Trophic level asynchrony in rates of phenological change for marine, freshwater and terrestrial environments


Stephen J. Thackeray, tel. +44 01524 595852, fax +44 01524 61536, e-mail:


Recent changes in the seasonal timing (phenology) of familiar biological events have been one of the most conspicuous signs of climate change. However, the lack of a standardized approach to analysing change has hampered assessment of consistency in such changes among different taxa and trophic levels and across freshwater, terrestrial and marine environments. We present a standardized assessment of 25 532 rates of phenological change for 726 UK terrestrial, freshwater and marine taxa. The majority of spring and summer events have advanced, and more rapidly than previously documented. Such consistency is indicative of shared large scale drivers. Furthermore, average rates of change have accelerated in a way that is consistent with observed warming trends. Less coherent patterns in some groups of organisms point to the agency of more local scale processes and multiple drivers. For the first time we show a broad scale signal of differential phenological change among trophic levels; across environments advances in timing were slowest for secondary consumers, thus heightening the potential risk of temporal mismatch in key trophic interactions. If current patterns and rates of phenological change are indicative of future trends, future climate warming may exacerbate trophic mismatching, further disrupting the functioning, persistence and resilience of many ecosystems and having a major impact on ecosystem services.


Recent climate warming has dramatically altered the distribution, abundance and population dynamics of many aquatic and terrestrial organisms (IPCC, 2007). Acting in concert with other ecological stressors, climatic change is projected to weaken the resilience of many ecosystems, leading to sudden re-organizations of communities and drastic alterations to ecosystem structure and function (Scheffer & Carpenter, 2003). Recorded changes in the seasonal timing, or phenology, of life history events have formed a key part of this assessment, with spring and summer events generally becoming earlier as temperatures have risen. Species-specific variation in phenological responses to climate can disrupt the synchrony of ecological interactions (Harrington et al., 1999; Visser & Both, 2005) and potentially affect community persistence. To assess the potential risk of phenological asynchrony affecting community structure, it is crucial to determine how rates of phenological change vary among species, functional groups and trophic levels.

Assessments of broad-scale taxonomic variation in phenological change have generally involved meta-analyses in which previously published results on changes in the timing of biological events are extracted from the literature and combined to provide overall rates of change. While there has been broad unanimity that spring events in the northern hemisphere have become earlier, with estimated mean advances ranging from 2.3 days per decade (Parmesan & Yohe, 2003) to 5.5 days per decade (Root et al., 2003), there has been a lack of consensus on which taxonomic groups show the most, and least, change. Differences in the data selection criteria adopted by different meta-analyses have been cited as a contributory factor to this lack of consistency (Englund et al., 1999; Parmesan, 2007). Reliance on published studies creates two further potential problems. Firstly, positive publication bias (Gurevitch et al., 2001) leads to the under-representation of taxa showing little or no phenological change. Researchers have sought to overcome this by restricting analyses to multispecies studies that include nonresponding species (Parmesan & Yohe, 2003), though there is no assurance that such assemblages are representative of the wider community. Secondly, despite the fact that rates (and even directions) of phenological change are known to vary over time (Crick & Sparks, 1999; Thackeray et al., 2008), the lengths and periods of phenological time series used in meta-analyses are rarely standardized (Badeck et al., 2004; Parmesan, 2007). Thus, apparent taxonomic differences in rates of change can be generated by differences in time series start year, end year and length, rather than by biologically relevant characteristics.

To date analyses overcoming such problems by using raw time series data and consistent time scales rather than published values have been mainly restricted to species comparisons within major taxonomic or functional groups, such as plankton (Edwards & Richardson, 2004), terrestrial plants (Fitter & Fitter, 2002; Menzel et al., 2006), insects (Roy & Sparks, 2000; Stefanescu et al., 2003) and birds (Cotton, 2003; Both & Artemyev, 2004). Here, we apply this approach across environments to obtain the first directly comparable estimates of phenological change for UK terrestrial, freshwater and marine taxa between 1976 and 2005. This has been a period of major environmental change with factors such as eutrophication (including atmospheric deposition of nutrients), acid rain, land-use change, overfishing and, of course, climate change impacting on natural systems (Smart et al., 2003; Millenium Ecosystem Assessment, 2005).

A number of functional traits might be expected to influence the rate of phenological change demonstrated by organisms. It might be expected that the mode of thermal physiology would be particularly influential. Developmental rates of ectotherms are, by definition, directly affected by habitat warming (Precht et al., 1973). One might hypothesize therefore that they have shown the most rapid mean rates of phenological advance. It has been argued that generation time should affect evolutionary rates in response to selective pressures imposed by environmental change (Rosenheim & Tabashnik, 1991; Berteaux et al., 2004) and so we might expect this trait to affect rates of phenological change, assuming a role for microevolution. At the scale of single ecosystems, differential phenological change among trophic levels has frequently been observed (Harrington et al., 1999; Visser & Both, 2005) though it is not clear whether mean rates of change differ significantly among trophic levels at larger scales of ecological organization (Blenckner & Hillebrand, 2002), or whether these patterns are idiosyncratic. Also, choice of phenological metric (Miller-Rushing et al., 2008; Van Buskirk et al., 2009) may affect observed rates of change.

The aims of this study were to (i) provide a fully standardized and unbiased assessment of rates of phenological change in the United Kingdom and (ii) to resolve differences in rates of change among environments and major taxonomic and functional groups. We explicitly considered the following functional attributes; thermal physiology, generation time, environment (where each species feeds/gathers resources during the focal phenological event) and trophic level. The latter attribute allowed us to compute average rates for each trophic level and thus evaluate, across the three environments, likely risks of trophic mismatching i.e. reduced synchrony between the seasonal occurrence of food resources and the peak demand for those resources.

Materials and methods

Quantifying rates of change

All rates of change were calculated from the raw observational data for each nonoverlapping 10-year period between 1976 and 2005, and for the whole 30-year period.

We selected only records indicating the beginning or middle of phenological events, occurring on average between January and August. Thus, our emphasis was on spring and summer events rather than those occurring in autumn or winter. All phenological indicators were the standard metrics used by each of the monitoring programmes represented in the analysis (supporting information, Table S1). Linear regression was used to quantify phenological trends over each specified time period. Annual seasonal timing (day of year) of phenological events was analysed as a function of year by linear regression. Slope coefficients quantified rates of change in days per year. Regressions were performed for each nonoverlapping 10-year period between 1976 and 2005 and for the whole 30-year period. Trends were only retained for meta-analysis if phenological data were available for 80% or more of the years within the respective time period. Considering all taxa, sites and time periods this yielded 25 532 rates of phenological change for 726 distinct taxa (supporting information, Table S2) including terrestrial plants, freshwater and marine phytoplankton and zooplankton, insects, amphibians, birds, fish and mammals (supporting information, Table S2). Phenological records included dates of flowering and leafing, plankton population growth, insect flight periods, births and migration. The analysis included data from across the whole United Kingdom, spanning the latitudinal range 49.9–60.9°N.

Categorization of phenological trends

For the meta-analysis, phenological trends were categorized by major taxonomic group (plants/phytoplankton, invertebrates, vertebrates) and by functional attributes: environment (marine, freshwater, terrestrial), trophic level (primary producer, primary consumer, secondary consumer), thermal physiology (ectothermic, endothermic) and generation time (<1 week, 1 week–1 month, 1 month–1 year, >1 year). Previous studies have shown a tendency for first events to shift more rapidly than middle or peak events (Miller-Rushing et al., 2008; Van Buskirk et al., 2009), and so we also included this dichotomy as a factor in our analyses. Species that used more than one major environment (e.g. amphibians, dragonflies, seabirds) were assigned to the environment with which they were most closely associated for feeding, during the focal phenological event.

Differences in time series lengths among monitoring programmes

Some of the monitoring programmes represented could only provide phenological data for a single decade from the 30-year period (supporting information, Table S1). It is possible that among-decade differences in phenological change were a statistical artefact of having a different assemblage of monitoring programmes represented in each decade. We therefore calculated mean rates of change separately for more finely resolved taxonomic groups that were each represented by a single monitoring programme in all three decades. Patterns of phenological change present in these more consistently recorded taxa were used to confirm patterns of change observed in the whole dataset.

Summarizing rates of change

For each major taxonomic group and habitat type, both the mean rate of change and variability in rates of change were calculated. To test whether trends have on balance shifted either towards earliness or lateness, two-tailed binomial tests were used to test the null hypothesis that trends were symmetrically distributed about zero. Differences in trend variability among groups were assessed using Levene's tests, after Anderson–Darling tests showed that data were not normally distributed.

Linear mixed effects modelling

Linear mixed effects models were used to test for significant differences in mean rates of change among taxonomic groups, and with organism functional traits. This was carried out in sas 9.1 using proc mixed. The following random effects were specified: data source, species nested in data source, site nested in data source, and brood/generation nested in data source. The collated datasets, each of which focussed on a particular taxonomic or functional group, varied in size. By including data source as a random effect, each dataset was given equal weight in the analysis so that results were not unduly influenced by any one, well represented, group. In addition, this explicitly allowed nonindependence among trends from closely related organisms, recorded using the same sampling protocols. Data points (i.e. regression slopes) were weighted by inverse variance. Tested fixed effects were major taxonomic group, metric type (beginning or middle event) and organism functional attributes. Denominator degrees of freedom for fixed effects were estimated using Satterthwaite's approximation. All significance levels given are from type III tests. Figures show marginal least-squares means ± 1 SE.


Widespread phenological change across environments

Rapid climate change occurred over the study period: all three environments experienced very similar warming trends (0.04–0.05 °C yr−1), with temperatures frequently above the 1976–2005 mean since the late 1980s (Fig. 1). During this period, the seasonal timing of biological events in all major taxonomic groups in UK terrestrial, freshwater and marine environments advanced, on average, by 0.39 d yr−1 (equivalent to 11.7 days over the whole period). Overall, 83.8% of phenological trends were towards earlier seasonal timing. In all groups the proportion of negative trends (i.e. towards earliness) was significantly greater than expected if trends were symmetrical about zero (Fig. 2a) and the percentage of statistically significant phenological trends exceeded that expected by chance alone (5%) at the P=0.05 significance level.

Figure 1.

 January–August mean temperatures from 1976 to 2005 for the United Kingdom. Presented are UK mean air temperatures (blue line), mean sea surface temperature 10°W–5°E, 50°N–60°N (red line) and mean lake surface water temperatures (brown line). All series presented as anomalies from the 1976 to 2005 mean. Air temperatures provided by the UK Met Office and sea surface anomalies by the UK Met Office Hadley Centre. Lake temperatures are means of those recorded at the sites included in this study: the northern and southern basins of Windermere, Esthwaite Water and Loch Leven.

Figure 2.

 Phenological change for the UK flora and fauna from 1976 to 2005. (a) percentages of advancing (below horizontal) and delaying (above horizontal) trends for each taxon–environment combination. Statistically significant advancing and delaying trends are indicated by black shading. Nonsignificant trends are indicated by white shading. The number of trends analysed for each taxon-environment combination (n) is given above each bar. Also shown is the significance level (P) of a two-tailed binomial test of the null hypothesis that negative and positive trends are equally likely. (b) Mean ± SEM rates of change for plants/phytoplankton (plant; green bars), invertebrates (invert; yellow bars) and vertebrates (vert; blue bars) in marine, freshwater and terrestrial environments. All mean trends are negative, indicating an advance of phenological events. The taxa included in each taxonomic group-environment combination, and the number of trends per taxon, are given in supporting information, Table S3.

A taxonomically resolved analysis showed that mean rates of change were significantly different among taxonomic groups, though not among environments (linear mixed effects model, LME, n=3419, interaction: F4,71.8=1.47, P=0.22, environment: F2,152=0.34, P=0.71, taxonomic group: F2,73.4=3.58, P=0.033). Leafing, flowering and fruiting dates of terrestrial plants showed the most rapid mean rate of change (0.58 d yr−1, Fig. 2b) and the highest percentage of advancing trends (92.5%, Fig. 2a). Freshwater plants (phytoplankton bloom timings) had the slowest rate of response (0.23 d yr−1 earlier, Fig. 2b) and showed the lowest percentage of advancing trends (62.2%, Fig. 2a). There was also significant heterogeneity of variance in phenological trends among taxa (Levene's test, W8,3411=39.73, P<0.001) with marine and freshwater plants and invertebrates exhibiting much higher variability in rates of change (SD range 0.74–0.98) than other groups (SD range 0.30–0.53).

Decadal variation in rates of change

Examination of phenological changes at a finer temporal resolution indicated that at the decadal scale, trends varied markedly among taxonomic groups and environments (Fig. 3). For many taxa, rates of phenological change have accelerated during recent decades. In general, timing advanced least rapidly before 1986 except for marine plants (phytoplankton bloom timings) and marine vertebrates (migration and spawning/egg laying in marine fish and marine birds). In some cases there was a reversal of phenological trends between the 1976–1985 and 1986–1995 periods i.e. a period of phenological delay was followed by a period of phenological advance (marine and freshwater invertebrates, freshwater plants). For some groups mean rates of advance were consistently higher during 1986–2005 than during 1976–1985 (e.g. freshwater and terrestrial plants, marine and freshwater invertebrates), for others rates of advance declined considerably post 1996 (e.g. terrestrial vertebrates). These results were robust when only analysing taxa that were consistently recorded throughout the whole 30-year period (Table 1). Results based on these taxa confirmed the acceleration of phenological change observed in the full dataset: rates of advancing phenology were again lowest during 1976–1985, with higher rates of advance occurring during 1986–1995 (e.g. butterfly first flight and bird arrival dates) or 1986–2005 (e.g. freshwater plankton and terrestrial plants). Flowering dates and aphid flight dates advanced particularly rapidly, by more than 1.0 d yr−1 (10.0 days per decade) after 1986, as did moth flight dates during 1986–1995. The major exception to this overall trend was marine vertebrates, which showed little phenological change after 1986.

Figure 3.

 Decadal variations in rates of phenological change. Decadal mean ± SEM rates of change for plants/phytoplankton (plant; green bars), invertebrates (invert; yellow bars) and vertebrates (vert; blue bars) in marine, freshwater and terrestrial environments. For each taxon–environment combination, decades appear chronologically: 1976–1985 (1), 1986–1995 (2) and 1996–2005 (3). Negative trends indicate advance, and positive trends delaying, of phenological events. The number of trends analysed for each taxon–environment combination, in each decade, is given above each bar.

Table 1.   Mean ± SEM trends (sample size) for various monitoring schemes and phenological metrics. Trends are summarized for each of the 10-year periods 1976–1985, 1986–1995 and 1996–2005, and overall for the 30-year period 1976–2005
Recording schemePhenological metric1976–1985 (d yr−1)1986–1995 (d yr−1)1996–2005 (d yr−1)1976–2005 (d yr−1)
  1. UKBMS: UK Butterfly Monitoring Scheme, BTO: British Trust for Ornithology, CEH: Centre for Ecology and Hydrology, RIS: Rothamsted Insect Survey, SAHFOS: Sir Alister Hardy Foundation for Ocean Science, UKPN: UK Phenology Network

RIS: AphidsFirst flight0.10 ± 0.18 (114)−2.35 ± 0.18 (114)−1.64 ± 0.19 (114)−0.87 ± 0.04 (114)
5th percentile cumulative annual catch0.53 ± 0.14 (114)−1.51 ± 0.14 (114)−1.28 ± 0.12 (114)−0.59 ± 0.05 (114)
UKBMS: ButterfliesFirst flight−0.58 ± 0.08 (1046)−0.98 ± 0.05 (2279)−0.58 ± 0.03 (5410)−0.42 ± 0.02 (1156)
Mean flight date−0.11 ± 0.05 (1044)−0.85 ± 0.03 (2272)−0.51 ± 0.02 (5391)−0.35 ± 0.01 (1154)
BTO: BirdsMean first egg date0.05 ± 0.10 (51)−0.47 ± 0.09 (63)−0.27 ± 0.11 (55)−0.19 ± 0.03 (56)
County bird reportsFirst arrival−0.26 ± 0.07 (238)−0.64 ± 0.07 (254)−0.10 ± 0.06 (250)−0.31 ± 0.02 (239)
CEH: Freshwater plankton50% max. monthly mean abundance0.42 ± 0.40 (114)−0.13 ± 0.30 (150)−0.60 ± 0.38 (130)−0.23 ± 0.09 (126)
Bloom centre of gravity0.75 ± 0.34 (94)−0.38 ± 0.27 (129)−0.48 ± 0.32 (120)−0.22 ± 0.09 (108)
RIS: MothsDay of median catch0.14 ± 0.09 (39)−1.01 ± 0.13 (39)−0.53 ± 0.09 (38)−0.33 ± 0.04 (39)
SAHFOS: Marine planktonBloom centre of gravity−0.20 ± 0.31 (101)−0.54 ± 0.26 (152)−0.51 ± 0.26 (164)−0.38 ± 0.07 (136)
UKPN: PlantsFirst flowering−0.02 ± 0.07 (270)−1.31 ± 0.09 (300)−1.75 ± 0.22 (105)−0.57 ± 0.07 (50)

Trophic and functional variation in rates of change

Phenological trends varied significantly among trophic levels over the whole 30-year period (LME, n=3419, F2,63.7=4.16, P=0.020). In particular, events associated with secondary consumers advanced less rapidly than those for both primary producers and primary consumers (Fig. 4a). The nonsignificant interaction between trophic level and environment (LME, n=3419, F4,139=0.99, P=0.42) would suggest that this has been a general phenomenon in marine, freshwater and terrestrial ecosystems. Analysis of decadal phenological trends confirmed the difference in rates of change among trophic levels (LME, n=22113, F2,116=7.46, P=0.0009). There was a significant acceleration in overall rates of advance post 1986 (LME, n=22113, F2,16*103=37.7, P<0.0001) but this was less pronounced for secondary consumers than for the remaining trophic levels (LME, trophic level–decade interaction, n=22113, F4,4392=12.9, P<0.0001, Fig. 4b). Both consumer categories showed similar rates of advance post 1986. Therefore, the relatively weak acceleration of phenological change for secondary consumers would seem to account for their lower mean rates of change over the whole 30-year period.

Figure 4.

 Mean trends in phenology over the 30-year period 1976–2005, based on linear mixed effects models. (a) Secondary consumers have advanced less than primary producers and primary consumers over the whole period. Patterns of change have varied among decades for major functional groups. (b) Trophic levels. (c) Major ecosystems. (d) Generation time. Values shown are marginal least-squares means ± 1 SE.

Over the 30-year period we found no significant effect of environment (LME, n=3419, F2,152=0.48, P=0.62) and the decadal acceleration of phenological advance was similar across environments (Fig. 4c). There was no significant effect of generation time on rates of changing phenology over the 30-year period (LME, n=3419, F3,33.2=1.92, P=0.15) but variability in rates of change was significantly higher for organisms with a generation time of <1 week (Levene's test, W3,3416=84.09, P<0.001). The effect of generation time interacted with decade (LME, n=22113, F6,7323=16.4, P<0.0001), with the increase in rates of change post 1986 not apparent in organisms with a generation time shorter than 1 week (Fig. 4d).

While mean rates of change indicated a more pronounced advance in ectotherms than endotherms over the 30-year period (0.40 and 0.29 d yr−1 respectively), variation in rates of change was considerable (SD 0.59 for ectotherms, 0.31 for endotherms). There was no significant effect of thermal physiology on rates of change over the whole 30-year period (LME, n=3419, F1,56.4=0.75, P=0.39) or in the decadal analysis (LME, n=22113, F1,121=0.19, P=0.67). Rates of change were more rapid for phenological first events than for middle and peak events in both the 30-year analysis (LME, n=3419, F1,2912=28.0, P<0.0001) and the decadal analysis (LME, n=22113, F1,19*103=41.6, P<0.0001).


Phenological changes and developing asynchronies between trophic pairings (e.g. predators and their prey) have been linked to reductions in individual fitness and declines in the population size of focal species (Platt et al., 2003; Winder & Schindler, 2004; Both et al., 2006; Møller et al., 2008), increasing the risk of population extinctions and biodiversity loss. Such changes are also of great economic and societal importance due to the influence of phenological synchrony on processes such as pollination (Elzinga et al., 2007), fisheries production (Cushing, 1990) and herbivory by agricultural pests (Harrington et al., 2007). Our analysis of the most comprehensive UK phenological data collection to date has shown that the timing of spring and summer events has become earlier for the majority of taxa, and at a more rapid rate than previously reported. The phenomenon is widespread across the terrestrial, freshwater and marine environments and rates of change differ significantly among trophic levels. Our results therefore suggest a heightened risk that phenological asynchrony may disrupt the stability and functioning of aquatic and terrestrial ecosystems, and the delivery of key ecosystem services.

Previous analyses, restricted to taxa showing significant trends in the northern temperate zone, suggest mean rates of advance of 4.4–5.5 days per decade (Root et al., 2003; Root et al., 2005). A recent synthesis (Parmesan, 2007) calculated an ‘unbiased’ estimated mean phenological advance of 2.8 days per decade by analysing published rates of change from taxa showing both significant and nonsignificant phenological trends. Our unbiased, standardized analysis reports a substantially higher rate of advance. This comes contrary to expectation: our avoidance of potentially biased published results, and inclusion of trends from ‘nonresponding’ taxa, should result in a lower mean rate of change. The discrepancy probably arises due to the standardization of time series lengths, and the focus on a period with a consistent warming trend. In previous meta-analyses some taxa were represented over longer time scales, including periods of stable and decreasing mean temperatures. Since warming trends are projected to intensify (IPCC, 2007), we believe our results to be indicative of a period of continuing, possibly increasing, phenological change.

For the first time we have explicitly demonstrated that there has been systematic variation in rates of phenological change among trophic levels, and that this has been apparent across marine, freshwater and terrestrial environments in the United Kingdom. Specifically, phenological advances have been slowest for secondary consumers over the 30-year period. This appears to result from a slower acceleration of phenological advance for this group, compared with the lower trophic levels. Numerous potential trophic interactions exist between the primary and secondary consumers represented in our analysis e.g. amphibians and fish consume both plankton and aquatic invertebrate larvae such as dragonflies and alderflies, adult dragonflies can consume members of the Lepidoptera and Hymeno-ptera, many of the terrestrial bird species provision their chicks on Lepidoptera, Hemiptera and Coleoptera while marine fish and birds utilize larval fish such as sandeels Ammodytes marinus. Our analysis would therefore suggest an increasing likelihood of trophic mismatching. Although developing trophic asynchronies have been observed within single communities (Harrington et al., 1999; Visser & Both, 2005), generalizations regarding rates of change among trophic levels have proven elusive. Patterns of change within food webs have appeared idiosyncratic, with different trophic levels showing the most rapid rates of change in different communities and ecosystems (Visser & Both, 2005). This has prohibited the detection of systematic variation in the effects of climate on phenological change at different trophic levels (Blenckner & Hillebrand, 2002). Our standardized, unbiased analysis has made such detection possible. We also show for the first time that accelerating rates of change, previously identified for single species or small groups of species (Crick & Sparks, 1999; Thackeray et al., 2008), have been coherent across ecosystems and have potentially widened existing trophic asynchronies. The timing of acceleration corresponds with a transition between periods when temperatures were frequently below, and frequently above, the 1976–2005 mean. This suggests that this increase in rates of phenological advance has been a response to climate warming over the 30-year period.

There is much variability in the direction and magnitude of phenological trends. This occurs because taxa/populations (i) differ in the extent to which life history events are able to accelerate with warming, (ii) experience different warming trends due to variations in mean seasonal timing and microhabitat use, (iii) vary in the extent to which their phenological responses are driven/constrained by factors other than increases in temperature and (iv) may respond to changing climate in other ways, e.g. distributional changes. Alternative drivers of phenological change will vary regionally and include atmospheric nutrient deposition and precipitation (Peñuelas et al., 2002; Cleland et al., 2006), nutrient enrichment of aquatic ecosystems (Thackeray et al., 2008) and variations in population size (Langvatn et al., 2004; Miller-Rushing et al., 2008) and age structure (Langvatn et al., 2004; Gillet & Dubois, 2007). Resolution of the mechanisms underpinning variations in rates of change is crucial to projecting impacts of these changes. We adopted a functional approach to this problem, though the sources of variation outlined above and correlations among functional traits render mechanistic understanding a major challenge. We examined expectations that rates of change would be highest for ectotherms and short generation time organisms; attributes which should enhance the potential for phenological advance with warming.

Ectotherms had a higher mean rate of change than endotherms, as might be expected. The phenology of ectotherm life history events will be directly influenced by warming while endotherms will respond more indirectly, via ecosystem effects. The difference will be exacerbated because birds, the most well-represented endotherms in the analysis, demonstrate constrained phenological plasticity due to photoperiodic induction of gonad maturation and migration, especially for long-distance migrants (Dawson, 2008). Despite the apparent difference in rates of change between ectotherms and endotherms, there was considerable variability around the mean rates of change for each thermal physiology category and the difference was not significant. Variability was highest for phytoplankton, which also showed the highest incidences of phenological delays. Despite having the shortest generation times of the analysed taxa, the phytoplankton showed comparatively low mean rates of change over the 30-year period. While individual phytoplankton taxa can show very rapid phenological shifts, great trend variability appears to have prevented mean rates of change from being exceptional. Additionally, although phytoplankton growth is directly affected by temperature, bloom timings reflect population dynamics that are also influenced by light and nutrient availability, grazing and sedimentation (Thackeray et al., 2008) and interspecific interactions that have potentially unpredictable outcomes (Benincàet al., 2008).

In contrast to previous meta-analyses (Parmesan, 2007), we found that terrestrial plants have demonstrated the fastest mean phenological advance and the highest proportion of phenological advances. This perhaps indicates the primacy of temperature as a limiting factor for plant growth at mid northern latitudes (Badeck et al., 2004) and that temperature change has not been great enough to induce delaying effects such as heat stress in summer (Sherry et al., 2007) or insufficient winter chilling (Thompson & Clark, 2008). However, warming is also believed to be the predominant driver of change for other phenological metrics, such as butterfly flight periods (Roy & Sparks, 2000). The rapid and consistent phenological advances for terrestrial plants cannot therefore be explained by the fact that they respond the least to alternative drivers. Plant development may simply accelerate more rapidly with warming than developmental rates for other taxa. It is also possible that, since individual plants are sessile, they cannot modulate the warming they experience by moving among microhabitats. This might result in stronger and more consistent phenological trends for this group.

It is possible that effects of thermal physiology and generation time were masked by correlations among functional traits. Trophic level and thermal physiology are, to an extent, correlated. Endotherms were represented only in the consumer categories, such that the fastest rates of phenological change and acceleration occurred in the only trophic level occupied solely by ectotherms. It is plausible that consumers showed lower rates of advance and acceleration because of the influence of endotherms, but that the independent effect of thermal physiology could not be resolved in the analysis. Generation time covaries strongly with trophic level in aquatic ecosystems, generally increasing higher up the food web, but it would seem that this trait had a limited influence on rates of change. The taxonomic group demonstrating the most rapid change, terrestrial plants, did not have the shortest generation times. Furthermore, evidence for rapid microevolution in response to warming is equivocal, with phenotypic plasticity playing a dominant role in phenological adjustment (Berteaux et al., 2004; Charmantier et al., 2008). Interestingly, the expected decrease in rates of change with increasing generation time appears to be maintained post 1986, during the period of most rapid overall phenological change, and for organisms with generation times in excess of 1 week. Given the correlations among functional traits, a major challenge for phenology research is to determine the traits responsible for rates of change and to establish the ecological scale at which independent effects of these traits can be disentangled.

Climate change has been identified as one of the most critical threats to the maintenance of global biodiversity. Our findings indicating earlier occurrence of many spring and summer biological events across a wide range of terrestrial, freshwater and marine taxa should aid current and future assessment of the impacts of climate change. Although the precise consequences of phenological changes on ecosystem functioning are not clear, our study highlights the potential risk of desynchronizing trophic linkages between primary and secondary consumers. Such disruption could have catastrophic environmental consequences and affect delivery of important ecosystem services, particularly food production. Given the broad taxonomic and environmental scales at which our analyses were conducted, it is reasonable to expect that our findings may generalize to other mid latitude regions and have relevance at a global level.


We thank the following for collecting and allowing access to data: Miss Ursula Allitt, Belfast HSC Trust, Cardiff School of Health Sciences UWIC, Prof. Jean Emberlin, Environmental and Public Protection Offices Islington, the Freshwater Biological Association, Lancashire Immunology Service (Royal Preston Hospital), the National Dormouse Monitoring Programme, University of Plymouth School of Geography. We thank the following organizations for funding: BBSRC, Countryside Council for Wales, Defra, Forestry Commission, Joint Nature Conservation Committee (JNCC), The Lawes Agricultural Trust, Natural England, NERC, Northern Ireland Environment Agency, Scottish Natural Heritage. The Nest Record Scheme is supported by the JNCC/BTO partnership. We thank Suzanne Clark and Peter Rothery for assistance with coding statistical models, and Björn Beckmann for assistance with data extraction. We are indebted to the many thousands of volunteer recorders who have contributed records to many of the datasets we have analysed. The authors have no conflicts of interest to declare. We thank Prof. Mark Bailey and Dr Alex Elliott for their comments on an earlier version of the manuscript. This analysis was funded by the Centre for Ecology and Hydrology Environmental Change Integrating Fund project SPACE (Shifting Phenology: Attributing Change across Ecosystems).