The effects of four decades of climate change on the breeding ecology of an avian sentinel species across a 1,500‐km latitudinal gradient are stronger at high latitudes

Abstract Global warming affects breeding phenology of birds differentially with latitude, but there is contrasting evidence about how the changing climate influences the breeding of migrating songbirds at their northern breeding range. We investigate the effect of climate warming on breeding time and breeding success of European pied flycatchers Ficedula hypoleuca in Sweden during a period of 36 years using nest reports from bird ringing. To account for the latitudinal variation, we divided Sweden into three latitudinal bands (northern, intermediate, and southern). We applied a sliding window approach to find the most influential period and environment characteristics (temperature, vegetation greenness, and precipitation), using linear mixed models and model averaging. Our results show a long‐term advancement of breeding time related to increasing spring temperature and vegetation greenness during a period before hatching. Northern breeders revealed a larger advancement over the years (8.3 days) compared with southern breeders (3.6 days). We observed a relatively stronger effect of temperature and greenness on breeding time in the north. Furthermore, northern birds showed an increase in breeding success over time, while birds breeding at southern and intermediate latitudes showed reduced breeding success in years with higher prehatching temperatures. Our findings with stronger environment effects on breeding time advancement in the north suggest that pied flycatchers are more responsive to weather cues at higher latitudes. Breeding time adjustment and, potentially, low competition help explain the higher long‐term success observed in the north. Reduced breeding success at more southerly latitudes suggests an inability to match breeding time to very early and warm springs, a fate that with continued climate change could also be expected for pied flycatchers and other long‐distance migrants at their very northern breeding range.


| INTRODUC TI ON
Evidence of the impact of recent climate change on phenology of animals and plants is ample today (Both et al., 2006;Dunn & Winkler, 1999;Hällfors et al., 2020;Helm et al., 2019;Parmesan & Yohe, 2003;Root et al., 2003;Saino et al., 2011;Schwartz et al., 2006;Shave et al., 2019;Townsend et al., 2013), but the underlying mechanisms of the changing patterns are poorly understood (Knudsen et al., 2011), especially considering the variation of climate change with latitude (Parmesan, 2007). In birds, phenological shifts, like advancement of breeding time due to warmer and earlier springs, may facilitate larger clutches and the opportunity to rear more young (Dunn, 2004). However, phenological changes may also have negative consequences on breeding success and population trends when, for example, sudden changes in weather (Dunn, 2004;Visser et al., 2015) or ecological mismatch occur (Both et al., 2006;Harrington et al., 1999;Saino et al., 2011;Sanz et al., 2003). Thus, in the face of rapid climate change, birds may experience reduced breeding success that can result in population decline (Both et al., 2006;Sanz et al., 2003), but if there is time for adaptive behaviors to emerge, populations could adapt to the novel environments (Charmantier et al., 2008;Helm et al., 2019;Visser et al., 2015). In fact, microevolution on breeding time advancement as a consequence of climate change has been proven in pied flycatchers in Germany (Helm et al., 2019).
Phenological response to climate change has been found to be stronger at higher latitudes due to a more intense effect of global warming toward the north, with temperature increase as the main part of the mechanism (Intergovernmental Panel on Climate Change, IPCC, 2001;Root et al., 2003;Shukla et al., 2019). Still, some studies at northern sites experienced more stable spring and summer temperatures during the last decades (Ram et al., 2019) or even temperature decrease (Both et al., 2004;Both & Marvelde, 2007) finding nonsignificant trends on breeding time change in several bird species. Nevertheless, studies on the effect of climate change on breeding time and breeding success over a large latitudinal gradient are scarce (Parmesan, 2007; but see examples in Dunn & Winkler, 1999;Hällfors et al., 2020;Meller et al., 2018;Shave et al., 2019) and need further investigation.
In seasonal habitats, insectivorous birds generally time their breeding season to coincide with high food availability to match their offspring's needs and thus maximize breeding success (Van Noordwijk et al., 1995). As springs become earlier with climate change, some species have adjusted their breeding time, matching the advancement of local-insect peak (Both et al., 2006;Charmantier et al., 2008;Emmenegger et al., 2014). In particular, Afro-Palearctic long-distance migrants, that fly thousands of km from Tropical Africa, breeding at mid-latitude forests of Europe may need more generation time than resident birds to adjust breeding to earlier food peaks at breeding areas; to avoid ecological mismatch (Both et al., 2010;Saino et al., 2011;Sanz et al., 2003). Nevertheless, it is unclear whether the underlying mechanism of long-term avian productivity decrease observed in the UK is directly related to trophic asynchrony (Franks et al., 2017). Furthermore, in boreal forest of Europe, warm spring temperature was positively associated with breeding success suggesting no mismatch between chick food demands and food availability (Meller et al., 2018). Moreover, the tendency to breed early can be triggered by intraspecific competition, rather than trophic synchrony, as suggested for some populations of willow tits (Pakanen et al., 2016) and pied flycatchers (Ahola et al., 2012) in Finland.
The European pied flycatcher has become a model species when studying the effects of climate change on the annual cycle of longdistance migratory birds (Ahola et al., 2004(Ahola et al., , 2012Both et al., 2006;Both & Visser, 2001;Cadahía et al., 2017;Helm et al., 2019;Sanz et al., 2003;Tomotani et al., 2018;Visser et al., 2015). In general, its breeding time has advanced during the last decades in central and western Europe, presumably driven by an increase in spring temperature (Both et al., 2004;Both & Visser, 2001;Helm et al., 2019;Tomotani et al., 2018). At more northern sites, there are reports of less of a change in breeding time (Ahola et al., 2004;Both et al., 2004;Källander et al., 2017;Ram et al., 2019), but a recent study showed earlier date of egg laying over the last three decades that closely associated with birch leafing and higher May temperature (Järvinen, 2020). In southern and central Europe, several populations of pied flycatchers that did not advance their breeding time despite the increase in spring temperatures had lower breeding success (Both et al., 2006;Sanz et al., 2003), while populations that adapted to earlier breeding did not decline (Both et al., 2006). However, early breeding can cause lower adult survival during possible subsequent cold springs, especially after long-distance migration (Visser et al., 2015) or due to emergent competition effects (Ahola et al., 2007;Samplonius & Both, 2019). Nevertheless, contrasting results are typically assigned to different rates of climate change between latitudes and study periods (Parmesan, 2007;Visser et al., 2015), and there are also population and habitat-specific differences (Vatka et al., 2011). In addition, differences in the methods used may add to the variability of results observed (Parmesan, 2007).
Although the combination of warm spring temperature and high precipitation will trigger primary productivity, measurements of seasonal vegetation greenness give a more direct information on vegetation development, in a given space and time (Gordo, 2007;Pettorelli et al., 2005). Moreover, several studies indicate association between vegetation greenness and spring phenology advancement as well as insectivore food resources (Fernández-Tizón et al., 2020;Jørgensen et al., 2016;Thorup et al., 2017;Vega & Thorup, 2018; but see Rybinski et al., 2016;Veen et al., 2010).
Altogether, a set of environmental variables such as temperature, precipitation, and vegetation greenness can be used to disentangle the mechanism of climate change effects on breeding phenology and reproduction.
Here, we investigate the influence of spring advancement on timing of breeding and reproductive success of the pied flycatcher breeding in Sweden during a period of 36 years. Our study is unique in covering a large span of latitudes (the whole range of Sweden with more than 1,500 km) and over a long time span, allowing us to investigate differences in the effects of climate change depending on latitude in the northern breeding range of small songbirds (but see other studies covering latitudinal gradients : Dunn & Winkler, 1999and Shave et al., 2019for North America, and Meller et al., 2018and Hällfors et al., 2020. Sweden has since the 1980s experienced higher year-round temperature and earlier and longer vegetation periods, especially in the north where the vegetation period lengthened by two weeks during 40 years (Swedish Meteorological andHydrological Institute, SMHI, 2021b, 2021d). Being a migratory bird, breeding time and success of pied flycatchers might be affected by environmental factors during migration, but we chose to restrict our analyses to study effects of environmental conditions at the breeding ground.
We used standardized nest reports collected all over Sweden with information on ringing date of chicks, number of chicks alive, and developmental stage of the chicks per nest. As environmental variables, we used temperature, precipitation and a remotesensing measurement of primary productivity, the normalized difference vegetation index (NDVI), measured all over Sweden, giving us the opportunity to better understand the geographical differences in breeding response to climate change found to date (Both & Marvelde, 2007;Källander et al., 2017;Ram et al., 2019;Samplonius et al., 2018), and shed light on the mixed results found for Scandinavia. The potential of NDVI use in this kind of studies has been generally overlooked. To investigate the mechanism of breeding response to global warming at a regional scale, we use NDVI as an estimate of seasonal vegetation development (and consequently spring advancement; see, e.g., Emmenegger et al., 2016;Jørgensen et al., 2016;Kristensen et al., 2013;Thorup et al., 2017 for studies on a more global scale). We quantified the relationship between the environmental variables and breeding time, and the effect of these on reproductive success at different latitudes over Sweden. Within a model-averaging framework and sliding window approach, we identify important variables and period of the year influencing breeding time and success.
We predict earlier breeding time in relation to increasing spring temperatures and vegetation greenness throughout the study period since warmer conditions and higher vegetation greenness (the latter influenced by more precipitation early in spring) enable an earlier onset of breeding. However, we hypothesize that more precipitation near the breeding season could delay breeding time since it would trigger lower temperature and insect availability in a short term. Furthermore, we test whether breeding success of pied flycatchers has been retained, suggesting a sufficient adjustment of breeding time in relation to climate change. In our analysis, we test the hypothesis of the effect of the changing climate in a long term on breeding time and success with the interaction between year and the three environmental variables respectively, thus giving information on the magnitude of climate change effect over time. In addition, we analyze the individual effect of the environment variables to grasp the effects due to their yearly variation, thus regardless of the time trend. In terms of latitudinal variation of climate change, we predict stronger response to climate change toward northern Sweden where, overall, the rate of climate change has been larger.
Since spring arrival is later at higher latitudes, we accounted for latitudinal effects in our analyses to avoid confounding effects on the ecological responses to climate change. In the same way, we took into account the possible effect of the sea, given that spring phenology is generally earlier at the coast, by using a distance to the coast variable. TA B L E 1 Chick developmental stages used to calculate age of nestlings, and corresponding age range derived from pictures of nestlings of known age (e.g., in Figure 1) Research Station, in south of Sweden in 2018 (e.g., in Figure 1). We calculated the hatching day-of-year of each clutch by subtracting the mean age in days for their developmental stage from the ringing date (Table 1). To study breeding success, we used the number of ringed chicks per nest and considered this a surrogate of breeding success up to ringing time. We did not use data with stage E (newly fledged, 100 nests) since we wanted to record breeding success (as number of ringed chicks in the nest), or nest data with code X (mixed stages in the same nest, 270 nests).

| Data collection and processing
Environmental data consisting of weather and vegetation greenness for the study period were included in the analyses as proxies of climate since these factors closely vary with climate change (IPCC, 2001;Pettorelli et al., 2005), and biological systems generally have capacity to adjust to environmental conditions (Root et al., 2003;Saino et al., 2011). Regional reanalysis weather data were collected from the Copernicus Climate Change Service (https://cds. clima te.coper nicus.eu/cdsap p#!/datas et/reana lysis -uerra -europ esingl e-level s?tab=overview) that provides modeled data of weather observations from the SMHI and statistical simulations. As weather variables, we used surface-air average and minimum temperature at 2-m height above the surface and total precipitation, at a spatial resolution of 121 km 2 (UERRA-HARMONIE system). As a measurement of primary productivity, we used NDVI from Global Inventory Modeling and Mapping Studies (GIMMS; NDVI3g version 1.1; Pinzon & Tucker, 2014) at a spatial resolution of 85.6 km 2 . NDVI is a remotesensing measurement that indicates levels of photosynthetic activity, and it is frequently represented in a scale between 0 and 1, with higher values corresponding to increasing vegetation greenness.
NDVI is widely used as a measure of vegetation phenology and indirect indication of food resources that vary with climate conditions (Pettorelli et al., 2005;Sanz et al., 2003;Thorup et al., 2017). It is usual to find a small fraction of not available NDVI data because of atmospheric conditions (e.g., cloud cover and aerosols) even though efforts are dedicated to minimize this (Pinzon & Tucker, 2014). Since the smallest available resolution for the GIMMS NDVI data is biweekly, we also used this temporal resolution for temperature and precipitation. The geographical location of the nest boxes was given by a variable named 'location ID' that contained the same longitude and latitude coordinates for several nest boxes, and there were 1,077 location IDs in the final data set ( Figure 2). We obtained the environment data from netCDF files using the R package ncdf4 and extracted values at the corresponding nest location ID and biweekly periods using the R package raster. After filtering out NDVI and chick developmental stage NAs (around 1,500 and 7,370 data points, respectively), the final data set contained 29,035 nest reports.
We controlled for the latitude effect in two ways: dividing the south because of earlier spring phenology, and, consequently, breeding success would be affected. Distance to the (nearest) coast was calculated for all nest locations using a shapefile of coastlines of Sweden. Due to mixed effects linked to topography (e.g., mountains in the north and the two coastlines in the south), we transformed this variable into a factor with three levels fitting our analysis with a general measure of the influence of the sea. The three levels were the closest to the coast (level 1), the intermediate (level 2), and the most inland (level 3). For the analysis at the southernmost latitudinal band, a distance to the coast factor with two levels substituted the three-level factor given the small amount of data points at level 3.

| Statistical analysis
To study the effect of climate change on breeding time and breeding success throughout the study period, we used the individual nest data to do model-averaging analyses with linear mixed-effect models. We also calculated the change in annual-average hatching dayof-year, breeding success, and seasonal environmental conditions over time with linear regressions to understand their general trends.  (Tables S1 and S2).
Akaike weights (relative importance), confidence intervals, and coefficients of model-averaged parameters were calculated per latitudinal band. We used the environment at the most influential periods to show the annual-average trends.
We calculated Moran's Index on the best-model residuals confirming that there was no spatial autocorrelation (p ≥ .05).

Collinearity between independent variables was assessed with
Pearson's coefficients (threshold ≤0.7), variance inflation factor for breeding time analysis (VIF; threshold ≤2.5), and using univariate model comparison for breeding success analysis (Dormann et al., 2013). The pairwise correlation of all our model terms within a particular model was below the mentioned correlation thresholds; thus, we considered that there were not highly correlated. Models with minimum temperature were not used in the final analysis since they were less supported than models using mean temperature. We used the R packages lme4 for mixed-effect models and MuMIn for model averaging. All statistical analyses were carried out in R 3.5.2 64-bit (R Core Team, 2019).

| Environmental parameters
At the southern and intermediate latitudinal bands, environmental means during May (average of 30 and 15 days before mean hatching time; Figure 3) had a higher impact on the breeding time and breeding success than the other periods analyzed (averaged models with the environment during May had lower AIC; Table 2 and Table S1).
At the northern latitudinal band, environmental means from May to first half of June had more effect on breeding time (Table 2 and   Table S1). However, in the best models of the northern band, the environment at the periods analyzed did not explain variation of breeding success (Table 3 and Table S2). Thus, the most influential period was from May until mean hatching time for the three latitudinal bands. We will from here on call this period 'pre-hatching period' and refer to the environmental conditions during this period as prehatching temperature/NDVI/precipitation.   Figure 5a). In accordance with this, the model averaging showed a high relative importance of year (w i = 1) at the three latitudinal bands indicating earlier breeding during the study period, but the year effect was weaker at the northern band (Table 2). Furthermore, the model averaging showed a high relative importance of prehatching temperature at all latitudinal bands indicating earlier breeding with higher temperatures and with the strongest effect at the northern band (Table 2; estimates from −0.28 to −0.43; Figure 6a). The interaction between temperature and year had high relative importance at all latitudinal bands, with positive estimates indicating that the effect of the temperature depended on the year and its effect diminished lightly over time (  Figure 3. 'Full' average method was used. Parameters with w i < 0.2 are not included. a Distance to the coast compares relative effects in relation to the variable levels; for example, distance to the coast (2-1) shows the effect of level 2 compared with level 1.

| Effects on breeding success
One of the most important variables associated with breeding success across all latitudinal bands was hatching day-of-year, indicating that earlier breeding birds had higher reproductive success (Table 3 and Figure 7). At the northern band, pied flycatchers increased annual-average breeding success throughout the study period with  (Table 3). In addition, at the southern band we found a relatively high effect of latitude suggesting an increase in success toward the north (Table 3). Finally, across all latitudinal bands, the younger the chicks were when ringed, the higher breeding success was found (more chicks were alive per nest; Table 3; chick age).

| Climate change over the study period
Overall Although the temperature increase found during the prehatching period is lower than expected at high latitudes (Root et al., 2003;SMHI, 2021bSMHI, , 2021d, earlier vegetation development over the years suggests overall warmer conditions in the north as well. Cumulative effects of warmer temperature and precipitation during earlier or longer periods of the year than measured in this study may have influenced the more intense change observed in vegetation greenness (Schwartz et al., 2006).

| Effects on breeding time
Our study shows that pied flycatchers breeding in Sweden advanced  Note: Model-averaged parameters with estimates, standard errors (SE), confidence intervals (CI), and Akaike weights (w i ). Results per latitudinal band where the period from May until mean breeding time (the so-called prehatching period) was more supported based on lowest AIC of competing models. A & B refer to 15-day periods in relation to mean hatching period as in Figure 3. may be more responsive to weather cues at higher latitudes, as they face more stochastic weather conditions and are more time constrained due to shorter breeding seasons (Conklin et al., 2010;Järvinen & Väisänen, 1984). Our results are in line with a long-term study in North America where it was shown that purple martins Progne subis lay eggs earlier in warmer springs and this effect was stronger toward the north (Shave et al., 2019). The fact that there is less human activity influencing the dynamics of vegetation greenness in the north of Sweden may make the greenness index a better measurement of spring phenology in the north compared with more southern latitudes (Pettorelli et al., 2005). Our observation of advanced breeding time with increasing temperature in northern Sweden contrasts with previous long-term studies of pied flycatchers in the northern distribution range that found stable (Kluen et al., 2017;Nyholm, 2011;Ram et al., 2019) or later (Both et al., 2004;Laaksonen et al., 2006) breeding time over the years. Different rates of climate change depending on latitude and study period (Both & Marvelde, 2007;Visser et al., 2015) as well as the geographical scale under study (Laaksonen et al., 2006;Ram et al., 2019) may influence the response found on breeding time.
Compared with the previous studies, we sampled more scattered locations over Sweden, thus maybe representing a larger variation of the pied flycatcher breeding situation at the northern breeding range. Our findings for southern Sweden are, however, supported by studies of more southerly populations of pied flycatchers that showed earlier spring migration and egg laying in connection with temperature increase (Both et al., 2004;Cadahía et al., 2017;Hüppop & Winkel, 2006).  (Gullefors, 2008), where precipitation increases from coast to inland (Greiser et al., 2020;SMHI, 2021a).
We can thus expect that pied flycatchers breeding in this area face more variable habitat in terms of vegetation and climate than at other latitudes due to the transition zone.

| Effects on breeding success
Birds breeding early in the year may have larger clutches and longer seasons to rear their young (Dunn, 2004;Shave et al., 2019;Tomotani et al., 2018), and specifically, early breeding pied flycatchers may lay more eggs (Järvinen, 1989), giving opportunity for a higher reproductive success compared with later breeding birds.
Accordingly, our results show that pied flycatchers that bred early in the season had more chicks in the nest at ringing than later breeders.
Furthermore, number of ringed chicks in the nest, our measurement of breeding success, increased in the north throughout the study period while it did not change at more southern latitudes. Our results with stronger response of breeding time advancement in pied flycatchers at northernmost Sweden may have contributed to the observed increase in breeding success. Northern pied flycatchers did on average increase breeding success with 0.36 chick during the study period of 36 years, indicating that they nowadays perform as the more southerly populations under study or maybe even better.
Perhaps, reproductive success of northernmost populations of pied flycatchers may be favored by higher temperatures as suggested for songbirds breeding in boreal areas in Finland (Meller et al., 2018).
At subarctic areas, harsher climate conditions and a more generalist bird community (Eeva et al., 2000) might result in lower competition for territories and food (Herrera, 1978). This may have contributed to such increase in breeding success in the north, enhanced by the recent milder climate.
Breeding earlier may, however, not always correspond to higher breeding success as other factors like trophic mismatch, extreme weather events (Dunn, 2004;Visser et al., 1998Visser et al., , 2015, competition at high breeding densities (Ahola et al., 2012), and more frequent competitive interactions (Samplonius & Both, 2019) may render lower reproductive success. At southern and mid-latitudes, we found no change in breeding success, despite earlier breeding time in the south. Interestingly, at these latitudes, in accordance with a study of a more southern population of pied flycatchers (González-Braojos et al., 2017), high prehatching temperatures rendered a reduced breeding success. This result suggests that in years with early and warm spring conditions, even early breeding flycatchers might be too late to time the peak of food availability for their chicks (Both et al., 2006). Furthermore, especially during years with warm winters and early springs, it has been shown that resident and short-distance migrants can have an advantage over long-distance migrants when competing for good quality territories (Herrera, 1978). This could thus negatively influence breeding success of pied flycatchers if they have to occupy lower quality territories, as suggested in studies of interactions between tits and flycatchers (Ahola et al., 2007;Samplonius et al., 2018;Samplonius & Both, 2019).
In summary, our study indicates that a long-distance migratory songbird breeding at the north of its range is able to adapt breeding time to climate change by tracking the earlier vegetation greenness over the years despite the long distance between breeding and wintering grounds. Furthermore, birds breeding in the very north of Sweden showed an increase in breeding success and the largest advancement in breeding time over the study period, indicating that migration time schedule does not constrain the breeding time response to spring advancement. However, in southern and midlatitudes of Sweden, negative effects of warm spring temperatures on breeding success suggest an inability to fully adjust breeding time to the prevailing spring conditions at such occasions. The observed stronger response of breeding time and breeding success in the north may relate to a tighter adjustment of breeding time to the environment due to shorter breeding seasons than birds breeding at more southern latitudes in Sweden.
Even if some long-distance migrants can adjust their breeding time to climate change so far and evolutionary changes in spring timing have been shown in pied flycatcher (Helm et al., 2019), a critical question is whether they can breed even earlier in the future. On the basis of rapid global warming projected in the future (Diffenbaugh & Field, 2013), with stronger effects toward higher latitudes in the Northern Hemisphere (IPCC, 2001;Parmesan, 2007;Shukla et al., 2019), there is a high risk that long-distance migrants, especially in their northernmost breeding range, at some point will be unable to fully adjust breeding time in response to climate change. The reduced breeding success in Swedish pied flycatchers at southern and mid-latitudes during warm springs reported in our study could thus be expected to also occur in more northern breeding birds with continued climate change and indicates a fate that could be true for several long-distance migratory species (Radchuk et al., 2019;Visser, 2019).

ACK N OWLED G M ENTS
We deeply acknowledge all volunteer bird ringers for collecting the nest data. The research was funded by the Bolin Centre for Climate Research at Stockholm University (Research Area 8). The authors thank the SMHI and Copernicus Regional Reanalysis for Europe through Semjon Schimanke for providing weather data and the National Aeronautics and Space Administration/Goddard Space Flight Center GIMMS institution through Jorge E. Pinzon for providing NDVI data.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data supporting the study are available at the Dryad Digital Repository https://doi.org/10.5061/dryad.v9s4m w6v4.