤ tP% var Log(no. arthropods) 0.16 1.07 0.303 4 No. sunlit trunks 0.73 5.27 Ͻ0.001 59 Altitude 0.38 2.56 0.022 13 No. trunks Ͼ30 cm in diameter Ϫ0.06 0.46 0.650 0 Note: Regression analysis: ␤ϭpartial standardized regression coef®cient; % var ϭ percentage of variance explained, calculated as 100 ϫ␤ϫPearson correlation coef®cient; df ϭ 15. controlled for the effects of tree density, prey avail- el temperature in the shade (r ϭϪ0.778; Fig. 4), in ability, and altitude (Table 1). Treecreeper abundance the sun (r ϭϪ0.742), or average weighted model tem- was also higher at higher altitudes, but it was unrelated perature (r ϭϪ0.787, n ϭ 36 and P Ͻ 0.001 in all to tree density or prey availability (Table 1). These four cases). The regression line of electivity vs. Tshade in- variables explained 76.2% of the variance in treecree- tercepts the 0 value of electivity at 6.5ЊC (95% con- per abundance (F4,15 ϭ 12.0, P Ͻ 0.001), the avail- ®dence interval: 3.9±9.0ЊC). Thus, foraging treecree- ability of sunlit trunks being its most useful predictor pers actively selected sun-exposed patches at low am- (59% of the variance explained). bient temperatures (below ϳ4ЊC in the shade), but they

preferred shaded locations when Tshade was higher than Selection of sunlit and shaded trunk surfaces by ϳ9ЊC. foraging treecreepers Fig. 4 also shows the comparison of our results with The use of sunlit patches by focal treecreepers was those predicted by the hypotheses of ``only metabolic negatively correlated with model temperature (weight- bene®ts'' (zero±one rule; Fig. 1C) and ``trade-off with ed by the availability of sunlit and shaded trunk patch- predation risk'' (Fig. 1F), based on the clear-cut graphic es; partial standardized regression coef®cient, ␤ϭ differences between both patterns. The data clearly lead Ϫ0.787, P Ͻ 0.0001) when we controlled for the effects us to reject the predictions of the ``only metabolic ben- of sun availability (␤ϭ0.111, P ϭ 0.459) and prey e®ts'' model, because electivity for sunlit patches was abundance, both in the sun (␤ϭ0.126, P ϭ 0.437) negative (i.e., birds avoided sunlit patches) at Tshade val- and in the shade (␤ϭ0.038, P ϭ 0.811). These four ues well below the thermoneutral zone. Moreover, tree- variables explained 51% of the variance in sun expo- creepers did not continuously exploit sunlit patches sure among focal birds (F4,31 ϭ 8.07, P Ͻ 0.001). when Tshade was lower than Tlc (ϳ21.5ЊC), and a con- The selection for sunlit trunk patches exerted by for- sistent pattern of shade seeking was observed above aging birds (Ivlev's electivity index) was negatively 7±8ЊC. On the other hand, our results partially mirror correlated with ambient temperature, independent of the predictions of the ``trade-off with predation risk'' the particular temperature statistic employed, i.e., mod- model, because we found a negative relationship be-

tween electivity for sunlit patches and Tshade at tem-

peratures below Tlc. Nevertheless, data were inconsis- tent with predictions in that we did not observe a full selection of sunlit patches at temperatures lower than

Tlc minus ⌬T (see Introduction). Selection for sunlit patches increased as temperature decreased, but no fo- cal bird used only sunlit patches for foraging.

Treecreeper detectability on sunlit and shaded trunk surfaces The location of the stuffed bird in sunlit vs. shaded patches (Fig. 2) had a signi®cant effect on the image contrast of its surrounding area (one-way ANOVA with

image contrast log-transformed: F1,18 ϭ 57.85, P Ͻ FIG. 4. Relationship between treecreepers' electivity for 0.001), the contrast being higher in the sun (inter- sun-exposed bark surfaces and the operative temperature in quartile range on a gray-scale single-channel image shaded patches. The thick, continuous line shows the predic- tions derived from the ``only metabolic bene®ts'' hypothesis consisting of up to 256 levels of gray: xÅ Ϯ 1 SD ϭ 59.4 (zero±one rule). The discontinuous lines show the predictions Ϯ 12.2, n ϭ 10 slides) than in the shade (28.7 Ϯ 6.0, of the ``trade-off with predation risk'' hypothesis. For the n ϭ 10 slides). Color disparity between the bird mount temperature range between 8.0Њ and 21.5ЊC, the dashed line and its background (Fig. 5) was more pronounced in is only approximate, i.e., we predict a negative relationship between electivity and temperature, but the exact shape of full sun than in the shade for the three color channels such a relationship depends on the intensity of the predation (Rao's R ϭ 4.42, P ϭ 0.019, df ϭ 3, 16; Wilks' ␭ϭ pressure. 0.55). Thus, crypsis diminished in the sun not only June 2001 THERMOREGULATION AND CRYPSIS IN BIRDS 1651

between-plots level, treecreeper abundance was posi- tively related to the availability of sunlit trunks (de- termined by geomorphology and cardinal orientation of slopes). At the within-plot level, treecreepers pre- ferred to forage on sunlit bark surfaces of trunks and thick branches when ambient temperature was low

(Tshade lower than ϳ6.5ЊC), and the intensity of the se- lection for sunlit patches was higher at lower temper- atures. These patterns remained unchanged when we controlled holding for the effects of prey availability. Although several authors have shown an association between temperature and the winter distribution of FIG. 5. Effects of the location of bird mounts (in full sun or in the shade) on the color discrepancy between bird and birds (northern limits and range size; Root 1988, Re- bark background. The y-axis shows the absolute value of pasky 1991), little is known about the role that solar differences in color levels in the RGB (red, green, blue) scale. radiation might play in explaining and quantifying their Bars show means Ϯ 1 SE (n ϭ 10 in all cases). distribution and abundance. This study, by document- ing the in¯uence of the radiative environment on the because of the higher image contrast, but also because variation in treecreeper densities under conditions of of an increased difference in perceivable coloration be- cold stress, demonstrates the population effects of the tween bird and background. thermal quality of the habitat (Huey 1991). All mounts in the sun were located by all observers Solar radiation could exert its in¯uence on the winter in less than 1 min, and only three birdwatchers failed biology of treecreepers in three different ways. First, to locate the stuffed bird in less than 1 min when it the radiation-induced increase in temperature could fa- was in the shade. Average detection times were sig- vor a reduction of thermoregulatory costs, because be- ni®cantly lower in the sun (1.2 Ϯ 1.08 s, xÅ Ϯ 1 SD; n low the thermoneutral zone, metabolism falls as tem- ϭ 10 slides; the ®gure for each slide is the mean de- perature rises (Calder and King 1974; see Fig. 1B). tection time for 14 observers) than in the shade (7.0 Second, temperature could have a positive, indirect ef- Ϯ 3.8 s, n ϭ 10; one-way ANOVA with detection times fect on foraging ef®ciency through the activation of arthropods at higher temperatures (Avery and Krebs log-transformed: F1,18 ϭ 24.59, P Ͻ 0.001). 1984, see also Results), which would lead birds in sunlit Vigilance and foraging on sunlit patches to achieve higher foraging success. Finally, the and shaded trunk surfaces use of sunlit patches could also entail a higher predation The proportion of time spent vigilant by foraging risk via enhanced detectability on sun-exposed sub- treecreepers (Table 2) was signi®cantly higher in full strates (Endler 1987). sun than in the shade. Nevertheless, pecking rates did The concomitant effects of solar radiation on ther- not differ signi®cantly between sun and shade loca- moregulation and foraging success were not controlled tions. The same analysis using net pecking rates (cal- experimentally in this study, because all data were col- culated discounting the time spent vigilant; i.e., rough- lected under natural conditions. However, our analyt- ly equivalent to net intake rates), yielded a similar non- ical approach allows us to statistically account for, and signi®cant difference. Therefore, food intake rates did to estimate the magnitude of, each effect. Our conclu- not change with exposure to sun or shade, but tree- sion is that arthropod availability did not signi®cantly creepers seemed to perceive a higher predation risk affect treecreeper distribution and behavior, and that when foraging at sun-exposed locations. the magnitude of its effect was negligible. Three sourc- es of evidence support this conclusion. First, between- DISCUSSION plot differences in food availability had a very low and Our results show that Short-toed Treecreepers were nonsigni®cant in¯uence on treecreeper density (ac- selective in their use of sun±shade patches. This result counting for 4% of variance). Second, exposure to sun was consistent at two different spatial scales. At the or shade accounted for a low amount of variance (18%)

TABLE 2. Proportion of time spent vigilant and pecking rates of treecreepers while foraging in the shade (n ϭ 28 observations) and in full sun (n ϭ 24).

Variable Shade Sun F1,50 P Time vigilant (s/min) 4.5 Ϯ 4.1 10.5 Ϯ 4.4 22.92 Ͻ0.001 Pecking rate (no./min) 10.9 Ϯ 7.9 9.1 Ϯ 5.9 0.42 0.518 Net pecking rate (no./min) 11.6 Ϯ 8.2 11.2 Ϯ 7.9 0.05 0.832 Notes: Net pecking rate was calculated discounting the time spent vigilant. Data are means Ϯ 1 SD. 1652 LUIS M. CARRASCAL ET AL. Ecology, Vol. 82, No. 6 in arthropod counts. Instead, food availability depend- treecreepers while using sunlit patches. Four lines of ed mainly on individual tree characteristics that should evidence support this conclusion. First, results shown be dif®cult for a ¯ying bird to forecast from a distance in Fig. 4 are consistent with the negative relationship prior to its arrival at any given particular trunk. Finally, between electivity for sunlit patches and Tshade predicted the feeding success attained by focal birds did not differ by the ``trade-off with predation risk'' hypothesis. Sec- between sunlit and shaded patches (0.8% variance ex- ond, our photogrammetric approach demonstrates that plained by the effect of patch category (sunlit vs. shad- the image contrast and color discrepancy between bird ed) on pecking rates). mount and trunk background were more pronounced On the other hand, differences in model temperature in full sun than in the shade. Third, such differences between locations in full sun and in the shade were in¯uenced the perceptual abilities of visually guided highly predictable (82% of variance explained) and ``predators'': birds on sunlit patches were easier to de- substantially large (difference of 14.6 Ϯ 6.6ЊC, xÅ ϭ 1 tect by humans, and detection times were shorter in the SD). Considering the equations that relate metabolic sun. Finally, birds spent more time vigilant on sunlit rate to environmental temperature in small passerines patches, which con®rms that they actually perceived a (Fig. 1A), the observed increase of operative temper- higher risk of being detected by potential predators. atures at sunlit patches should allow birds to save up Although predation risk may be traded off against to ϳ45% of their thermoregulatory costs, a ®gure ob- the thermal quality of patches or habitats, there are tained using the average of regression parameters for relatively few studies that document such trade-offs Parus species in Kendeigh et al. (1977). This energy (see review by Lima and Dill 1990). Grubb and Green- saving would become less important when Tshade ap- wald (1982), for instance, interpreted the patch selec- proaches the lower critical temperature, Tlc (recall that tion patterns found in House Sparrows (Passer do- the highest model temperatures recorded during the mesticus) as re¯ecting a compromise between the dis- study period, roughly 33ЊC in full sun, were lower than tance to cover that may represent a source of predators the critical thermal maximum expected for small pas- (farther is safer) and the energy savings afforded by serines; Calder and King 1974). Thus, the advantage cover (closer is warmer). Similar trade-offs between of sun seeking should be higher under lower environ- thermoregulation and predation risk have been postu- mental temperatures. lated in lizards (J. A. DõÂaz 1992, MartõÂn and LoÂpez We hypothesize that solar radiation should exert a 1999), salamanders (Holomuzki 1986), and ®sh (Fi- striking in¯uence on patch selection by treecreepers scher et al. 1987). (and, ultimately, on their distribution and abundance) Fig. 4 also shows that our data do not perfectly ®t through its thermal effect on the bird's environment. the predictions of the predation risk hypothesis, be- For our three-year sample of birds, over 95% of the cause treecreepers did not restrict themselves only to readings of Tshade were below 21.5ЊC. This means that sunlit patches at temperatures below Tlc Ϫ⌬T. In fact, only rarely would some shaded locations fall within the slope of the observed negative regression between the thermoneutral zone. In this scenario, what is re- electivity for sunlit patches and temperature was less markable is not that the electivity of birds for sunlit steep than predicted. We can explain such discrepancy patches was greater at lower temperatures, but that by considering that, in the early morning, when the sun birds actively selected to forage in the shade at tem- is low and shadows are long, temperatures are low but peratures as low as ϳ6.5ЊC. These temperatures are there is a scarcity of trunk surfaces exposed to sunlight. well below the expected Tlc for Certhia brachydactyla This could increase the travel costs faced by treecree- (21.5ЊC; Mozetich 1995) and, indeed, for most holarc- pers when ¯ying to a sunlit trunk. Such interpretation tic passerines in winter (Kendeigh et al. 1977). Assum- is supported by the fact that the percentage of time ing that water losses do not increase with temperature spent ¯ying by focal treecreepers was negatively cor- within the thermoneutral zone (Weathers and Greene related with Tshade (r ϭϪ0.476, P ϭ 0.003, n ϭ 36), 1998), the ``only metabolic bene®ts'' hypothesis pre- which suggests higher travel costs when electivity for dicts that, between lower and upper critical tempera- sunlit patches is higher. An alternative explanation is tures, the use of sun±shade patches should be propor- that the increased length of ¯ying times at low tem- tional to their availability (i.e., electivity ϭ 0; Fig. 1C). peratures could reduce thermoregulatory costs (Web- Alternatively, the ``trade-off with predation risk'' hy- ster and Weathers 1990). pothesis predicts a shade-seeking behavior when tem- In summary, we interpret the selection of shade as pertaure in the shade equals Tlc (i.e., electivity ϳϪ1; a behavioral mechanism for enhancing crypsis. Al- Fig. 1F). Thus, the observed shade-seeking behavior though treecreepers were dif®cult to detect because of rejects the ``only metabolic bene®ts'' hypothesis, be- their cryptic coloration, our results show that crypsis cause it re¯ects a cost to the use of sunlit trunks that was much more effective in shaded patches than in must be large enough to counterbalance the energy sav- sunlit ones. Increased detectability would force birds ings afforded by sunlit patches. to ignore the thermal bene®ts of sun-exposed locations. Our results strongly suggest that this cost results Moreover, the selection for shaded sites, and thus cryp- from the increased risk of predation experienced by sis enhancement, increased as the metabolic advantage June 2001 THERMOREGULATION AND CRYPSIS IN BIRDS 1653 of sun selection became smaller when operative tem- creeper predators. This study was funded by DGICYT project peratures approached T . Thus, our results exemplify PB95-0102-C02-01. This paper is a contribution to the ``El lc Ventorrillo'' Field Station of the Museo Nacional de Ciencias the importance of the predator detection subcomponent Naturales (CSIC). of risk (Lima and Dill 1990) not only in promoting the evolution of morphological adaptations, but also in LITERATURE CITED shaping behavioral decisions that modulate the short- Avery, M. I., and J. R. Krebs. 1984. Temperature and for- aging success of great tits Parus major hunting for spiders. term effectiveness of crypsis (Bauwens and Thoen Ibis 126:33±38. 1981, Dill and Fraser 1984). Austin, G. T. 1976. Behavioral adaptations of the Verdin to In temperate latitudes during winter, possibilities for the desert. Auk 93:245±262. minimizing costs of thermoregulation by selecting sun- Bakken, G. S. 1989. Arboreal perch properties and operative temperature experienced by small . Ecology 70: lit patches are scarce because of the few clear days 922±930. available. Geographical variation in winter cloudiness Bakken, G. S., W. R. Santee, and D. J. Erskine. 1985. Op- de®nes the background where the interactions that we erative and standard operative temperature: tools for ther- have studied may be ecologically relevant for this spe- mal energetics studies. American Zoologist 25:933±943. cies. Cloudiness varies strikingly with latitude, from Bauwens, D., and C. Thoen 1981. Escape tactics and vul- nerability to predation associated with reproduction in the 35±40% of the 9±10 h of winter daylight in southern lizard Lacerta vivipara. Journal of Animal Ecology 50:733± Europe to 70% of less than5hofdaylight at latitudes 743. higher than 50±55Њ (W. Cramer, personal communi- Bednekoff, P. A., H. Biebach, and J. R. Krebs. 1994. Great cation). Within this wide geographical range, many Tit fat reserves under unpredictable temperatures. Journal of Avian Biology 25:156±160. passerine species are winter residents in areas with high Blem, C. R. 1990. Avian energy storage. Current Ornithology cloudiness (for example, the Brown Creeper Certhia 7:59±113. familiaris at northern latitudes), which implies that Bouskila, A. 1995. Interactions between predation risk and cloudiness does not necessarily prevent winter resi- competition: a ®eld study of kangaroo rats and snakes. dence. Therefore, the applicability of the thermoreg- Ecology 76:165±178. Calder, W. A., and J. R. King. 1974. Thermal and caloric ulatory behaviors described in this paper cannot be relations of birds. Pages 259±413 in D. S. Farner and J. generalized to any environmental condition or geo- R. King, editors. Avian biology. Volume 4. Academic graphical location. They could play an important role Press, New York, New York, USA. favoring winter energy savings in southern latitudes of Carlson, A., and J. Moreno. 1992. Cost of short ¯ights in the Willow Tit measured with doubly-labeled water. Auk cold continental and/or montane climate, with low 109:389±393. cloudiness and under windless conditions (Walsberg Carmi-Winkler, N., A. A. Degen, and B. Pinshow. 1987. Sea- 1986, Wolf and Walsberg 1996). We also predict that sonal time±energy budgets of free-living Chukars in the such behaviors should be more important in species Negev Desert. Condor 89:594±601. absent from high latitudes in winter, which tend to be Carrascal, L. M. 1987. Relaciones entre avifauna y estructura de la vegetacioÂn en el PaõÂs Vasco atlaÂntico. Dissertation. less tolerant of cold stress than their northern coun- Universidad Complutense, Madrid, Spain. terparts. This is probably the case for the Short-toed Carrascal, L. M. 1988. In¯uencia de las condiciones am- Treecreeper, with a northern limit in its distribution bientales sobre la organizacioÂn de la comunidad de aves range that does not reach 55ЊN (10±13Њ below the en un bosque subalpino mediterraÂneo. DonÄana, Acta Ver- tebrata 15:111±131. northern limit of Certhia familiaris). An interesting Carrascal, L. M., P. Lopez, J. Martin, and A. Salvador. 1992. extension of this argument is that thermoregulation Basking and antipredator behaviour in a high altitude liz- might play a prominent role in explaining the impor- ard: implications of heat-exchange rate. Ethology 92:143± tance of southern mediterranean latitudes as winter 154. quarters for many European species. Thus, the large Carrascal, L. M., J. Potti, and F. J. SaÂnchez-Aguado. 1987. Spatio-temporal organization of the bird communities in winter densities of small passerines in mediterranean two Mediterranean montane forests. Holarctic Ecology 10: areas (Carrascal et al. 1998, TellerõÂa et al. 1999) could 185±192. also be explained by adding thermoregulation bene®ts Carrascal, L. M., J. C. Senar, I. Mozetich, F. Uribe, and J. to the food availability hypothesis, especially if win- Domenech. 1998. Interactions among environmental stress, body condition, nutritional status, and dominance in tering mediterranean populations are not limited by Great Tits. Auk 115:727±738. food availability. Clark, C. W., and D. A. Levy. 1988. Diel vertical migrations by juvenile sockeye salmon and the antipredation window. ACKNOWLEDGMENTS American Naturalist 131:271±290. Alvaro RamõÂrez gave us valuable ®eld assistance. Martinito Clarke, J. A. 1983. Moonlight's in¯uence on predator/prey (L. M. Carrascal's six-month-old son), while enjoying the day interactions between short-eared owls (Asio ¯ammeus) and out, remained obligingly quiet while we sampled in the Val- deermice (Peromyscus maniculatus). Behavioral Ecology sain±La Granja forest. The de®nition of our model was sub- and Sociobiology 13:205±209. stantially improved by the suggestions of Vicente Polo and Cooper, R. J., and R. C. Whitmore. 1990. Arthropod sampling Luis M. Bautista. Two anonymous reviewers provided useful methods in ornithology. Studies in Avian Biology 13:29± recommendations. Roberto Carbonell, E. de Juana, J. A. Del- 37. gado, M. DõÂaz, E. FernaÂndez, M. FernaÂndez-Cruz, J. T. Cramp, S., and C. M. Perrins, editors. 1993. The birds of GarcõÂa, F. J. GarcõÂa,D.LoÂpez, J. PeÂrez-Tris, T. Santos, J. L. the Western Palearctic. Volume VII. Oxford University TellerõÂa, E. VirgoÂs, and J. GonzaÂlez acted as ef®cient tree- Press, London, UK. 1654 LUIS M. CARRASCAL ET AL. Ecology, Vol. 82, No. 6

De Jong, A. A. 1976. The in¯uence of simulated solar ra- thermoregulation: orientation toward the sun in herring diation on the metabolic rate of White-crowned Sparrows. gulls. Science 200:81±83. Condor 78:174±179. Lustick, S., S. Talbot, and E. L. Fox. 1970. Absorption of DõÂaz, J. A. 1992. Choice of compass directions around shrub radiant energy in Red-winged Blackbirds (Agelaius phoen- patches by the heliothermic lizard Psammodromus algirus. iceus). Condor 72:471±473. Herpetologica 48:293±300. MartõÂn, J., and P. LoÂpez. 1999. An experimental test of the DõÂaz, M. 1992. Rodent seed predation in cereal crop areas costs of antipredatory refuge use in the wall lizard, Po- of central Spain: effects of physiognomy, food availability, darcis muralis. Oikos 84:499±505. and predation risk. Ecography 15:77±85. Moreno, J., A. Carlson, and R. V. Alatalo. 1988. Winter en- Dill, L. M., and A. H. G. Fraser. 1984. Risk of predation and ergetics of coniferous forest tits, Paridae, in the north: the the feeding behavior of juvenile coho salmon (Oncorhyn- implications of body size. Functional Ecology 2:163±170. chus kisutch). Behavioral Ecology and Sociobiology 16: Mozetich, I. M. 1995. EcologõÂa Invernal de Passeriformes 65±71. en ambientes forestales mediterraÂneos. Dissertation. Univ- Endler, J. A. 1987. Predation, light intensity and courtship ersidad AutoÂnoma, Madrid, Spain. behaviour in Poecilia reticulata (Pisces: Poeciliidae). An- Ohmart, R. D., and R. C. Lasiewski. 1971. Roadrunners: imal Behaviour 35:1376±1385. energy conservation by hypothermia and absorption of sun- Fischer, R. U. Jr., E. A. Standora, and J. R. Spotila. 1987. light. Science 172:67±69. Predator-induced changes in thermoregulation of bluegill, Pianka, E. R. 1986. Ecology and natural history of desert Lepomis macrochirus, from a thermally altered reservoir. lizards: analyses of the ecological niche and community Canadian Journal of Fisheries and Aquatic Sciences 44: structure. Princeton University Press, Princeton, New Jer- 1629±1634. sey, USA. Repasky, R. R. 1991. Temperature and the northern distri- Grubb, T. C. Jr., and L. Greenwald. 1982. Sparrows and a butions of wintering birds. Ecology 72:2274±2285. brushpile: foraging responses to different combinations of Root, T. 1988. Energy constraints on avian distributions and predation risk and energy cost. Animal Behaviour 30:637± abundances. Ecology 69:330±339. 640. Stephens, D. W., and J. R. Krebs. 1986. Foraging theory. Hagemeijer, W. J. M., and M. J. Blair, editors. 1997. The Princeton University Press, Princeton, New Jersey. EBCC atlas of European breeding birds. Their distribution Tatner, P., and D. M. Bryant. 1986. Flight cost of a small and abundance. Poyser, London, UK. passerine measured using doubly labeled water: implica- Hamilton, W. J., and F. Heppner. 1967. Radiant solar energy tions for energetics studies. Auk 103:169±180. and the function of black homeotherm pigment. Science TellerõÂa, J. L., B. Asensio, and M. DõÂaz. 1999. Aves IbeÂricas. 155:196±197. Volume 2. Passeriformes. J. M. Reyero Editor, Madrid, HedenstroÈm, A. 1992. Flight performance in relation to fuel Spain. load in birds. Journal of Theoretical Biology 158:535±537. Vitt, L. J., and S. S. Sartorius. 1999. HOBOs, Tidbits and Hogstad, O. 1988. Social rank and antipredator behaviour of lizard models: the utility of electronic devices in ®eld stud- willow tits Parus montanus in winter ¯ocks. Ibis 130:45± ies of ectotherm thermoregulation. Functional Ecology 13: 56. 670±674. Holomuzki, J. R. 1986. Predator avoidance and diel patterns Walsberg, G. E. 1983. Avian ecological energetics. Pages of microhabitat use by larval tiger salamanders. Ecology 161±220 in D. S. Farner, J. R. King, and K. C. Parkes, 67:737±748. editors. Avian biology. Volume 7. Academic Press, New Huey, R. B. 1991. Physiological consequences of habitat se- York, New York, USA. lection. American Naturalist 137:91±115. Walsberg, G. E. 1986. Thermal consequences of roost-site Ivlev, V. S. 1961. Experimental ecology of the feeding of selection: the relative importance of three modes of heat ®shes. Yale University Press, New Haven, Connecticut, conservation. Auk 103:1±7. USA. Watanuki, Y. 1986. Moonlight avoidance behavior in Leach's Kendeigh, S. C., V. R. Dol'nik, and V. M. Gavrilov. 1977. Storm-Petrels as a defense against Slaty-backed Gulls. Auk Avian energetics. Pages 127±204 in J. Pinowski and S. C. 103:14±22. Kendeigh, editors. Granivorous birds in ecosystems. Cam- Weathers, W. W., W. A. Buttemer, A. M. Hayworth, and K. bridge University Press, New York, New York, USA. A. Nagy. 1984. An evaluation of time-budget estimates of Kotler, B. P., Y. Ayal, and A. Subach. 1994. Effects of pred- daily energy expenditure in birds. Auk 101:459±472. Weathers, W. W. and E. Greene. 1998. Thermoregulatory atory risk and resource renewal on the timing of foraging responses of Bridled and Juniper Titmice to high temper- activity in a gerbil community. Oecologia 100:391±396. ature. Condor 100:365±372. Kotler, B. P., J. S. Brown, R. J. Smith, and W. O. Wirtz. 1988. Webster, M. D., and W. W. Weathers. 1990. Heat produced The effects of morphology and body size on rates of owl as a by-product of foraging activity contributes to ther- predation on desert rodents. Oikos 53:145±152. moregulation by Verdins, Auriparus ¯aviceps. Physiolog- Lima, S. L., and L. M. Dill. 1990. Behavioral decisions made ical Zoology 63:777±794. under the risk of predation: a review and prospectus. Ca- Wolf, B. O., and G. E. Walsberg. 1996. Thermal effects of nadian Journal of Zoology 68:619±640. radiation and wind on a small bird and implications for Lustick, S., M. Adam, and A. Hinko. 1980. Interaction be- microsite selection. Ecology 77:2228±2236. tween posture, color, and the radiative heat load in birds. Wolfe, J. L., and C. T. Summerlin. 1989. The in¯uence of Science 208:1052±1053. lunar light on nocturnal activity of the old-®eld mouse. Lustick, S., B. Battersby, and M. Kelty. 1978. Behavioral Animal Behaviour 37:410±414.