We analysed 92,828 records gathered from throughout the United Kingdom between 1939 and 1995 by the British Trust for Ornithology's Nest Record Scheme (see Supplementary Information). Annual median laying dates for 36 species, for which we had at least 1,000 records each, were used to measure shifts in the timing of whole nesting seasons. In most cases, the upper and lower tails of the distributions shifted in parallel to the medians, but the latter were less subject to random variation when annual samples were relatively small. Of these 36 species, 19 (53%) exhibited significant long-term trends, most of which were curvilinear, with laying dates becoming later in the 1960s and 1970s, and then earlier through the 1980s and 1990s (Fig. 1). Trends towards earlier laying in recent years are due to increased numbers of early nests and were never due solely to fewer records of late nests.

Figure 1: Weather and laying dates of the chaffinch, Fringilla coelebs.
figure 1

a, Temporal changes in annual median laying date. b, Temporal changes in the mean of March and April monthly mean Central England Temperatures (CETs). c, Relation between annual median laying date and mean of March and April CETs (r =−0.76, P <0.001). Laying date is numbered such that day 110 is 20 April and day 121 is 1 May. The smoothed lines are calculated using a LOWESS (locally weighted scatterplot smoother) method.

This data set does not allow us to distinguish first clutches from subsequent ones, or from renests after failure, and species might extend their nesting seasons as a consequence of earlier nesting opportunities. Although this study is therefore statistically conservative, and so might underestimate trends towards earliness (and the effect of weather on laying dates), only the robin Erithacus rubecula (which has several broods a year) showed evidence of significant increases in the incidence of early and late nests, using a measure of the width of the two tails of the distribution.

We investigated relationships of median laying date with temperature and rainfall using the Central England Temperature (CET)5 and the England and Wales Precipitation (EWP)6 records, which are broadly representative of UK weather7. We used forward stepwise regression to identify the mean monthly temperure and precipitation variables that accounted for variation in median laying date, weighted by sample size. Weather variables were entered only for months in which laying had been recorded and for the previous month. Significant CET or EWP effects were found for 31 species (86% of the total); the five species without effects were among those with relatively small average annual sample sizes, which may have limited the statistical power. Given the number of associations between month and weather that were tested for significance, some results might be due to chance effects, although more than half the species exhibited weather relationships at P <0.001 and only six species were solely dependent upon effects at P >0.01 (see Supplementary Information).

Of 19 species with long-term trends in laying date, 17 species (89%) exhibited significant weather effects (Fig. 1). The key weather variables were the temperatures in March and April, which show a strong curvilinear trend since the 1940s, having cooled in the 1960s and 1970s (Fig. 1). When trend variables (year, year2 or year3, as appropriate) were included in the weather models for 7 (41%) of these 17 species, trend variables were no longer significant. Thus, variation in CET and EWP was suf-ficient to account for these long-term trends, describing, on average, 29% more of the variation in laying date than the trend models alone. These seven species (winter wren, Troglodytes troglodytes; dunnock, Prunella modularis; blackcap, Sylvia atricapilla; willow warbler, Phylloscopus trochilus; spotted flycatcher, Muscicapa striata; long-tailed tit, Aegithalos caudatus; and greenfinch, Carduelis chloris) are all widespread in Britain, and it is likely that their records provide the best match with the weather data. For a further ten species, trend variables had additional effects to weather variables, accounting for, on average, 16% more than the weather models alone.

These residual trends (and those of the two species without any CET or EWP effects) could be due to the choice of weather variables (for example, the temperature of the soil, rather than the air, might have been more appropriate for some species), or because birds may respond to environmental factors (such as food supply) for which CET and EWP are to some extent surrogates; in addition, the residual trends may be unrelated to climate (for example, large-scale changes in land use might have affected some farmland birds, and magpies may nest earlier because of their rapid expansion into urban areas).

There appear to be no obvious taxonomic, ecological or life-history correlates of the species that are more or less affected by weather. Within each of the groups identified in the Supplementray Information there are, allowing for the size of the groups, mixtures of both residents and migrants, insectivores and granivores, and single- and multibrooded species. The only possible correlate might be body size, as the small number of larger-bodied species (corvids and water birds) appear to show little response to temperature.

We have used the ‘medium-high’ UKCIP98 climate scenario4 for Britain to predict how laying dates for each species may change by 2080, based on the (linear) coefficients of temperature and precipitation. For the 27 species that are predicted to lay earlier, the average advancement is 8 days, ranging up to 18 days (see Supplementary Information). The significance of these findings for nesting success and the subsequent survival of young needs now to be investigated, especially with respect to the possibility of mistimed reproduction in relation to food supplies2,8.