Best environmental predictors of breeding phenology differ with elevation in a common woodland bird species

Abstract Temperatures in mountain areas are increasing at a higher rate than the Northern Hemisphere land average, but how fauna may respond, in particular in terms of phenology, remains poorly understood. The aim of this study was to assess how elevation could modify the relationships between climate variability (air temperature and snow melt‐out date), the timing of plant phenology and egg‐laying date of the coal tit (Periparus ater). We collected 9 years (2011–2019) of data on egg‐laying date, spring air temperature, snow melt‐out date, and larch budburst date at two elevations (~1,300 m and ~1,900 m asl) on a slope located in the Mont‐Blanc Massif in the French Alps. We found that at low elevation, larch budburst date had a direct influence on egg‐laying date, while at high‐altitude snow melt‐out date was the limiting factor. At both elevations, air temperature had a similar effect on egg‐laying date, but was a poorer predictor than larch budburst or snowmelt date. Our results shed light on proximate drivers of breeding phenology responses to interannual climate variability in mountain areas and suggest that factors directly influencing species phenology vary at different elevations. Predicting the future responses of species in a climate change context will require testing the transferability of models and accounting for nonstationary relationships between environmental predictors and the timing of phenological events.

coal tit (Periparus ater). We collected 9 years (2011-2019) of data on egg-laying date, spring air temperature, snow melt-out date, and larch budburst date at two elevations (~1,300 m and ~1,900 m asl) on a slope located in the Mont-Blanc Massif in the French Alps. We found that at low elevation, larch budburst date had a direct influence on egg-laying date, while at high-altitude snow melt-out date was the limiting factor. At both elevations, air temperature had a similar effect on egg-laying date, but was a poorer predictor than larch budburst or snowmelt date. Our results shed light on proximate drivers of breeding phenology responses to interannual climate variability in mountain areas and suggest that factors directly influencing species phenology vary at different elevations. Predicting the future responses of species in a climate change context will require testing the transferability of models and accounting for nonstationary relationships between environmental predictors and the timing of phenological events.

K E Y W O R D S
budburst, climate change, coal tit, elevation, French Alps, laying date, mountain, snow meltout Tinbergen, & Lessells, 1998). Breeding phenology is often decisive for the reproductive success of many vertebrate species, as birth or hatching dates have to coincide with resource emergence which itself depends on plant phenology. As air temperature at the beginning of spring increases, species tend to give birth or lay eggs earlier in the season (Cresswell & McCleery, 2003;Dunn, 2004) to maintain synchronization between hatching date and peak in food abundance (Cole, Long, Zelazowski, Szulkin, & Sheldon, 2015;Crick, Dudley, Glue, & Thomson, 1997;Visser et al., 2006) Changes in species phenology may therefore lead to cascading effects in the community and negative consequences on species reproductive success (Dunn, 2004).
Our understanding of animal reproductive phenology in temperate mountain ecosystems is poor (Inouye & Wielgolaski, 2013) due to rough topography and strong seasonality, which render fieldwork difficult. Elevation influences phenological events mainly through a temperature decrease of about 0.6°C/100 m (Dumas, 2013;Körner, 2007;Rolland, 2003). However, in high-altitude (or high-latitude) ecosystems, air temperature is not the only forcing climatic factor influencing phenology. Duration and depth of snow cover vary with elevation and topography, and are important parameters influencing mountain species phenology. For example, snow meltout date determines the onset of the growing season for alpine plants (Kudo & Hirao, 2006;Vitasse et al., 2017;Wipf, 2010;Wipf, Stoeckli, & Bebi, 2009). The few studies that explored the effects of snow on bird breeding phenology in high elevation ecosystems (see Hendricks, 2003;Morton, 1978;Pereyra, 2011) showed that in North America, breeding is delayed in years of deep snowpack and late snow melt-out date. Similarly, in the Arctic, it is well established that snow, through depth and melting date, influences the timing of nest initiation of ground-nesting birds (Dickey, Gauthier, & Cadieux, 2008;Green, Greenwood, & Lloyd, 1977;Hildén, 1965;Liebezeit, Gurney, Budde, Zack, & Ward, 2014;Meltofte, Høye, Schmidt, & Forchhammer, 2007 (Soininen et al., 2018;Yates et al., 2018) of models developed within a narrow elevation band.
In this study, we aimed to quantify and understand relationships between environmental variables and the breeding phenology of a common mountain forest bird species the coal tit (Periparus ater), at different elevations on a mountain slope. Coal tits often nest close to the ground and are therefore expected to be dependent on snow conditions, particularly when snowmelt-out date is late, that is, at high elevation. First, we tested whether annual measures of local environmental variables (air temperature, snow melt-out date, and tree phenology) explain interannual variation in the timing of coal tit breeding phenology at two elevations. Second, we investigated how elevation modifies the direct and indirect effects of environmental variables on coal tit breeding phenology. Egglaying date and environmental variables (air temperature, snow melt-out date, and plant phenology) were locally measured every year in the field from 2011 to 2019 at two elevations (~1,300 m and ~1,900 m of elevation) in the Mont-Blanc Massif located in the French Alps.

| Study site
The study was carried out in Loriaz (46°1′N, 6°55′E), a mountain located above the town of Vallorcine, located in the Mont-Blanc Massif, France. Forest occurs from 1,300 m to 1,900 m asl and is mainly composed of Norway spruce (Picea abies) and European larch (Larix decidua). Common birch (Betula pendula), rowan (Sorbus aucuparia), common hazel (Corylus avellana), and European ash (Fraxinus excelsior) are mainly present in the lower part of the forest but are much less abundant than coniferous species.

| Survey of the breeding phenology of coal tits-Database 1
Coal tit (Periparus ater) breeding data spanning 9 years (2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019) were collected as a part of a long-term phenology program of the Research Center for Alpine Ecosystems (CREA Mont-Blanc, Stier et al., 2014Stier et al., , 2016. The coal tit was targeted as a study model as it is a common mountain forest species breeding across wide range of elevations. Coal tits make their nests in a hole in the ground or among tree roots, making them dependent on snow on the ground. From April to June, nest boxes were visited weekly. Tits generally lay one egg a day for approximately 8 days (Perrins, 1965 for great tit Parus major). Therefore, the laying date (the date when the first egg is laid in a nest) was estimated by counting backwards from the date that clutch was observed. Only the first clutch of the year was used in this study to exclude second breeding events (van Noordwijk, McCleery, & Perrins, 1995). We considered that clutches laid within 30 days of the first clutch of the year were first clutch events (van Noordwijk et al., 1995). From 10% to 45% of nest boxes were occupied in different years and elevations. We estimated the date of snow melt-out at each elevation using the temperature sensor located at the ground surface of each station. When snow cover is present, ground surface temperature does not vary and equals 0°C; otherwise, it is positive or negative depending on the air temperature (Gadek & Leszkiewicz, 2010).

| Climate data (air temperature and date of snow melt-out)-Database 2
Date of snow melt-out corresponds to the last day of snowpack. For F I G U R E 1 Location map of the nest boxes, climate stations, and larches surveyed. Larches surveyed at the low elevation site are not presented on the map as they are located 5.8 km away to the south, in Montroc each year and each elevation, the timing of snow melt-out was defined as the first snow-free day after >40-day snow-covered period (~6 weeks) from 1st of September until 31st of August, following Klein, Vitasse, Rixen, Marty, and Rebetez (2016). Given the strong correlation obtained for 2 m height air temperature values between the two highest stations (1,915 m and 1,970 m of elevation), and as there were some missing values for the date of snow melt-out, we used for the snowmelt-out date at the high elevation site (ESM 2) the average date of snowmelt-out by year at the 1,915 m and 1,970 m stations. Snowmelt-out date at the low elevation site was obtained with the station located at 1,340 m.

| Survey of larch budburst-Database 3
As we did not have direct measurement of invertebrate phenology at our study sites, we used larch phenology as a proxy. There is evidence that difference in tree phenological responses may affect bird phenology (Veen et al., 2010), but in our case using . We are aware that the sample size is small at each elevation, but different lines of evidence support that the data provides robust estimate of between-year variation in budburst dates as (a) the same individuals were surveyed every year, and the differences in budburst dates that we observe among years come mainly from environmental variability and less from interindividual variation (ESM 3), (b) the order of budburst dates among individuals does not vary from one year to another (i.e., the individual that opened its buds the earliest one year always opened its buds the earliest (or the same day as the other trees) the other years).

| Data analysis
To estimate temporal trends of mean egg-laying dates at the two elevations, we used a linear model (egg-laying date ~ year*elevation), where "*" denotes an interaction between year as a continuous variable and elevation as a categorical variable (high/low) (i.e., different slopes for the year effect). We calculated elevation-dependent delays for each environmental variable (air temperature, snow meltout date, and larch budburst date), that is, the difference between the high and low elevation mean values. To address our first objective, we used linear models to assess the relationships between each environmental variable and coal tit egg-laying date. We included an interaction term between each variable and elevation to assess elevation-specific effects. For our second objective, we used structural equation modeling (SEM; Shipley, 2002) to estimate the magnitude and significance of hypothesized causal connections between variables (egg-laying date, air temperature, date of snow melt-out, and budburst date), and to identify the most important drivers of mean egg-laying date. In order to detect the elevation-specific direct and indirect effects of the environmental variables involved in egg-laying date, we used multigroup analysis applied to SEM. This method includes the interaction between the environmental variables and elevation (high/low), and indicates whether the effect of each environmental variable varies with elevation. We first considered a model in which all environmental variables (snowmelt, larch, and temperature) could directly influence egg-laying date with a model in which only snow and larch budburst directly influenced egglaying date using the data at two elevations. Indeed, temperature could directly influence egg-laying date (Pereyra, 2011) through its influence on gonadal growth (Jones, 1986) and thermoregulation costs of females (Meijer, Nienaber, Langer, & Trillmich, 1999). We then assessed if air temperature had a direct effect on snow meltout date and larch budburst date (Asse et al., 2018;Migliavacca et al., 2008), and if snow melt-out date had an influence on larch budburst date. We accepted the models if their relative goodness of fit p-values (Chi-square tests) were superior to 0.10 (Grace & Keeley, 2006;Pillar et al., 2013) and selected the model with the lowest AIC, a predictive criterion (Akaike Information Criterion; Claeskens & Hjort, 2008). We then applied multigroup analysis to the selected model ("free" model) by sequentially constraining the coefficients of each path by a single value determined by the entire dataset and refitting the model ("constrained" models). Then, we compared the free and constrained model using a likelihoodratio test. If models were significantly different, it implied that the tested path should not be constrained and was differing between sites. Otherwise, it meant that the constrained model was valid and that a single slope value obtained from the entire dataset is suitable for this tested path. Those steps were done for each path. Finally, if several constrained models were valid (p-value > .10, Grace & Keeley, 2006;Pillar et al., 2013), we selected the one with the lowest AIC which indicates that it is the best predictive model. In addition, the path coefficients of the other valid constrained model with the second lowest AIC were compared with the selected one to assess the robustness of the results.

| RE SULTS
Strong interannual variations were observed for egg-laying date, larch budburst date, snow melt-out date, and mean air temperature, with events always occurring later at high elevation ( Figure 2, Average egg-laying date for the 9 years was negatively correlated with air temperature and positively correlated with larch budburst date and snow melt-out date (Table 2). In addition, larch budburst date was negatively correlated with air temperature, and snow meltout date was negatively correlated with air temperature (Table 2) Table 2). selected the most parsimonious model (i.e., without the direct path F I G U R E 2 Egg-laying date, mean larch budburst date, snow melt-out date, and mean spring temperature variation between 2011 and 2019. 95% confidence intervals are shown for egg-laying date between snow melt-out date and larch budburst date). In addition, this path ("Larch budburst ~ Snow melt-out date") was not significant at both elevations, and the effect sizes were low (LE: β = 0.15 ± 0.08 [SE], HE: β = −0.31 ± 0.33 [SE]).
The valid model involving the constrained path "Larch budburst ~ Temperature" was selected as it had the lowest AIC (Table 3).
Hence, at low elevation, egg-laying date is directly influenced by larch budburst date: When larch budburst date is one day later, then egg-laying date is delayed by the same amount (Figure 3b). At high elevation, snow is the main direct driver of egg-laying date: When the date of snow melt-out is one day later, then egg-laying date is 0.38 days later (Figure 3c). The later the snow melts out, the smaller the delay between snow melt-out date and egg-laying date. Air temperature had an indirect effect on egg-laying date either through snowmelt-out date at high elevation or larch budburst date at low elevation (Figure 3b,c), and based on the SEM, egg-laying date occurs from 3.8 to 3.3 days earlier for every degree (°C) increase, respectively, at low and high elevation. The model including the only  (Table 3). In addition, it explained 60% and 40% of the variance of egg-laying date, respectively, at low and high elevation; while the model with the indirect effect of temperature explained 68% and 64% of the variance of egg-laying date, respectively, at low and high elevation. Hence, the model with an indirect effect of the temperature on egg-laying date through snowmelt-out date and larch budburst date better explain variation of egg-laying date than the model with a single direct effect of temperature. The free model M1, which had the second lowest AIC (Table 3)

| D ISCUSS I ON
In this study, we aimed to understand direct and indirect relationships between environmental variables and the breeding phenology of a common mountain forest bird species, the coal tit (Periparus ater), at two elevations (~1,300 and ~1,900 m). At low elevation, larch budburst date (plant phenology) directly influenced egg-laying date, and at high elevation, snow melt-out date was the main predictor. Lower availability of snow-free ground during years of late snow melt-out may delay bird breeding phenology of ground-nesting species such as the coal tit, and delay the availability of ground plant material useful for nest construction (moss and animal hairs for coal tit nests, Pereyra, 2011;Saracco et al., 2019). Alternatively, at low elevation, to maintain synchronization, later plant phenology (larch budburst date) may delay bird breeding phenology due to later insect emergence (Finn & Poff, 2008;Marshall, Cooper, DeCecco, Strazanac, & Butler, 2002). Here, we cannot determine whether it is plant phenology acting as an environmental cue for coal tit individuals Hinks et al., 2015;Thomas, Bourgault, Shipley, Perret, & Blondel, 2010) or other factors co-occurring with budburst, such as insect emergence, that initiate egg-laying date. Future work should be conducted at a larger spatial scale to better identify how local environment influences bird nesting, as some experimental evidence suggests that plant phenology may not be the main environmental cue for the onset of bird breeding (Schaper, Rueda, Sharp, Dawson, & Visser, 2011;Visser, Silverin, Lambrechts, & Tinbergen, 2002). In addition, a larger sample size of budburst dates, both within species (larch) and across species, could be surveyed to explore the relationship between interindividual and interspecific variation in budburst and breeding phenology. While the relationships found in our study have been independently detected for other bird species (effect of plant phenology at low elevation in tits in Thomas et al., 2010, Hinks et al., 2015and Shutt, Cabello, et al., 2019ef-fect of snow at high elevation in American pipit, Dusky flycatcher and White-crowned sparrows, respectively, in Hendricks, 2003, Pereyra, 2011and Morton, 1978 our study shows that elevation determines which predictors are best (Saracco et al., 2019).
TA B L E 3 AIC, chi-square test difference, and adjustment quality of the free, entirely constrained and sequentially constrained paths in tested model. In addition, our results concur with other studies on woodland passerines indicating that air temperature has a strong causal effect on egg-laying dates (Pereyra, 2011;Phillimore, Leech, Pearce-Higgins, & Hadfield, 2016;Shutt, Burgess, et al., 2019;Shutt, Cabello, et al., 2019;Simmonds, Cole, & Sheldon, 2019;Visser, Holleman, & Caro, 2009 (Parmesan, 2006), a pattern confirmed by the absence of temporal trend of air temperature. However, the strong interannual climate variability led us to hypothesize high plasticity in the breeding phenology of the coal tit in response to these environmental variations, which was already observed at the individual level for the great tit (Charmantier et al., 2008).
According to the scenario RCP4.5 (scenario of long-term global emissions of greenhouse gases which stabilizes radiative forcing at 4.5 W/m 2 in the year 2100), temperature is expected to increase by 2 to 3°C in the Alps by the end of the 21st century, with potentially increased rates of warming at higher elevation (Gobiet et al., 2014;Jacob et al., 2014). Simultaneously, climate models predict that the snow amount and duration will drastically decrease below 1,500-2,000 m of elevation, in particular, due to earlier snow melt-out in spring (Beniston, 2012;Beniston et al. 2018;Beniston, Keller, Koffi, & Goyette, 2003;Castebrunet, Eckert, Giraud, Durand, & Morin, 2014;Gobiet et al. 2014;Verfaillie et al. 2018). Furthermore, models predict increasing rainfall during the winter months as well as higher frequency of heavy precipitation events, which could contribute to a reduced and faster melting snowpack (Jacob et al. 2014).
Given these climate projections, we expect that bird breeding phenology will advance until the end of the 21st century. At our high elevation site (~1,900 m asl), a thinner winter snowpack and earlier spring snow melt-out may lead to earlier onset of breeding.
However, we hypothesize that under a certain threshold of snow melt-out date, this climatic factor will not have as much influence compared with current conditions. Hence, we expect that the relationships between variables of the high-altitude site (i.e., direct snow melt-out date effect on egg-laying date) may move toward the relationships of the low altitude site (i.e., direct larch budburst date effect on egg-laying date). Finally, considering that observed warming is stronger at higher elevations, we expect that the delay in breeding phenology between both elevations might decrease (Vitasse, Signarbieux, & Fu, 2018).
In conclusion, our study demonstrates how the predictors of bird breeding phenology can vary with elevation. Larch budburst date and snow melt-out date are significant direct predictors of egg-laying dates of the coal tit, respectively, at low and high elevation, while temperature has a significant indirect effect in both cases. Variable phenological responses according to elevation raise the question of transferability (Soininen et al. 2018;Yates et al. 2018) of models built at a single elevation as the predictors of phenology at one elevation may not be consistent at higher elevations, or relevant in the context of future climate changes. Additional long-term studies on animal phenology in mountain environments should be conducted in order to better predict how interactions between species will change in the context of climate change in mountain areas, where responses can differ along elevation gradients.

ACK N OWLED G M ENTS
This study has been funded by the AURA Region, Department of
We also warmly thank all the volunteers that helped us to survey the coal tit breeding phenology in the Mont-Blanc Massif during these 9 years.

CO N FLI C T O F I NTE R E S T
None declared.

DATA ACCE SS I B I LIT Y
Data are deposited on Dryad and accessible here: https://datad ryad.