Marcel Visser

Variation in age structure influences population dynamics, yet we have limited understanding of the spatial scale at which its fluctuations are synchronised between populations. Using 32 great tit populations, spanning 4° W–33° E and 35°–65° N involving > 130,000 birds across 67 years, we quantify spatial synchrony in breeding demographic structure (subadult vs. adult breeders) and its drivers. We show that larger clutch sizes, colder winters, and larger beech crops lead to younger populations. We report distance-dependent synchrony of demographic structure, maintained at approximately 650 km. Despite covariation with demographic structure, we do not find evidence for environmental variables influencing the scale of synchrony, except for beech masting. We suggest that local ecological and density-dependent dynamics impact how environmental variation interacts with demographic structure, influencing estimates of the environment's effect on synchrony. Our analyses demonstrate the operation of synchrony in demographic structure over large scales, with implications for age-dependent demography in populations.

A major aim of evolutionary biology is to understand why patterns of genomic diversity vary within taxa and space. Large-scale genomic studies of widespread species are useful for studying how environment and demography shape patterns of genomic divergence. Here, we describe one of the most geographically comprehensive surveys of genomic variation in a wild vertebrate to date; the great tit (Parus major) HapMap project. We screened ca 500,000 SNP markers across 647 individuals from 29 populations, spanning ~30 degrees of latitude and 40 degrees of longitude – almost the entire geographical range of the European subspecies. Genome-wide variation was consistent with a recent colonisation across Europe from a South-East European refugium, with bottlenecks and reduced genetic diversity in island populations. Differentiation across the genome was highly heterogeneous, with clear ‘islands of differentiation’, even among populations with very low levels of genome-wide differentiation. Low local recombination rates were a strong predictor of high local genomic differentiation (F_ST), especially in island and peripheral mainland populations, suggesting that the interplay between genetic drift and recombination causes highly heterogeneous differentiation landscapes. We also detected genomic outlier regions that were confined to one or more peripheral great tit populations, probably as a result of recent directional selection at the species' range edges. Haplotype-based measures of selection were related to recombination rate, albeit less strongly, and highlighted population-specific sweeps that likely resulted from positive selection. Our study highlights how comprehensive screens of genomic variation in wild organisms can provide unique insights into spatio-temporal evolutionary dynamics.

Artificial light at night (ALAN) widely affects wildlife by blurring light-dark differences, including transitions such as sunrise and sunset, thereby affecting regulation of diel rhythms. As a result, activity onsets in many wild diurnal songbirds advance under ALAN. From chronobiological studies, it is known that the direction and strength of the response to light depends on when during the night exposure takes place. However, these experiments are mostly done under continuous light conditions, when animals have free-running rhythms. It remains unclear whether phase-dependence also holds in entrained, wild songbirds; i.e., does the effect of ALAN on activity patterns differ between exposure in the morning compared to the evening? This information is essential to assess the effects of mitigation measures by limiting ALAN to selected times of the night. We exposed incubating great tits (Parus major) inside the nest-box to 4 h of dim light, of which 1 h overlapped with dawn before sunrise or dusk after sunset. We found a small advancing effect of morning-light on activity onset and of evening-light on offset compared to dark controls but not vice versa. Breeding success and chick condition were unaffected by the light treatments. However, light-treated females had lower weights 9–18 days after the end of the treatment compared to the controls, independent of whether ALAN occurred in the morning or the evening, indicating possible costs of ALAN. Despite the weak behavioral response, ALAN might have affected the females' circadian clock or physiology resulting in lower body condition.

Our planet endures a progressive increase in artificial light at night (ALAN), which affects virtually all species, and thereby biodiversity. Mitigation strategies include reducing its intensity and duration, and the adjustment of light spectrum using modern light emitting diode (LED) light sources. Here, we studied ground-dwelling invertebrate (predominantly insects, arachnids, molluscs, millipedes, woodlice and worms) diversity and community composition after 3 or 4 years of continued nightly exposure (every night from sunset to sunrise) to experimental ALAN with three different spectra (white-, and green- and red-dominated light), as well as for a dark control, in natural forest-edge habitat. Diversity of pitfall-trapped ground-dwelling invertebrates, and the local contribution to beta diversity, did not differ between the dark control and illuminated sites, or between the different spectra. The invertebrate community composition, however, was significantly affected by the presence of light. Keeping lights off during single nights did show an immediate effect on the composition of trapped invertebrates compared to illuminated nights. These effects of light on species composition may impact ecosystems by cascading effects across the food web. This article is part of the theme issue 'Light pollution in complex ecological systems'.

Mate choice is a key component of reproductive biology. Females often prefer certain males but do females modulate their reproductive investment depending on whether they are mated with their preferred partner? We investigated this question in great tits (Parus major) where we subjected 36 females to a six-choice mate preference test. Male morphological traits and the female's own characteristics did not influence the preference females expressed. We however found that females spent more time near more exploratory males. We then paired females with one of the males in indoor aviaries, and subsequently monitored their reproductive investment (through measurement of plasma 17β-oestradiol concentrations, first egg date, clutch size and egg size). Females that were mated with a male for which they had a strong preference laid their first clutch significantly earlier in the season than females paired with a male they less preferred. Our results show that mate preference influences reproductive investment in great tits, thereby linking mate choice to bird reproductive decisions.

Climate change can severely impact species that depend on temporary resources by inducing phenological mismatches between consumer and resource seasonal timing. In the winter moth, warmer winters caused eggs to hatch before their food source, young oak leaves, became available. This phenological mismatch changed the selection on the temperature sensitivity of egg development rate. However, we know little about the fine-scale fitness consequences of phenological mismatch at the individual level and how this mismatch affects population dynamics in the winter moth. To determine the fitness consequences of mistimed egg hatching relative to timing of oak budburst, we quantified survival and pupation weight in a feeding experiment. We found that mismatch greatly increased mortality rates of freshly hatched caterpillars, as well as affecting caterpillar growth and development time. We then investigated whether these individual fitness consequences have population-level impacts by estimating the effect of phenological mismatch on population dynamics, using our long-term data (1994-2021) on relative winter moth population densities at four locations in The Netherlands. We found a significant effect of mismatch on population density with higher population growth rates in years with a smaller phenological mismatch. Our results indicate that climate change-induced phenological mismatch can incur severe individual fitness consequences that can impact population density in the wild.

Oscine birds preferentially respond to certain sounds over others from an early age, which focuses subsequent learning onto sexually relevant songs.^1,2,3 Songs vary both across species and, due to cultural evolution, among populations of the same species. As a result, early song responses are expected to be shaped by selection both to avoid the fitness costs of cross-species learning⁴ and to promote learning of population-typical songs.⁵ These sources of selection are not mutually exclusive but can result in distinct geographic patterns of song responses in juvenile birds: if the risks of interspecific mating are the main driver of early song discrimination, then discrimination should be strongest where closely related species co-occur.⁴ In contrast, if early discrimination primarily facilitates learning local songs, then it should be tuned to songs typical of the local dialect.^5,6,7 Here, we experimentally assess the drivers of song discrimination in nestling pied flycatchers (Ficedula hypoleuca). We first demonstrate that early discrimination against the songs of the closely related collared flycatcher (F. albicollis) is not strongly affected by co-occurrence. Second, across six European populations, we show that nestlings’ early song responses are tuned to their local song dialect and that responses to the songs of collared flycatchers are similarly weak as to those of other conspecific dialects. Taken together, these findings provide clear experimental support for the hypothesis that cultural evolution, in conjunction with associated learning predispositions, drives the emergence of pre-mating reproductive barriers.

Climate change will strongly affect the developmental timing of insects, as their development rate depends largely on ambient temperature. However, we know little about the genetic mechanisms underlying the temperature sensitivity of embryonic development in insects. We investigated embryonic development rate in the winter moth (Operophtera brumata), a species with egg dormancy which has been under selection due to climate change. We used RNA sequencing to investigate which genes are involved in the regulation of winter moth embryonic development rate in response to temperature. Over the course of development, we sampled eggs before and after an experimental change in ambient temperature, including two early development weeks when the temperature sensitivity of eggs is low and two late development weeks when temperature sensitivity is high. We found temperature-responsive genes that responded in a similar way across development, as well as genes with a temperature response specific to a particular development week. Moreover, we identified genes whose temperature effect size changed around the switch in temperature sensitivity of development rate. Interesting candidate genes for regulating the temperature sensitivity of egg development rate included genes involved in histone modification, hormonal signalling, nervous system development and circadian clock genes. The diverse sets of temperature-responsive genes we found here indicate that there are many potential targets of selection to change the temperature sensitivity of embryonic development rate. Identifying for which of these genes there is genetic variation in wild insect populations will give insight into their adaptive potential in the face of climate change.

Natural selection can only occur if individuals differ in fitness. For this reason, the variance in relative fitness has been equated with the ‘opportunity for selection’ ((Formula presented.)), which has a long, albeit somewhat controversial, history. In this paper we discuss the use/misuse of (Formula presented.) and related metrics in evolutionary ecology. The opportunity is only realised if some fraction of (Formula presented.) is caused by trait variation. Thus, (Formula presented.) does not imply that selection is occurring, as sometimes erroneously assumed, because all fitness variation could be independent of phenotype. The selection intensity on any given trait cannot exceed (Formula presented.), but this upper limit will never be reached because (a) stochastic factors always affect fitness, and (b) there might be multiple traits under selection. The expected magnitude of the stochastic component of (Formula presented.) is negatively correlated with mean fitness. Uncertainty in realised (Formula presented.) is also larger when mean fitness or population/sample size are low. Variation in (Formula presented.) across time or space thus can be dominated (or solely driven) by variation in the strength of demographic stochasticity. We illustrate these points using simulations and empirical data from a population study on great tits Parus major. Our analysis shows that the scope for fecundity selection in the great tits is substantially higher when using annual number of recruits as the fitness measure, rather than fledglings or eggs, even after adjusting for the dependence of (Formula presented.) on mean fitness. This suggests nonrandom survival of juveniles across families between life stages. Indeed, previous work on this population has shown that offspring recruitment is often nonrandom with respect to clutch size and laying date of parents, for example. We conclude that one cannot make direct inferences about selection based on fitness data alone. However, examining variation in (Formula presented.) (the opportunity for fecundity selection adjusted for mean fitness) across life stages, years or environments can offer clues as to when/where fecundity selection might be strongest, which can be useful for research planning and experimental design.

The rate of adaptive evolution, the contribution of selection to genetic changes that increase mean fitness, is determined by the additive genetic variance in individual relative fitness. To date, there are few robust estimates of this parameter for natural populations, and it is therefore unclear whether adaptive evolution can play a meaningful role in short-term population dynamics. We developed and applied quantitative genetic methods to long-term datasets from 19 wild bird and mammal populations and found that, while estimates vary between populations, additive genetic variance in relative fitness is often substantial and, on average, twice that of previous estimates. We show that these rates of contemporary adaptive evolution can affect population dynamics and hence that natural selection has the potential to partly mitigate effects of current environmental change.

The profiling of epigenetic marks like DNA methylation has become a central aspect of studies in evolution and ecology. Bisulphite sequencing is commonly used for assessing genome-wide DNA methylation at single nucleotide resolution but these data can also provide information on genetic variants like single nucleotide polymorphisms (SNPs). However, bisulphite conversion causes unmethylated cytosines to appear as thymines, complicating the alignment and subsequent SNP calling. Several tools have been developed to overcome this challenge, but there is no independent evaluation of such tools for non-model species, which often lack genomic references. Here, we used whole-genome bisulphite sequencing (WGBS) data from four female great tits (Parus major) to evaluate the performance of seven tools for SNP calling from bisulphite sequencing data. We used SNPs from whole-genome resequencing data of the same samples as baseline SNPs to assess common performance metrics like sensitivity, precision, and the number of true positive, false positive, and false negative SNPs for the full range of variant and genotype quality values. We found clear differences between the tools in either optimizing precision (Bis-SNP), sensitivity (biscuit), or a compromise between both (all other tools). Overall, the choice of SNP caller strongly depends on which performance parameter should be maximized and whether ascertainment bias should be minimized to optimize downstream analysis, highlighting the need for studies that assess such differences.

The phenology of many species shows strong sensitivity to climate change; however, with few large scale intra-specific studies it is unclear how such sensitivity varies over a species’ range. We document large intra-specific variation in phenological sensitivity to temperature using laying date information from 67 populations of two co-familial European songbirds, the great tit (Parus major) and blue tit (Cyanistes caeruleus), covering a large part of their breeding range. Populations inhabiting deciduous habitats showed stronger phenological sensitivity than those in evergreen and mixed habitats. However, populations with higher sensitivity tended to have experienced less rapid change in climate over the past decades, such that populations with high phenological sensitivity will not necessarily exhibit the strongest phenological advancement. Our results show that to effectively assess the impact of climate change on phenology across a species’ range it will be necessary to account for intra-specific variation in phenological sensitivity, climate change exposure, and the ecological characteristics of a population.

The predictability of evolution is expected to depend on the relative contribution of deterministic and stochastic processes. This ratio is modulated by effective population size. Smaller effective populations harbor less genetic diversity and stochastic processes are generally expected to play a larger role, leading to less repeatable evolutionary trajectories. Empirical insight into the relationship between effective population size and repeatability is limited and focused mostly on asexual organisms. Here, we tested whether fitness evolution was less repeatable after a population bottleneck in obligately outcrossing populations of Caenorhabditis elegans. Replicated populations founded by 500, 50, or five individuals (no/moderate/strong bottleneck) were exposed to a novel environment with a different bacterial prey. As a proxy for fitness, population size was measured after one week of growth before and after 15 weeks of evolution. Surprisingly, we found no significant differences among treatments in their fitness evolution. Even though the strong bottleneck reduced the relative contribution of selection to fitness variation, this did not translate to a significant reduction in the repeatability of fitness evolution. Thus, although a bottleneck reduced the contribution of deterministic processes, we conclude that the predictability of evolution may not universally depend on effective population size, especially in sexual organisms.

Globally increasing levels of artificial light at night (ALAN) are associated with shifting rhythms of behaviour in many wild species. However, it is unclear whether changes in behavioural timing are paralleled by consistent shifts in the molecular clock and its associated physiological pathways. Inconsistent shifts between behavioural and molecular rhythms, and between different tissues and physiological systems, disrupt the circadian system, which coordinates all major body functions. We therefore compared behavioural, transcriptional and metabolomic responses of captive great tits (Parus major) to three ALAN intensities or to dark nights, recording activity and sampling brain, liver, spleen and blood at mid-day and midnight. ALAN advanced wake-up time, and this shift was paralleled by advanced expression of the clock gene BMAL1 in all tissues, suggesting close links between behaviour and clock gene expression across tissues. However, further analysis of gene expression and metabolites revealed that clock shifts were inconsistent across physiological systems. Untargeted metabolomic profiling showed that only 9.7% of the 755 analysed metabolites followed the behavioural shift. This high level of desynchronization indicates that ALAN disrupted the circadian system on a deep, easily overlooked level. Thus, circadian disruption could be a key mediator of health impacts of ALAN on wild animals.

Space-based tracking technology using low-cost miniature tags is now delivering data on fine-scale animal movement at near-global scale. Linked with remotely sensed environmental data, this offers a biological lens on habitat integrity and connectivity for conservation and human health; a global network of animal sentinels of environmental change.

Artificial light at night (ALAN) is closely associated with modern societies and is rapidly increasing worldwide. A dynamically growing body of literature shows that ALAN poses a serious threat to all levels of biodiversity—from genes to ecosystems. Many “unknowns” remain to be addressed however, before we fully understand the impact of ALAN on biodiversity and can design effective mitigation measures. Here, we distilled the findings of a workshop on the effects of ALAN on biodiversity at the first World Biodiversity Forum in Davos attended by several major research groups in the field from across the globe. We argue that 11 pressing research questions have to be answered to find ways to reduce the impact of ALAN on biodiversity. The questions address fundamental knowledge gaps, ranging from basic challenges on how to standardize light measurements, through the multi-level impacts on biodiversity, to opportunities and challenges for more sustainable use.

The introduction of artificial light at night (ALAN) into natural and urbanised landscapes is a known and highly pervasive disruptor of invertebrate communities. However, the effect of variation in intensity and spectra of ALAN on invertebrate communities inhabiting different spatial niches is little understood. Further, the remarkable ability of ALAN to continue to disrupt biodiversity even in chronically illuminated urban landscapes is not often acknowledged. Here, we simultaneously sampled airborne and ground-dwelling invertebrate assemblages under and between urban street lights to explore the effects on community composition and abundance of (a) proximity to decadal (i.e. long-illuminated) nocturnal street lighting and (b) variation in the spectral output of light. The two assemblages responded differently. For airborne invertebrates, night-time abundance doubled, and night-time assemblage composition was significantly different for traps under, compared with between, street lights. These differences in abundance were not affected by street light intensity, and were absent in day samples, suggesting that even weak ALAN may be causing short-term redistribution of nocturnal invertebrates. Further, the abundance (but not composition) effects of ALAN on airborne invertebrates increased when the street lights emitted a higher proportion of short-wavelength light. In contrast, for ground-dwelling invertebrates, we found only marginal effects of proximity and spectrum of lighting on abundance and no effect on assemblage composition. However, more intense street lighting reduced abundance and altered composition at traps both under and between lights. Synthesis and Applications. Public lighting managers must consider ALAN impacts on invertebrate communities not only when introducing ALAN to naïve environments, but also when changing lighting in areas that are highly urbanised and exposed to decades of ALAN. Further, lighting proposals and environmental monitoring of invertebrate communities must take into account the effects on both ground-dwelling and airborne assemblages, as these may respond very differently to the presence, intensity and spectrum of ALAN.

In migratory species, the timing of arrival at the breeding grounds is a life-history trait with major fitness consequences. The optimal arrival date varies from year-to-year, and animals use cues to adjust their arrival dates to match this annual variation. However, which cues they use to time their arrival and whether these cues actually predict the annual optimal arrival date is largely unknown. Here, we integrate causal and evolutionary analysis by identifying the environmental variables used by a migratory songbird to time its arrival dates and testing whether these environmental variables also predicted the optimal time to arrive. We used 11 years of male arrival data of a pied flycatcher population. Specifically, we tested whether temperature and normalized difference vegetation index (NDVI) values from their breeding grounds in the Netherlands and from their wintering grounds in Ivory Coast explained the variation in arrival date, and whether these variables correlated with the position of the annual fitness peak at the breeding grounds. We found that temperature and NDVI, both from the wintering and the breeding grounds, explained the annual variation in arrival date, but did not correlate with the optimal arrival date. We explore three alternative explanations for this lack of correlation. Firstly, the date of the fitness peak may have been incorrectly estimated because a potentially important component of fitness (i.e., migration date dependent mortality en route or directly upon arrival) could not be measured. Secondly, we focused on male timing but the fitness landscape is also likely to be shaped by female timing. Finally, the correlation has recently disappeared because climate change disrupted the predictive value of the cues that the birds use to time their migration. In the latter case, birds may adapt by altering their sensitivity to temperature and NDVI.

Urbanisation is increasing worldwide, and there is now ample evidence of phenotypic changes in wild organisms in response to this novel environment. Yet, the genetic changes and genomic architecture underlying these adaptations are poorly understood. Here, we genotype 192 great tits (Parus major) from nine European cities, each paired with an adjacent rural site, to address this major knowledge gap in our understanding of wildlife urban adaptation. We find that a combination of polygenic allele frequency shifts and recurrent selective sweeps are associated with the adaptation of great tits to urban environments. While haplotypes under selection are rarely shared across urban populations, selective sweeps occur within the same genes, mostly linked to neural function and development. Collectively, we show that urban adaptation in a widespread songbird occurs through unique and shared selective sweeps in a core-set of behaviour-linked genes.

Background: DNA methylation is likely a key mechanism regulating changes in gene transcription in traits that show temporal fluctuations in response to environmental conditions. To understand the transcriptional role of DNA methylation we need simultaneous within-individual assessment of methylation changes and gene expression changes over time. Within-individual repeated sampling of tissues, which are essential for trait expression is, however, unfeasible (e.g. specific brain regions, liver and ovary for reproductive timing). Here, we explore to what extend between-individual changes in DNA methylation in a tissue accessible for repeated sampling (red blood cells (RBCs)) reflect such patterns in a tissue unavailable for repeated sampling (liver) and how these DNA methylation patterns are associated with gene expression in such inaccessible tissues (hypothalamus, ovary and liver). For this, 18 great tit (Parus major) females were sacrificed at three time points (n = 6 per time point) throughout the pre-laying and egg-laying period and their blood, hypothalamus, ovary and liver were sampled.

Results: We simultaneously assessed DNA methylation changes (via reduced representation bisulfite sequencing) and changes in gene expression (via RNA-seq and qPCR) over time. In general, we found a positive correlation between changes in CpG site methylation in RBCs and liver across timepoints. For CpG sites in close proximity to the transcription start site, an increase in RBC methylation over time was associated with a decrease in the expression of the associated gene in the ovary. In contrast, no such association with gene expression was found for CpG site methylation within the gene body or the 10 kb up- and downstream regions adjacent to the gene body.

Conclusion: Temporal changes in DNA methylation are largely tissue-general, indicating that changes in RBC methylation can reflect changes in DNA methylation in other, often less accessible, tissues such as the liver in our case. However, associations between temporal changes in DNA methylation with changes in gene expression are mostly tissue- and genomic location-dependent. The observation that temporal changes in DNA methylation within RBCs can relate to changes in gene expression in less accessible tissues is important for a better understanding of how environmental conditions shape traits that temporally change in expression in wild populations.

Seasonal timing of reproduction is a key life-history trait, but we know little about the mechanisms underlying individual variation in female endocrine profiles associated with reproduction. In birds, 17β-oestradiol is a key reproductive hormone that links brain neuroendocrine mechanisms, involved in information processing and decision-making, to downstream mechanisms in the liver, where egg-yolk is produced. Here, we test, using a simulated induction of the reproductive system through a Gonadotropin-Releasing Hormone (GnRH) challenge, whether the ovary of pre-breeding female great tits responds to brain stimulation by increasing oestradiol. We also assess how this response is modified by individual-specific traits like age, ovarian follicle size, and personality, using females from lines artificially selected for divergent levels of exploratory behaviour. We show that a GnRH injection leads to a rapid increase in circulating concentrations of oestradiol, but responses varied among individuals. Females with more developed ovarian follicles showed stronger responses and females from lines selected for fast exploratory behaviour showed stronger increases compared to females from the slow line, indicating a heritable component. This study shows that the response of the ovary to reproductive stimulation from the brain greatly varies among individuals and that this variation can be attributed to several commonly measured individual traits, which sheds light on the mechanisms shaping heritable endocrine phenotypes.

BACKGROUND: A widely used approach in next-generation sequencing projects is the alignment of reads to a reference genome. Despite methodological and hardware improvements which have enhanced the efficiency and accuracy of alignments, a significant percentage of reads frequently remain unmapped. Usually, unmapped reads are discarded from the analysis process, but significant biological information and insights can be uncovered from these data. We explored the unmapped DNA (normal and bisulfite treated) and RNA sequence reads of the great tit (Parus major) reference genome individual. From the unmapped reads we generated de novo assemblies, after which the generated sequence contigs were aligned to the NCBI non-redundant nucleotide database using BLAST, identifying the closest known matching sequence.

RESULTS: Many of the aligned contigs showed sequence similarity to different bird species and genes that were absent in the great tit reference assembly. Furthermore, there were also contigs that represented known P. major pathogenic species. Most interesting were several species of blood parasites such as Plasmodium and Trypanosoma.

CONCLUSIONS: Our analyses revealed that meaningful biological information can be found when further exploring unmapped reads. For instance, it is possible to discover sequences that are either absent or misassembled in the reference genome, and sequences that indicate infection or sample contamination. In this study we also propose strategies to aid the capture and interpretation of this information from unmapped reads.

A 31-year study of pied flycatchers shows that it is the temperature at arrival when the offspring return to breed up to two years later that drives selection on breeding time.

Climate change has differentially affected the timing of seasonal events for interacting trophic levels, and this has often led to increased selection on seasonal timing. Yet, the environmental variables driving this selection have rarely been identified, limiting our ability to predict future ecological impacts of climate change. Using a dataset spanning 31 years from a natural population of pied flycatchers (Ficedula hypoleuca), we show that directional selection on timing of reproduction intensified in the first two decades (1980–2000) but weakened during the last decade (2001–2010). Against expectation, this pattern could not be explained by the temporal variation in the phenological mismatch with food abundance. We therefore explored an alternative hypothesis that selection on timing was affected by conditions individuals experience when arriving in spring at the breeding grounds: arriving early in cold conditions may reduce survival. First, we show that in female recruits, spring arrival date in the first breeding year correlates positively with hatch date; hence, early-hatched individuals experience colder conditions at arrival than late-hatched individuals. Second, we show that when temperatures at arrival in the recruitment year were high, early-hatched young had a higher recruitment probability than when temperatures were low. We interpret this as a potential cost of arriving early in colder years, and climate warming may have reduced this cost. We thus show that higher temperatures in the arrival year of recruits were associated with stronger selection for early reproduction in the years these birds were born. As arrival temperatures in the beginning of the study increased, but recently declined again, directional selection on timing of reproduction showed a nonlinear change. We demonstrate that environmental conditions with a lag of up to two years can alter selection on phenological traits in natural populations, something that has important implications for our understanding of how climate can alter patterns of selection in natural populations.